GCC(1)

🦋 GCC(1) — GNU 🦋

📛 NAME

gcc — GNU project C and C++ compiler 💅

---

📜 SYNOPSIS

    [-g] [-pg] [-Olevel]
    [-Wwarn...] [-Wpedantic]
    [-Idir...] [-Ldir...]
    [-Dmacro[=defn]...] [-Umacro]
    [-foption...] [-mmachine-option...]
    [-o outfile] [@file] infile...

okay so this is just the appetizer, bestie — only the most useful options are listed here. the full menu is in the sections below. and g++? it accepts mostly the same options as gcc, so you're covered either way. 🫶

---

🌸 DESCRIPTION 🌸

so when you run gcc, it's not just one thing — it's a whole pipeline: preprocessing, compilation, assembly, and linking. the “overall options” let you pause that process at any intermediate stage. like, -c says “nope, don't run the linker,” and then the output is object files from the assembler. iconic behavior, honestly.

other options get passed on to one or more stages of processing. some control the preprocessor, others the compiler itself. and yet other options control the assembler and linker — but most of those aren't documented here because you rarely need them. fr fr, you can ignore them.

most command-line options work for C programs. if an option is only useful with another language (usually C++), the explanation will say so explicitly. if the description doesn't mention a source language, you can use that option with all supported languages. no cap.

the usual way to run GCC is to call the executable gcc, or machine-gcc when cross-compiling, or machine-gcc-version to run a specific version of GCC. and when you compile C++ programs? bestie, invoke GCC as g++ instead. slay.

the gcc program accepts options and file names as operands. many options have multi-letter names, so you can't just group single-letter options together: -dv is very different from -d -v. periodt.

you can mix options and other arguments. for the most part, the order you use doesn't matter. but order does matter when you use several options of the same kind. for example, if you specify -L more than once, the directories are searched in the order specified. also, the placement of the -l option is significant. so yeah, sometimes order is the main character.

many options have long names starting with -f or with -W — for example, -fmove-loop-invariants, -Wformat and so on. most of these have both positive and negative forms; the negative form of -ffoo is -fno-foo. this manual documents only one of these two forms, whichever one is not the default. iykyk.

some options take one or more arguments, typically separated either by a space or by the equals sign (=) from the option name. unless documented otherwise, an argument can be either numeric or a string. numeric arguments must typically be small unsigned decimal or hexadecimal integers. hexadecimal arguments must begin with the 0x prefix. arguments to options that specify a size threshold of some sort may be arbitrarily large decimal or hexadecimal integers followed by a byte size suffix designating a multiple of bytes — such as kB and KiB for kilobyte and kibibyte, respectively, MB and MiB for megabyte and mebibyte, GB and GiB for gigabyte and gigibyte, and so on. those arguments are designated by byte-size in the following text. for the full listing and explanation of the binary and decimal byte size prefixes, refer to the NIST, IEC, and other relevant national and international standards. 💅

---

✨ OPTIONS ✨

📋 Option Summary

okay bestie, here's the whole rundown, grouped by type. the deep-dive explanations are in the following sections. 🎀

🎀 Overall Options

-c  -S  -E  -o file -dumpbase dumpbase  -dumpbase-ext auxdropsuf
-dumpdir dumppfx  -x language -v  -###  --help[=class[,...]]
--target-help  --version -pass-exit-codes  -pipe  -specs=file
-wrapper @file  -ffile-prefix-map=old=new  -fcanon-prefix-map
-fplugin=file  -fplugin-arg-name=arg -fdump-ada-spec[-slim]
-fada-spec-parent=unit  -fdump-go-spec=file

💖 C Language Options

-ansi  -std=standard  -aux-info filename -fno-asm -fno-builtin
-fno-builtin-function  -fcond-mismatch -ffreestanding  -fgimple
-fgnu-tm  -fgnu89-inline  -fhosted -flax-vector-conversions
-fms-extensions -fpermitted-flt-eval-methods=standard
-fplan9-extensions  -fsigned-bitfields  -funsigned-bitfields
-fsigned-char  -funsigned-char  -fstrict-flex-arrays[=n]
-fsso-struct=endianness

💅 C++ Language Options

-fabi-version=n  -fno-access-control -faligned-new=n
-fno-assume-sane-operators-new-delete -fchar8_t  -fcheck-new
-fconcepts  -fconstexpr-depth=n  -fconstexpr-cache-depth=n
-fconstexpr-loop-limit=n  -fconstexpr-ops-limit=n
-fno-elide-constructors -fno-enforce-eh-specs -fno-gnu-keywords
-fno-immediate-escalation -fno-implicit-templates
-fno-implicit-inline-templates -fno-implement-inlines
-fmodule-header[=kind] -fmodule-only -fmodules
-fmodule-implicit-inline -fno-module-lazy
-fmodule-mapper=specification -fmodule-version-ignore
-fms-extensions -fnew-inheriting-ctors -fnew-ttp-matching
-fno-nonansi-builtins  -fnothrow-opt  -fno-operator-names
-fno-optional-diags -fno-pretty-templates  -frange-for-ext-temps
-fno-rtti  -fsized-deallocation -fstrong-eval-order[=kind]
-ftemplate-backtrace-limit=n -ftemplate-depth=n
-fno-threadsafe-statics  -fuse-cxa-atexit -fno-weak  -nostdinc++
-fvisibility-inlines-hidden -fvisibility-ms-compat
-fext-numeric-literals -flang-info-include-translate[=header]
-flang-info-include-translate-not -flang-info-module-cmi[=module]
-stdlib=libstdc++,libc++ -Wabi-tag  -Wcatch-value  -Wcatch-value=n
-Wno-class-conversion  -Wclass-memaccess -Wcomma-subscript
-Wconditionally-supported -Wno-conversion-null
-Wctad-maybe-unsupported -Wctor-dtor-privacy  -Wdangling-reference
-Wno-defaulted-function-deleted -Wno-delete-incomplete
-Wdelete-non-virtual-dtor  -Wno-deprecated-array-compare
-Wdeprecated-copy -Wdeprecated-copy-dtor
-Wno-deprecated-enum-enum-conversion
-Wno-deprecated-enum-float-conversion -Weffc++
-Wno-elaborated-enum-base -Wno-exceptions -Wextra-semi
-Wno-global-module -Wno-inaccessible-base
-Wno-inherited-variadic-ctor  -Wno-init-list-lifetime
-Winvalid-constexpr -Winvalid-imported-macros -Wno-invalid-offsetof
-Wno-literal-suffix -Wmismatched-new-delete -Wmismatched-tags
-Wmultiple-inheritance  -Wnamespaces  -Wnarrowing -Wnoexcept
-Wnoexcept-type  -Wnon-virtual-dtor -Wpessimizing-move
-Wno-placement-new  -Wplacement-new=n -Wrange-loop-construct
-Wredundant-move -Wredundant-tags -Wreorder  -Wregister
-Wstrict-null-sentinel  -Wno-subobject-linkage  -Wtemplates
-Wno-non-template-friend  -Wold-style-cast -Woverloaded-virtual
-Wno-pmf-conversions -Wself-move -Wsign-promo -Wsized-deallocation
-Wsuggest-final-methods -Wsuggest-final-types  -Wsuggest-override
-Wno-template-body -Wno-template-id-cdtor  -Wtemplate-names-tu-local
-Wno-terminate  -Wno-vexing-parse  -Wvirtual-inheritance
-Wno-virtual-move-assign  -Wvolatile

🦋 Objective-C and Objective-C++ Language Options

-fconstant-string-class=class-name -fgnu-runtime  -fnext-runtime
-fno-nil-receivers -fobjc-abi-version=n -fobjc-call-cxx-cdtors
-fobjc-direct-dispatch -fobjc-exceptions -fobjc-gc -fobjc-nilcheck
-fobjc-std=objc1 -fno-local-ivars
-fivar-visibility=[public|protected|private|package]
-freplace-objc-classes -fzero-link -gen-decls -Wassign-intercept
-Wno-property-assign-default -Wno-protocol -Wobjc-root-class
-Wselector -Wstrict-selector-match -Wundeclared-selector

🔥 OpenMP and OpenACC Options

-foffload=arg  -foffload-options=arg -fopenacc  -fopenacc-dim=geom
-fopenmp  -fopenmp-simd  -fopenmp-target-simd-clone[=device-type]

🌈 Diagnostic Message Formatting Options

-fmessage-length=n -fdiagnostics-plain-output
-fdiagnostics-show-location=[once|every-line]
-fdiagnostics-color=[auto|never|always]
-fdiagnostics-urls=[auto|never|always]
-fdiagnostics-format=[text|sarif-stderr|sarif-file|json|json-stderr|json-file]
-fdiagnostics-add-output=DIAGNOSTICS-OUTPUT-SPEC
-fdiagnostics-set-output=DIAGNOSTICS-OUTPUT-SPEC
-fno-diagnostics-json-formatting -fno-diagnostics-show-option
-fno-diagnostics-show-caret -fno-diagnostics-show-event-links
-fno-diagnostics-show-labels  -fno-diagnostics-show-line-numbers
-fno-diagnostics-show-cwe -fno-diagnostics-show-rules
-fno-diagnostics-show-highlight-colors
-fdiagnostics-minimum-margin-width=width
-fdiagnostics-parseable-fixits  -fdiagnostics-generate-patch
-fdiagnostics-show-template-tree  -fno-elide-type
-fdiagnostics-path-format=[none|separate-events|inline-events]
-fdiagnostics-show-path-depths -fno-show-column
-fdiagnostics-column-unit=[display|byte]
-fdiagnostics-column-origin=origin
-fdiagnostics-escape-format=[unicode|bytes]
-fdiagnostics-text-art-charset=[none|ascii|unicode|emoji]

✨ Warning Options ✨

okay bestie, this is THE section. the whole reason you and gcc are in a situationship. warnings are gcc tapping you on the shoulder like "girl… i saw that." there are an absurd number of them, and they're listed here with zero shame:

• -fsyntax-only
• -fmax-errors=n
• -Wpedantic
• -pedantic-errors
• -fpermissive
• -w
• -Wextra
• -Wall
• -Wabi=n
• -Waddress
• -Wno-address-of-packed-member
• -Waggregate-return
• -Walloc-size
• -Walloc-size-larger-than=byte-size
• -Walloc-zero
• -Walloca
• -Walloca-larger-than=byte-size
• -Wno-aggressive-loop-optimizations
• -Warith-conversion
• -Warray-bounds
• -Warray-bounds=n
• -Warray-compare
• -Warray-parameter
• -Warray-parameter=n
• -Wno-attributes
• -Wattribute-alias=n
• -Wno-attribute-alias
• -Wno-attribute-warning
• -Wbidi-chars=[none|unpaired|any|ucn]
• -Wbool-compare
• -Wbool-operation
• -Wno-builtin-declaration-mismatch
• -Wno-builtin-macro-redefined
• -Wc90-c99-compat
• -Wc99-c11-compat
• -Wc11-c23-compat
• -Wc23-c2y-compat
• -Wc++-compat
• -Wc++11-compat
• -Wc++14-compat
• -Wc++17-compat
• -Wc++20-compat
• -Wno-c++11-extensions
• -Wno-c++14-extensions
• -Wno-c++17-extensions
• -Wno-c++20-extensions
• -Wno-c++23-extensions
• -Wcalloc-transposed-args
• -Wcast-align
• -Wcast-align=strict
• -Wcast-function-type
• -Wcast-qual
• -Wchar-subscripts
• -Wclobbered
• -Wcomment
• -Wcompare-distinct-pointer-types
• -Wno-complain-wrong-lang
• -Wconversion
• -Wno-coverage-mismatch
• -Wno-cpp
• -Wdangling-else
• -Wdangling-pointer
• -Wdangling-pointer=n
• -Wdate-time
• -Wno-deprecated
• -Wno-deprecated-declarations
• -Wno-designated-init
• -Wdisabled-optimization
• -Wno-discarded-array-qualifiers
• -Wno-discarded-qualifiers
• -Wno-div-by-zero
• -Wdouble-promotion
• -Wduplicated-branches
• -Wduplicated-cond
• -Wempty-body
• -Wno-endif-labels
• -Wenum-compare
• -Wenum-conversion
• -Wenum-int-mismatch
• -Werror
• -Werror=*
• -Wexpansion-to-defined
• -Wfatal-errors
• -Wflex-array-member-not-at-end
• -Wfloat-conversion
• -Wfloat-equal
• -Wformat
• -Wformat=2
• -Wno-format-contains-nul
• -Wno-format-extra-args
• -Wformat-nonliteral
• -Wformat-overflow=n
• -Wformat-security
• -Wformat-signedness
• -Wformat-truncation=n
• -Wformat-y2k
• -Wframe-address
• -Wframe-larger-than=byte-size
• -Wno-free-nonheap-object
• -Wheader-guard
• -Wno-if-not-aligned
• -Wno-ignored-attributes
• -Wignored-qualifiers
• -Wno-incompatible-pointer-types
• -Whardened
• -Wimplicit
• -Wimplicit-fallthrough
• -Wimplicit-fallthrough=n
• -Wno-implicit-function-declaration
• -Wno-implicit-int
• -Winfinite-recursion
• -Winit-self
• -Winline
• -Wno-int-conversion
• -Wint-in-bool-context
• -Wno-int-to-pointer-cast
• -Wno-invalid-memory-model
• -Winvalid-pch
• -Winvalid-utf8
• -Wno-unicode
• -Wjump-misses-init
• -Wlarger-than=byte-size
• -Wleading-whitespace=kind
• -Wlogical-not-parentheses
• -Wlogical-op
• -Wlong-long
• -Wno-lto-type-mismatch
• -Wmain
• -Wmaybe-uninitialized
• -Wmemset-elt-size
• -Wmemset-transposed-args
• -Wmisleading-indentation
• -Wmissing-attributes
• -Wmissing-braces
• -Wmissing-field-initializers
• -Wmissing-format-attribute
• -Wmissing-include-dirs
• -Wmissing-noreturn
• -Wmusttail-local-addr
• -Wmaybe-musttail-local-addr
• -Wno-missing-profile
• -Wno-multichar
• -Wmultistatement-macros
• -Wnonnull
• -Wnonnull-compare
• -Wnormalized=[none|id|nfc|nfkc]
• -Wnull-dereference
• -Wno-odr
• -Wopenacc-parallelism
• -Wopenmp
• -Wopenmp-simd
• -Wno-overflow
• -Woverlength-strings
• -Wno-override-init-side-effects
• -Wpacked
• -Wno-packed-bitfield-compat
• -Wpacked-not-aligned
• -Wpadded
• -Wparentheses
• -Wno-pedantic-ms-format
• -Wpointer-arith
• -Wno-pointer-compare
• -Wno-pointer-to-int-cast
• -Wno-pragmas
• -Wno-pragma-once-outside-header
• -Wno-prio-ctor-dtor
• -Wno-psabi
• -Wredundant-decls
• -Wrestrict
• -Wno-return-local-addr
• -Wreturn-type
• -Wno-scalar-storage-order
• -Wsequence-point
• -Wshadow
• -Wshadow=global
• -Wshadow=local
• -Wshadow=compatible-local
• -Wno-shadow-ivar
• -Wno-shift-count-negative
• -Wno-shift-count-overflow
• -Wshift-negative-value
• -Wno-shift-overflow
• -Wshift-overflow=n
• -Wsign-compare
• -Wsign-conversion
• -Wno-sizeof-array-argument
• -Wsizeof-array-div
• -Wsizeof-pointer-div
• -Wsizeof-pointer-memaccess
• -Wstack-protector
• -Wstack-usage=byte-size
• -Wstrict-aliasing
• -Wstrict-aliasing=n
• -Wstrict-overflow
• -Wstrict-overflow=n
• -Wstring-compare
• -Wno-stringop-overflow
• -Wno-stringop-overread
• -Wno-stringop-truncation
• -Wstrict-flex-arrays
• -Wsuggest-attribute=attribute-name
• -Wswitch
• -Wno-switch-bool
• -Wswitch-default
• -Wswitch-enum
• -Wno-switch-outside-range
• -Wno-switch-unreachable
• -Wsync-nand
• -Wsystem-headers
• -Wtautological-compare
• -Wtrailing-whitespace
• -Wtrailing-whitespace=kind
• -Wtrampolines
• -Wtrigraphs
• -Wtrivial-auto-var-init
• -Wno-tsan
• -Wtype-limits
• -Wundef
• -Wuninitialized
• -Wunknown-pragmas
• -Wunsuffixed-float-constants
• -Wunterminated-string-initialization
• -Wunused
• -Wunused-but-set-parameter
• -Wunused-but-set-variable
• -Wunused-const-variable
• -Wunused-const-variable=n
• -Wunused-function
• -Wunused-label
• -Wunused-local-typedefs
• -Wunused-macros
• -Wunused-parameter
• -Wno-unused-result
• -Wunused-value
• -Wunused-variable
• -Wuse-after-free
• -Wuse-after-free=n
• -Wuseless-cast
• -Wno-varargs
• -Wvariadic-macros
• -Wvector-operation-performance
• -Wvla
• -Wvla-larger-than=byte-size
• -Wno-vla-larger-than
• -Wvolatile-register-var
• -Wwrite-strings
• -Wno-xor-used-as-pow
• -Wzero-as-null-pointer-constant
• -Wzero-length-bounds

ik ik, that's A LOT. like, an obsessive amount. and bestie? this is just the index. every one of these has its own little lore later in the docs. we are not covering all of it here because this list is already giving "the compiler knows everything you've ever done." 💀

---

🔍 Static Analyzer Options 🔍

this is gcc in its detective era. 🕵️♀️ the static analyzer digs through your code and is like "hmm. that pointer. bestie. that pointer is NOT okay." main character energy fr fr:

• -fanalyzer
• -fanalyzer-call-summaries
• -fanalyzer-checker=name
• -fno-analyzer-feasibility
• -fanalyzer-fine-grained
• -fanalyzer-show-events-in-system-headers
• -fno-analyzer-state-merge
• -fno-analyzer-state-purge
• -fno-analyzer-suppress-followups
• -fanalyzer-transitivity
• -fno-analyzer-undo-inlining
• -fanalyzer-verbose-edges
• -fanalyzer-verbose-state-changes
• -fanalyzer-verbosity=level
• -fdump-analyzer
• -fdump-analyzer-callgraph
• -fdump-analyzer-exploded-graph
• -fdump-analyzer-exploded-nodes
• -fdump-analyzer-exploded-nodes-2
• -fdump-analyzer-exploded-nodes-3
• -fdump-analyzer-exploded-paths
• -fdump-analyzer-feasibility
• -fdump-analyzer-infinite-loop
• -fdump-analyzer-json
• -fdump-analyzer-state-purge
• -fdump-analyzer-stderr
• -fdump-analyzer-supergraph
• -fdump-analyzer-untracked
• -Wno-analyzer-double-fclose
• -Wno-analyzer-double-free
• -Wno-analyzer-exposure-through-output-file
• -Wno-analyzer-exposure-through-uninit-copy
• -Wno-analyzer-fd-access-mode-mismatch
• -Wno-analyzer-fd-double-close
• -Wno-analyzer-fd-leak
• -Wno-analyzer-fd-phase-mismatch
• -Wno-analyzer-fd-type-mismatch
• -Wno-analyzer-fd-use-after-close
• -Wno-analyzer-fd-use-without-check
• -Wno-analyzer-file-leak
• -Wno-analyzer-free-of-non-heap
• -Wno-analyzer-imprecise-fp-arithmetic
• -Wno-analyzer-infinite-loop
• -Wno-analyzer-infinite-recursion
• -Wno-analyzer-jump-through-null
• -Wno-analyzer-malloc-leak
• -Wno-analyzer-mismatching-deallocation
• -Wno-analyzer-null-argument
• -Wno-analyzer-null-dereference
• -Wno-analyzer-out-of-bounds
• -Wno-analyzer-overlapping-buffers
• -Wno-analyzer-possible-null-argument
• -Wno-analyzer-possible-null-dereference
• -Wno-analyzer-putenv-of-auto-var
• -Wno-analyzer-shift-count-negative
• -Wno-analyzer-shift-count-overflow
• -Wno-analyzer-stale-setjmp-buffer
• -Wno-analyzer-tainted-allocation-size
• -Wno-analyzer-tainted-assertion
• -Wno-analyzer-tainted-array-index
• -Wno-analyzer-tainted-divisor
• -Wno-analyzer-tainted-offset
• -Wno-analyzer-tainted-size
• -Wanalyzer-symbol-too-complex
• -Wanalyzer-too-complex
• -Wno-analyzer-undefined-behavior-ptrdiff
• -Wno-analyzer-undefined-behavior-strtok
• -Wno-analyzer-unsafe-call-within-signal-handler
• -Wno-analyzer-use-after-free
• -Wno-analyzer-use-of-pointer-in-stale-stack-frame
• -Wno-analyzer-use-of-uninitialized-value
• -Wno-analyzer-va-arg-type-mismatch
• -Wno-analyzer-va-list-exhausted
• -Wno-analyzer-va-list-leak
• -Wno-analyzer-va-list-use-after-va-end
• -Wno-analyzer-write-to-const
• -Wno-analyzer-write-to-string-literal

the way this list just keeps *going*… it's giving forensic accountant but make it a compiler. 🎀

---

🌸 C and Objective-C-only Warning Options 🌸

these ones are picky — they only show up for the C and Objective-C girlies, not the C++ crowd. exclusivity? we love to see it:

• -Wbad-function-cast
• -Wdeprecated-non-prototype
• -Wfree-labels
• -Wmissing-declarations
• -Wmissing-parameter-name
• -Wmissing-parameter-type
• -Wdeclaration-missing-parameter-type
• -Wmissing-prototypes
• -Wmissing-variable-declarations
• -Wnested-externs
• -Wold-style-declaration
• -Wold-style-definition
• -Wstrict-prototypes
• -Wtraditional
• -Wtraditional-conversion
• -Wdeclaration-after-statement
• -Wpointer-sign

ngl these are the "respect the classics" warnings and honestly? periodt. 💅

---

🎀 Debugging Options 🎀

okay so when your code is doing the absolute most and you need to see INSIDE its little head, you reach for these. debugging is the self-care of programming, bestie — you cannot fix what you cannot see, and -g is your bestie for that:

• -g
• -glevel
• -gdwarf
• -gdwarf-version
• -gbtf
• -gctf
• -gctflevel
• -gprune-btf
• -gno-prune-btf
• -ggdb
• -grecord-gcc-switches
• -gno-record-gcc-switches
• -gstrict-dwarf
• -gno-strict-dwarf
• -gas-loc-support
• -gno-as-loc-support
• -gas-locview-support
• -gno-as-locview-support
• -gcodeview
• -gcolumn-info
• -gno-column-info
• -gdwarf32
• -gdwarf64
• -gstatement-frontiers
• -gno-statement-frontiers
• -gvariable-location-views
• -gno-variable-location-views
• -ginternal-reset-location-views
• -gno-internal-reset-location-views
• -ginline-points
• -gno-inline-points
• -gvms
• -gz[=type]
• -gsplit-dwarf
• -gdescribe-dies
• -gno-describe-dies
• -fdebug-prefix-map=old=new
• -fdebug-types-section
• -fno-eliminate-unused-debug-types
• -femit-struct-debug-baseonly
• -femit-struct-debug-reduced
• -femit-struct-debug-detailed[=spec-list]
• -fno-eliminate-unused-debug-symbols
• -femit-class-debug-always
• -fno-merge-debug-strings
• -fno-dwarf2-cfi-asm
• -fvar-tracking
• -fvar-tracking-assignments

see how many -gno- versions there are? every -g something has a little -gno- twin that turns it right back off. iconic symmetry, no notes. 🦋

✨ Optimization Options ✨

okay bestie, this is THE section. this is where gcc stops being a compiler and starts being a personal trainer for your code. every flag here is a lil instruction that says "make my binary faster / smaller / hotter". you do not need to memorise all of these — most of them are switched on by -O levels automatically — but you need to know they exist, so here's the whole roster, no skips, fr fr. 🏋️‍♀️✨

• 🎀 the alignment clique (they will NOT tolerate messy memory): -faggressive-loop-optimizations, -falign-functions[=n[:m:[n2[:m2]]]], -falign-jumps[=n[:m:[n2[:m2]]]], -falign-labels[=n[:m:[n2[:m2]]]], -falign-loops[=n[:m:[n2[:m2]]]], -fmin-function-alignment=[n]
• 🧠 the foundational brainiacs: -fno-allocation-dce, -fallow-store-data-races, -fassociative-math, -fauto-profile, -fauto-profile[=path], -fauto-inc-dec, -fbranch-probabilities, -fcaller-saves
• 🧵 tiny housekeeping kings (stack adjustments, registers, crossjumping): -fcombine-stack-adjustments, -fconserve-stack, -ffold-mem-offsets, -fcompare-elim, -fcprop-registers, -fcrossjumping, -fcse-follow-jumps, -fcse-skip-blocks
• ➗ complex number girlies + section splitting: -fcx-fortran-rules, -fcx-limited-range, -fdata-sections, -fdce, -fdelayed-branch, -fdelete-null-pointer-checks, -fdevirtualize, -fdevirtualize-speculatively, -fdevirtualize-at-ltrans, -fdse
• 📞 inline everything, no thoughts, head empty: -fearly-inlining, -fipa-sra, -fexpensive-optimizations, -ffat-lto-objects, -ffast-math, -ffinite-math-only, -ffloat-store, -fexcess-precision=style, -ffinite-loops, -fforward-propagate, -ffp-contract=style, -ffunction-sections
• 🔁 global common subexpression elimination, the og slay: -fgcse, -fgcse-after-reload, -fgcse-las, -fgcse-lm, -fgraphite-identity, -fgcse-sm, -fhoist-adjacent-loads
• 🔀 if-conversion + inline culture: -fif-conversion, -fif-conversion2, -findirect-inlining, -finline-stringops[=fn], -finline-functions, -finline-functions-called-once, -finline-limit=n, -finline-small-functions
• 🫂 interprocedural analysis besties (the ipa girlies): -fipa-modref, -fipa-cp, -fipa-cp-clone, -fipa-bit-cp, -fipa-vrp, -fipa-pta, -fipa-profile, -fipa-pure-const, -fipa-reference, -fipa-reference-addressable, -fipa-reorder-for-locality, -fipa-stack-alignment, -fipa-icf
• 🪑 register allocation tea ☕: -fira-algorithm=algorithm, -flate-combine-instructions, -flifetime-dse, -flive-patching=level, -fira-region=region, -fira-hoist-pressure, -fira-loop-pressure, -fno-ira-share-save-slots, -fno-ira-share-spill-slots
• 🚧 isolating the problem paths (no drama please): -fisolate-erroneous-paths-dereference, -fisolate-erroneous-paths-attribute, -fivopts, -fkeep-inline-functions, -fkeep-static-functions, -fkeep-static-consts, -flimit-function-alignment, -flive-range-shrinkage
• 🌀 loop transformation era: -floop-block, -floop-interchange, -floop-strip-mine, -floop-unroll-and-jam, -floop-nest-optimize, -floop-parallelize-all, -flra-remat
• 🌐 link-time optimization (lto hive mind): -flto, -flto-compression-level, -flto-partition=alg, -flto-incremental=path, -flto-incremental-cache-size=n
• 🗑️ dedupe + constant-merging behavior: -fmalloc-dce, -fmerge-all-constants, -fmerge-constants, -fmodulo-sched, -fmodulo-sched-allow-regmoves, -fmove-loop-invariants, -fmove-loop-stores
• 🚫 the "no" girlies (they're iconic, don't come for them): -fno-branch-count-reg, -fno-defer-pop, -fno-fp-int-builtin-inexact, -fno-function-cse, -fno-guess-branch-probability, -fno-inline, -fno-math-errno, -fno-peephole, -fno-peephole2, -fno-printf-return-value
• 🙅‍♀️ more no's, we love boundaries: -fno-sched-interblock, -fno-sched-spec, -fno-signed-zeros, -fno-toplevel-reorder, -fno-trapping-math, -fno-zero-initialized-in-bss
• 🦴 frame pointer, crc, sibling calls, partial inlining: -fomit-frame-pointer, -foptimize-crc, -foptimize-sibling-calls, -fpartial-inlining, -fpeel-loops, -fpredictive-commoning, -fprefetch-loop-arrays
• 📊 profile-guided everything (data-driven queen): -fprofile-correction, -fprofile-use, -fprofile-use=path, -fprofile-partial-training, -fprofile-values, -fprofile-reorder-functions, -freciprocal-math, -free, -frename-registers
• 📦 block reordering + functions on deck: -freorder-blocks, -freorder-blocks-algorithm=algorithm, -freorder-blocks-and-partition, -freorder-functions, -frerun-cse-after-loop, -freschedule-modulo-scheduled-loops, -frounding-math, -fsave-optimization-record
• 🚋 scheduling pipeline, all aboard: -fsched2-use-superblocks, -fsched-pressure, -fsched-spec-load, -fsched-spec-load-dangerous, -fsched-stalled-insns-dep[=n], -fsched-stalled-insns[=n]
• 🧮 scheduler heuristics (the vibes committee): -fsched-group-heuristic, -fsched-critical-path-heuristic, -fsched-spec-insn-heuristic, -fsched-rank-heuristic, -fsched-last-insn-heuristic, -fsched-dep-count-heuristic
• 📆 scheduling + section anchors: -fschedule-fusion, -fschedule-insns, -fschedule-insns2, -fsection-anchors
• 🎯 selective scheduling (picky but make it fashion): -fselective-scheduling, -fselective-scheduling2, -fsel-sched-pipelining, -fsel-sched-pipelining-outer-loops
• 🫥 symbol visibility + shrink wrap + ieee float tea: -fsemantic-interposition, -fshrink-wrap, -fshrink-wrap-separate, -fsignaling-nans, -fsingle-precision-constant
• ✂️ splitting era: -fsplit-ivs-in-unroller, -fsplit-loops, -fsplit-paths, -fsplit-wide-types, -fsplit-wide-types-early, -fssa-backprop, -fssa-phiopt, -fstdarg-opt, -fstore-merging
• 🪄 aliasing + threading + tracing: -fstrict-aliasing, -fipa-strict-aliasing, -fthread-jumps, -ftracer
• 🌳 the whole tree family (they're all so iconic i can't): -ftree-bit-ccp, -ftree-builtin-call-dce, -ftree-ccp, -ftree-ch, -ftree-coalesce-vars, -ftree-copy-prop, -ftree-cselim, -ftree-dce, -ftree-dominator-opts, -ftree-dse, -ftree-forwprop, -ftree-fre, -fcode-hoisting
• 🔂 tree loop transformations: -ftree-loop-if-convert, -ftree-loop-im, -ftree-phiprop, -ftree-loop-distribution, -ftree-loop-distribute-patterns, -ftree-loop-ivcanon, -ftree-loop-linear, -ftree-loop-optimize, -ftree-loop-vectorize, -ftree-parallelize-loops=n
• 🔎 pre / partial-pre / pta / reassoc crew: -ftree-pre, -ftree-partial-pre, -ftree-pta, -ftree-reassoc, -ftree-scev-cprop, -ftree-sink, -ftree-slsr, -ftree-sra, -ftree-switch-conversion, -ftree-tail-merge, -ftree-ter, -ftree-vectorize, -ftree-vrp
• 🧊 misc glow-ups: -ftrivial-auto-var-init, -funconstrained-commons, -funit-at-a-time, -funroll-all-loops, -funroll-loops, -funsafe-math-optimizations, -funswitch-loops
• 🏎️ misc round two, electric boogaloo: -fipa-ra, -fvariable-expansion-in-unroller, -fvect-cost-model, -fvpt, -fweb, -fwhole-program, -fwpa, -fuse-linker-plugin, -fzero-call-used-regs, --param name=value
• 👑 THE OPTIMIZATION LEVELS, the main characters themselves: -O, -O0, -O1, -O2, -O3, -Os, -Ofast, -Og, -Oz

ngl -O0 is the vibe when you're debugging and you want your stack trace to make ANY sense. -O3 is the vibe when you want speed and you're okay with your debugger crying. choose your fighter. 💀

---

🩺 Program Instrumentation Options 🩺

this is the "let's put lil sensors on your program" section. profiling, coverage, sanitizers, stack protection, security hardening — all the girlies who watch your code while it runs and go "um, actually, that's a bug." 🫶

• 📈 the og profiling duo + coverage: -p, -pg, -fprofile-arcs, --coverage, -ftest-coverage, -fcondition-coverage, -fpath-coverage
• 📁 profile path plumbing: -fprofile-abs-path, -fprofile-dir=path, -fprofile-generate, -fprofile-generate=path, -fprofile-info-section, -fprofile-info-section=name, -fprofile-note=path, -fprofile-prefix-path=path
• 🔧 profile update + filtering (regex girlies rise): -fprofile-update=method, -fprofile-filter-files=regex, -fprofile-exclude-files=regex, -fprofile-reproducible=[multithreaded|parallel-runs|serial]
• 🧼 sanitizers!! my beloved: -fsanitize=style, -fsanitize-recover, -fsanitize-recover=style, -fsanitize-trap, -fsanitize-trap=style, -fasan-shadow-offset=number, -fsanitize-sections=s1,s2,..., -fsanitize-undefined-trap-on-error, -fbounds-check
• 🛡️ control flow protection + hardening: -fcf-protection, -fcf-protection=[full|branch|return|none|check], -fharden-compares, -fharden-conditional-branches, -fhardened, -fharden-control-flow-redundancy, -fhardcfr-skip-leaf, -fhardcfr-check-exceptions, -fhardcfr-check-returning-calls, -fhardcfr-check-noreturn-calls=[always|no-xthrow|nothrow|never]
• 🥞 stack protector squad (buffer overflows, not today): -fstack-protector, -fstack-protector-all, -fstack-protector-strong, -fstack-protector-explicit, -fstack-check, -fstack-limit-register=reg, -fstack-limit-symbol=sym, -fno-stack-limit, -fsplit-stack
• 🐛 stack scrubbing (strub) family: -fstrub=disable, -fstrub=strict, -fstrub=relaxed, -fstrub=all, -fstrub=at-calls, -fstrub=internal
• 🧙 vtable verification + debugging: -fvtable-verify=[std|preinit|none], -fvtv-counts, -fvtv-debug
• 🪝 function instrumentation hooks: -finstrument-functions, -finstrument-functions-once, -finstrument-functions-exclude-function-list=sym,sym,..., -finstrument-functions-exclude-file-list=file,file,...
• 🏷️ prefix mapping + patchable entries: -fprofile-prefix-map=old=new, -fpatchable-function-entry=N[,M]

---

🧵 Preprocessor Options 🧵

the preprocessor is that friend who reads your message before you send it and rewrites it entirely. macros, includes, conditionals, all of it. here's the full toolbox. 💜

• 🅰️ the define/undef basics: -Aquestion=answer, -A-question[=answer], -C, -CC, -Dmacro[=defn], -dD, -dI, -dM, -dN, -dU
• 🔍 debug-cpp + directives-only + charset tea: -fdebug-cpp, -fdirectives-only, -fdollars-in-identifiers, -fexec-charset=charset, -fextended-identifiers, -finput-charset=charset, -fmacro-prefix-map=old=new, -fmax-include-depth=depth
• 🗂️ canonical headers + pch behavior: -fno-canonical-system-headers, -fpch-deps, -fpch-preprocess, -fpreprocessed, -ftabstop=width, -ftrack-macro-expansion, -fwide-exec-charset=charset, -fworking-directory, -H
• 📎 include/macros at the command line: -imacros file, -include file
• 📋 the M family (dependency generation, they run this town): -M, -MD, -MF, -MG, -MM, -MMD, -MP, -MQ, -MT, -Mno-modules
• ⚙️ misc preprocessor switches: -no-integrated-cpp, -P, -pthread, -remap, -traditional, -traditional-cpp, -trigraphs, -Umacro, -undef, -Wp,option, -Xpreprocessor option

---

🔩 Assembler Options 🔩

just two, but they're lowkey powerful. pass stuff straight through to the assembler. 🦋

• -Wa,option, -Xassembler option

---

🔗 Linker Options 🔗

the linker is the final boss of the build. it's where all your .o files meet and either become besties or start beefing. check this roster. 🌈

• object-file-name, -fuse-ld=linker, -llibrary, -nostartfiles, -nodefaultlibs, -nolibc, -nostdlib, -nostdlib++, -e entry, --entry=entry, -pie, -pthread, -r, -rdynamic, -s, -static, -static-pie, -static-libgcc, -static-libstdc++, -static-libasan, -static-libtsan, -static-liblsan, -static-libubsan, -shared, -shared-libgcc, -symbolic, -T script, -Wl,option, -Xlinker option, -u symbol, -z keyword

---

📁 Directory Options 📁

where does gcc look for headers and libraries? girl, it has OPINIONS, and also a search path you can absolutely rearrange. 🔥

• -Bprefix, -Idir, -I-, -idirafter dir, -imacros file, -imultilib dir, -iplugindir=dir, -iprefix file, -iquote dir, -isysroot dir, -isystem dir, -iwithprefix dir, -iwithprefixbefore dir, --embed-dir=dir, -Ldir, -no-canonical-prefixes, --no-sysroot-suffix, -nostdinc, -nostdinc++, --sysroot=dir

---

🏭 Code Generation Options 🏭

this section is about how the code comes out the other side: registers, exceptions, pic/pie, structs, visibility, all the structural decisions. it's giving "final look in the mirror before we go out." 💅

• 🪑 the register rules: -fcall-saved-reg, -fcall-used-reg, -ffixed-reg
• 🚨 exception handling behavior: -fexceptions, -fnon-call-exceptions, -fdelete-dead-exceptions, -funwind-tables, -fasynchronous-unwind-tables, -fno-gnu-unique, -finhibit-size-directive, -fcommon, -fno-ident
• 🧱 struct return conventions: -fpcc-struct-return, -fpic, -fPIC, -fpie, -fPIE, -fno-plt, -fno-jump-tables, -fno-bit-tests, -frecord-gcc-switches, -freg-struct-return
• 🧬 enum / wchar / asm / packing: -fshort-enums, -fshort-wchar, -fverbose-asm, -fpack-struct[=n], -fleading-underscore, -ftls-model=model, -fstack-reuse=reuse_level
• 🕳️ trampolines + trapv + wrapv: -ftrampolines, -ftrampoline-impl=[stack|heap], -ftrapv, -fwrapv
• 👁️ visibility + volatile bitfields + sync + padding: -fvisibility=[default|internal|hidden|protected], -fstrict-volatile-bitfields, -fsync-libcalls, -fzero-init-padding-bits=value

---

🛠️ Developer Options 🛠️

this is the backstage pass section. dumps, debug counters, pass enabling/disabling, memory reports, timing reports — everything you need when you're like "wait WHY did the compiler do that". these are not for casual use, they're for when you and gcc need to have a serious talk. 😭✨

• 🐣 the little -d dicers + identity crisis: -dletters, -dumpspecs, -dumpmachine, -dumpversion, -dumpfullversion
• 📊 callgraph + checking + debug counters: -fcallgraph-info[=su,da], -fchecking, -fchecking=n, -fdbg-cnt-list, -fdbg-cnt=counter-value-list
• 🔌 disabling passes (ipe, rtl, tree): -fdisable-ipa-pass_name, -fdisable-rtl-pass_name, -fdisable-rtl-pass-name=range-list, -fdisable-tree-pass_name, -fdisable-tree-pass-name=range-list
• 🗄️ the dump family, pt. 1 (debug, addressing, numbering): -fdump-debug, -fdump-earlydebug, -fdump-noaddr, -fdump-unnumbered, -fdump-unnumbered-links, -fdump-final-insns[=file]
• 🗄️ the dump family, pt. 2 (ipa + lang dumps): -fdump-ipa-all, -fdump-ipa-cgraph, -fdump-ipa-inline, -fdump-lang-all, -fdump-lang-switch, -fdump-lang-switch-options, -fdump-lang-switch-options=filename
• 🗄️ the dump family, pt. 3 (passes + rtl dumps): -fdump-passes, -fdump-rtl-pass, -fdump-rtl-pass=filename, -fdump-statistics
• 🗄️ the dump family, pt. 4 (tree dumps, the finale): -fdump-tree-all, -fdump-tree-switch, -fdump-tree-switch-options, -fdump-tree-switch-options=filename
• 🆚 debug comparison + enabling passes: -fcompare-debug[=opts], -fcompare-debug-second, -fenable-kind-pass, -fenable-kind-pass=range-list
• 📢 verbose + reports (lto, mem, opt, stats): -fira-verbose=n, -flto-report, -flto-report-wpa, -fmem-report-wpa, -fmem-report, -fpre-ipa-mem-report, -fpost-ipa-mem-report, -fopt-info, -fopt-info-options[=file], -fmultiflags, -fprofile-report
• 🎲 seeds, scheduler verbosity, sel-sched debug: -frandom-seed=string, -fsched-verbose=n, -fsel-sched-verbose, -fsel-sched-dump-cfg, -fsel-sched-pipelining-verbose
• ⏱️ stats + stack usage + time reports: -fstats, -fstack-usage, -ftime-report, -ftime-report-details, -fvar-tracking-assignments-toggle, -gtoggle
• 🔍 the -print-* crew (gcc answering your questions): -print-file-name=library, -print-libgcc-file-name, -print-multi-directory, -print-multi-lib, -print-multi-os-directory, -print-prog-name=program, -print-search-dirs, -Q, -print-sysroot, -print-sysroot-headers-suffix
• 💾 keep your temps, bestie: -save-temps, -save-temps=cwd, -save-temps=obj, -time[=file]

💅 Machine-Dependent Options 💅

okay so THIS is the part where GCC says "every chip is different and I love them all individually" 🫶. Below is the full menu of per-processor option families — every single flag, exactly as it is, no notes. If your target lives here, your target has opinions, and compiler girl is here to honor them 💖.

---

🦋 AArch64 Options

64-bit ARM, bestie. The whole roster, one bullet each:

• -mabi=name
• -mbig-endian
• -mlittle-endian
• -mgeneral-regs-only
• -mcmodel=tiny
• -mcmodel=small
• -mcmodel=large
• -mstrict-align
• -mno-strict-align
• -momit-leaf-frame-pointer
• -mtls-dialect=desc
• -mtls-dialect=traditional
• -mtls-size=size
• -mfix-cortex-a53-835769
• -mfix-cortex-a53-843419
• -mlow-precision-recip-sqrt
• -mlow-precision-sqrt
• -mlow-precision-div
• -mpc-relative-literal-loads
• -msign-return-address=scope
• -mbranch-protection=none|standard|pac-ret[+leaf +b-key]|bti|gcs
• -mharden-sls=opts
• -march=name
• -mcpu=name
• -mtune=name
• -moverride=string
• -mverbose-cost-dump
• -mstack-protector-guard=guard
• -mstack-protector-guard-reg=sysreg
• -mstack-protector-guard-offset=offset
• -mtrack-speculation
• -moutline-atomics
• -mearly-ldp-fusion
• -mlate-ldp-fusion
• -Wexperimental-fmv-target

---

🧚 Adapteva Epiphany Options

tiny many-core fairy energy ✨:

• -mhalf-reg-file
• -mprefer-short-insn-regs
• -mbranch-cost=num
• -mcmove
• -mnops=num
• -msoft-cmpsf
• -msplit-lohi
• -mpost-inc
• -mpost-modify
• -mstack-offset=num
• -mround-nearest
• -mlong-calls
• -mshort-calls
• -msmall16
• -mfp-mode=mode
• -mvect-double
• -max-vect-align=num
• -msplit-vecmove-early
• -m1reg-reg

---

🔥 AMD GCN Options

GPU hours, and the list is genuinely three flags long. iconic minimalism:

• -march=gpu
• -mtune=gpu
• -mstack-size=bytes

---

🌈 ARC Options

ARC said "let me write a whole novel" and honestly? respect 😭. All of it, in order:

• -mbarrel-shifter
• -mjli-always
• -mcpu=cpu
• -mA6
• -mARC600
• -mA7
• -mARC700
• -mdpfp
• -mdpfp-compact
• -mdpfp-fast
• -mno-dpfp-lrsr
• -mea
• -mno-mpy
• -mmul32x16
• -mmul64
• -matomic
• -mnorm
• -mspfp
• -mspfp-compact
• -mspfp-fast
• -msimd
• -msoft-float
• -mswap
• -mcrc
• -mdsp-packa
• -mdvbf
• -mlock
• -mmac-d16
• -mmac-24
• -mrtsc
• -mswape
• -mtelephony
• -mxy
• -misize
• -mannotate-align
• -marclinux
• -marclinux_prof
• -mlong-calls
• -mmedium-calls
• -msdata
• -mirq-ctrl-saved
• -mrgf-banked-regs
• -mlpc-width=width
• -G num
• -mvolatile-cache
• -mtp-regno=regno
• -malign-call
• -mauto-modify-reg
• -mbbit-peephole
• -mno-brcc
• -mcase-vector-pcrel
• -mcompact-casesi
• -mno-cond-exec
• -mearly-cbranchsi
• -mexpand-adddi
• -mindexed-loads
• -mlra
• -mlra-priority-none
• -mlra-priority-compact
• -mlra-priority-noncompact
• -mmillicode
• -mmixed-code
• -mq-class
• -mRcq
• -mRcw
• -msize-level=level
• -mtune=cpu
• -mmultcost=num
• -mcode-density-frame
• -munalign-prob-threshold=probability
• -mmpy-option=multo
• -mdiv-rem
• -mcode-density
• -mll64
• -mfpu=fpu
• -mrf16
• -mbranch-index

---

🎀 ARM Options

the ARM buffet, served on a lil plate, every single dish 🌸:

• -mapcs-frame
• -mno-apcs-frame
• -mabi=name
• -mapcs-stack-check
• -mno-apcs-stack-check
• -mapcs-reentrant
• -mno-apcs-reentrant
• -mgeneral-regs-only
• -msched-prolog
• -mno-sched-prolog
• -mlittle-endian
• -mbig-endian
• -mbe8
• -mbe32
• -mfloat-abi=name
• -mfp16-format=name
• -mthumb-interwork
• -mno-thumb-interwork
• -mcpu=name
• -march=name
• -mfpu=name
• -mtune=name
• -mprint-tune-info
• -mstructure-size-boundary=n
• -mabort-on-noreturn
• -mlong-calls
• -mno-long-calls
• -msingle-pic-base
• -mno-single-pic-base
• -mpic-register=reg
• -mnop-fun-dllimport
• -mpoke-function-name
• -mthumb
• -marm
• -mflip-thumb
• -mtpcs-frame
• -mtpcs-leaf-frame
• -mcaller-super-interworking
• -mcallee-super-interworking
• -mtp=name
• -mtls-dialect=dialect
• -mword-relocations
• -mfix-cortex-m3-ldrd
• -mfix-cortex-a57-aes-1742098
• -mfix-cortex-a72-aes-1655431
• -munaligned-access
• -mneon-for-64bits
• -mslow-flash-data
• -masm-syntax-unified
• -mrestrict-it
• -mverbose-cost-dump
• -mpure-code
• -mcmse
• -mfix-cmse-cve-2021-35465
• -mstack-protector-guard=guard
• -mstack-protector-guard-offset=offset
• -mfdpic
• -mbranch-protection=none|standard|pac-ret[+leaf][+bti]|bti[+pac-ret[+leaf]]

---

🍓 AVR Options

microcontroller gang, you get a LOT of levers 🫶:

• -mmcu=mcu
• -mabsdata
• -maccumulate-args
• -mcvt
• -mbranch-cost=cost
• -mfuse-add=level
• -mfuse-move=level
• -mfuse-move2
• -mcall-prologues
• -mgas-isr-prologues
• -mint8
• -mflmap
• -mdouble=bits
• -mlong-double=bits
• -mno-call-main
• -mn_flash=size
• -mfract-convert-truncate
• -mno-interrupts
• -mmain-is-OS_task
• -mrelax
• -mrmw
• -mstrict-X
• -mtiny-stack
• -mrodata-in-ram
• -msplit-bit-shift
• -msplit-ldst
• -mshort-calls
• -mskip-bug
• -muse-nonzero-bits
• -nodevicelib
• -nodevicespecs
• -Waddr-space-convert
• -Wmisspelled-isr

---

🖤 Blackfin Options

every flag, no skips, we are thorough in this house 💅:

• -mcpu=cpu[-sirevision]
• -msim
• -momit-leaf-frame-pointer
• -mno-omit-leaf-frame-pointer
• -mspecld-anomaly
• -mno-specld-anomaly
• -mcsync-anomaly
• -mno-csync-anomaly
• -mlow-64k
• -mno-low64k
• -mstack-check-l1
• -mid-shared-library
• -mno-id-shared-library
• -mshared-library-id=n
• -mleaf-id-shared-library
• -mno-leaf-id-shared-library
• -msep-data
• -mno-sep-data
• -mlong-calls
• -mno-long-calls
• -mfast-fp
• -minline-plt
• -mmulticore
• -mcorea
• -mcoreb
• -msdram
• -micplb

---

⭐ C6X Options

short and sweet, a lil snack:

• -mbig-endian
• -mlittle-endian
• -march=cpu
• -msim
• -msdata=sdata-type

---

🦋 CRIS Options

including the two that are just -sim and -sim2, and yes, one dash only, that's not a typo, don't come for me:

• -mcpu=cpu
• -march=cpu
• -mtune=cpu
• -mmax-stack-frame=n
• -metrax4
• -metrax100
• -mpdebug
• -mcc-init
• -mno-side-effects
• -mstack-align
• -mdata-align
• -mconst-align
• -m32-bit
• -m16-bit
• -m8-bit
• -mno-prologue-epilogue
• -melf
• -maout
• -sim
• -sim2
• -mmul-bug-workaround
• -mno-mul-bug-workaround

---

💜 C-SKY Options

a whole vibe check's worth of flags:

• -march=arch
• -mcpu=cpu
• -mbig-endian
• -EB
• -mlittle-endian
• -EL
• -mhard-float
• -msoft-float
• -mfpu=fpu
• -mdouble-float
• -mfdivdu
• -mfloat-abi=name
• -melrw
• -mistack
• -mmp
• -mcp
• -mcache
• -msecurity
• -mtrust
• -mdsp
• -medsp
• -mvdsp
• -mdiv
• -msmart
• -mhigh-registers
• -manchor
• -mpushpop
• -mmultiple-stld
• -mconstpool
• -mstack-size
• -mccrt
• -mbranch-cost=n
• -mcse-cc
• -msched-prolog
• -msim

---

🌸 Cygwin and MinGW Options

Windows-adjacent besties, this is your section:

• -mconsole
• -mcrtdll=library
• -mdll
• -mnop-fun-dllimport
• -mthread
• -municode
• -mwin32
• -mwindows
• -fno-set-stack-executable

---

🍎 Darwin Options

macOS hours 🫶. Heads up: this list has some repeats in it (-sectobjectsymbols and -segs_read_only_addr / -segs_read_write_addr each show up more than once) and I am leaving them EXACTLY as they are because the rule is the rule:

• -all_load
• -allowable_client
• -arch
• -arch_errors_fatal
• -arch_only
• -bind_at_load
• -bundle
• -bundle_loader
• -client_name
• -compatibility_version
• -current_version
• -dead_strip
• -dependency-file
• -dylib_file
• -dylinker_install_name
• -dynamic
• -dynamiclib
• -exported_symbols_list
• -filelist
• -flat_namespace
• -force_cpusubtype_ALL
• -force_flat_namespace
• -headerpad_max_install_names
• -iframework
• -image_base
• -init
• -install_name
• -keep_private_externs
• -multi_module
• -multiply_defined
• -multiply_defined_unused
• -noall_load
• -no_dead_strip_inits_and_terms
• -nodefaultrpaths
• -nofixprebinding
• -nomultidefs
• -noprebind
• -noseglinkedit
• -pagezero_size
• -prebind
• -prebind_all_twolevel_modules
• -private_bundle
• -read_only_relocs
• -sectalign
• -sectobjectsymbols
• -whyload
• -seg1addr
• -sectcreate
• -sectobjectsymbols
• -sectorder
• -segaddr
• -segs_read_only_addr
• -segs_read_write_addr
• -seg_addr_table
• -seg_addr_table_filename
• -seglinkedit
• -segprot
• -segs_read_only_addr
• -segs_read_write_addr
• -single_module
• -static
• -sub_library
• -sub_umbrella
• -twolevel_namespace
• -umbrella
• -undefined
• -unexported_symbols_list
• -weak_reference_mismatches
• -whatsloaded
• -F
• -gused
• -gfull
• -mmacosx-version-min=version
• -mkernel
• -mone-byte-bool

---

🌈 DEC Alpha Options

old school and elegant, honestly it's giving vintage 👑

• -mno-fp-regs
• -msoft-float
• -mieee
• -mieee-with-inexact
• -mieee-conformant
• -mfp-trap-mode=mode
• -mfp-rounding-mode=mode
• -mtrap-precision=mode
• -mbuild-constants
• -mcpu=cpu-type
• -mtune=cpu-type
• -mbwx
• -mmax
• -mfix
• -mcix
• -msafe-bwa
• -msafe-partial
• -mfloat-vax
• -mfloat-ieee
• -mexplicit-relocs
• -msmall-data
• -mlarge-data
• -msmall-text
• -mlarge-text
• -mmemory-latency=time

---

🌟 eBPF Options

kernel-side girlies, this one's yours 💖:

• -mbig-endian
• -mlittle-endian
• -mframe-limit=bytes
• -mxbpf
• -mco-re
• -mno-co-re
• -mjmpext
• -mjmp32
• -malu32
• -mv3-atomics
• -mbswap
• -msdiv
• -msmov
• -mcpu=version
• -masm=dialect
• -minline-memops-threshold=bytes

---

🎀 FR30 Options

two flags. two. the restraint, the discipline:

• -msmall-model
• -mno-lsim

---

🧁 FT32 Options

another lil bite-sized menu:

• -msim
• -mlra
• -mnodiv
• -mft32b
• -mcompress
• -mnopm

---

🔥 FRV Options

the toggles go ON and OFF and ON and OFF, a true duet of options:

• -mgpr-32
• -mgpr-64
• -mfpr-32
• -mfpr-64
• -mhard-float
• -msoft-float
• -malloc-cc
• -mfixed-cc
• -mdword
• -mno-dword
• -mdouble
• -mno-double
• -mmedia
• -mno-media
• -mmuladd
• -mno-muladd
• -mfdpic
• -minline-plt
• -mgprel-ro
• -multilib-library-pic
• -mlinked-fp
• -mlong-calls
• -malign-labels
• -mlibrary-pic
• -macc-4
• -macc-8
• -mpack
• -mno-pack
• -mno-eflags
• -mcond-move
• -mno-cond-move
• -moptimize-membar
• -mno-optimize-membar
• -mscc
• -mno-scc
• -mcond-exec
• -mno-cond-exec
• -mvliw-branch
• -mno-vliw-branch
• -mmulti-cond-exec
• -mno-multi-cond-exec
• -mnested-cond-exec
• -mno-nested-cond-exec
• -mtomcat-stats
• -mTLS
• -mtls
• -mcpu=cpu

---

🐧 GNU/Linux Options

li'l System V / libc family switcheroo section. Also -tno-android-cc and -tno-android-ld really have a t at the front, that's on purpose, leave it:

• -mglibc
• -muclibc
• -mmusl
• -mbionic
• -mandroid
• -tno-android-cc
• -tno-android-ld

---

🌸 H8/300 Options

short, punchy, eight flags exactly:

• -mrelax
• -mh
• -ms
• -mn
• -mexr
• -mno-exr
• -mint32
• -malign-300

---

👑 HPPA Options

a HUGE one. no notes, just vibes and a lot of flags:

• -march=architecture-type
• -matomic-libcalls
• -mbig-switch
• -mcaller-copies
• -mdisable-fpregs
• -mdisable-indexing
• -mordered
• -mfast-indirect-calls
• -mgas
• -mgnu-ld
• -mhp-ld
• -mfixed-range=register-range
• -mcoherent-ldcw
• -mjump-in-delay
• -mlinker-opt
• -mlong-calls
• -mlong-load-store
• -mno-atomic-libcalls
• -mno-disable-fpregs
• -mno-disable-indexing
• -mno-fast-indirect-calls
• -mno-gas
• -mno-jump-in-delay
• -mno-long-load-store
• -mno-portable-runtime
• -mno-soft-float
• -mno-space-regs
• -msoft-float
• -mpa-risc-1-0
• -mpa-risc-1-1
• -mpa-risc-2-0
• -mportable-runtime
• -mschedule=cpu-type
• -mspace-regs
• -msoft-mult
• -msio
• -mwsio
• -munix=unix-std
• -nolibdld
• -static
• -threads

---

🦋 IA-64 Options

Itanium girlies, this is a long one and it's all here, in order, no cap:

• -mbig-endian
• -mlittle-endian
• -mgnu-as
• -mgnu-ld
• -mno-pic
• -mvolatile-asm-stop
• -mregister-names
• -msdata
• -mno-sdata
• -mconstant-gp
• -mauto-pic
• -mfused-madd
• -minline-float-divide-min-latency
• -minline-float-divide-max-throughput
• -mno-inline-float-divide
• -minline-int-divide-min-latency
• -minline-int-divide-max-throughput
• -mno-inline-int-divide
• -minline-sqrt-min-latency
• -minline-sqrt-max-throughput
• -mno-inline-sqrt
• -mdwarf2-asm
• -mearly-stop-bits
• -mfixed-range=register-range
• -mtls-size=tls-size
• -mtune=cpu-type
• -milp32
• -mlp64
• -msched-br-data-spec
• -msched-ar-data-spec
• -msched-control-spec
• -msched-br-in-data-spec
• -msched-ar-in-data-spec
• -msched-in-control-spec
• -msched-spec-ldc
• -msched-spec-control-ldc
• -msched-prefer-non-data-spec-insns
• -msched-prefer-non-control-spec-insns
• -msched-stop-bits-after-every-cycle
• -msched-count-spec-in-critical-path
• -msel-sched-dont-check-control-spec
• -msched-fp-mem-deps-zero-cost
• -msched-max-memory-insns-hard-limit
• -msched-max-memory-insns=max-insns

---

🍓 LM32 Options

last lil family in this batch, five flags, all enabled-this and enabled-that:

• -mbarrel-shift-enabled
• -mdivide-enabled
• -mmultiply-enabled
• -msign-extend-enabled
• -muser-enabled

🐉 LoongArch Options (LoongArch Options)

okay so LoongArch has entered the chat and she brought a whole suitcase of flags 💜 every single one of these is a real option, no skips, here we go:

• -march=arch-type
• -mtune=tune-type
• -mabi=base-abi-type
• -mfpu=fpu-type
• -msimd=simd-type
• -msoft-float
• -msingle-float
• -mdouble-float
• -mlsx
• -mno-lsx
• -mlasx
• -mno-lasx
• -mbranch-cost=n
• -maddr-reg-reg-cost=n
• -mcheck-zero-division
• -mno-check-zero-division
• -mcond-move-int
• -mno-cond-move-int
• -mcond-move-float
• -mno-cond-move-float
• -memcpy
• -mno-memcpy
• -mstrict-align
• -mno-strict-align
• -G num
• -mmax-inline-memcpy-size=n
• -mexplicit-relocs=style
• -mexplicit-relocs
• -mno-explicit-relocs
• -mdirect-extern-access
• -mno-direct-extern-access
• -mcmodel=code-model
• -mrelax
• -mpass-mrelax-to-as
• -mrecip
• -mrecip=opt
• -mfrecipe
• -mno-frecipe
• -mdiv32
• -mno-div32
• -mlam-bh
• -mno-lam-bh
• -mlamcas
• -mno-lamcas
• -mld-seq-sa
• -mno-ld-seq-sa
• -mtls-dialect=opt
• -mannotate-tablejump
• -mno-annotate-tablejump

---

🧩 M32C Options (M32C Options)

short and sweet, three whole options, she's low maintenance fr:

• -mcpu=cpu
• -msim
• -memregs=number

---

🎛️ M32R/D Options (M32R/D Options)

• -m32r2
• -m32rx
• -m32r
• -mdebug
• -malign-loops
• -mno-align-loops
• -missue-rate=number
• -mbranch-cost=number
• -mmodel=code-size-model-type
• -msdata=sdata-type
• -mno-flush-func
• -mflush-func=name
• -mno-flush-trap
• -mflush-trap=number
• -G num

---

🕰️ M680x0 Options (M680x0 Options)

the 68k family tree is huge, everyone's here, it's giving family reunion 😭:

• -march=arch
• -mcpu=cpu
• -mtune=tune
• -m68000
• -m68020
• -m68020-40
• -m68020-60
• -m68030
• -m68040
• -m68060
• -mcpu32
• -m5200
• -m5206e
• -m528x
• -m5307
• -m5407
• -mcfv4e
• -mbitfield
• -mno-bitfield
• -mc68000
• -mc68020
• -mnobitfield
• -mrtd
• -mno-rtd
• -mdiv
• -mno-div
• -mshort
• -mno-short
• -mhard-float
• -m68881
• -msoft-float
• -mpcrel
• -malign-int
• -mstrict-align
• -msep-data
• -mno-sep-data
• -mshared-library-id=n
• -mid-shared-library
• -mno-id-shared-library
• -mxgot
• -mno-xgot
• -mlong-jump-table-offsets

---

🪶 MCore Options (MCore Options)

• -mhardlit
• -mno-hardlit
• -mdiv
• -mno-div
• -mrelax-immediates
• -mno-relax-immediates
• -mwide-bitfields
• -mno-wide-bitfields
• -m4byte-functions
• -mno-4byte-functions
• -mcallgraph-data
• -mno-callgraph-data
• -mslow-bytes
• -mno-slow-bytes
• -mno-lsim
• -mlittle-endian
• -mbig-endian
• -m210
• -m340
• -mstack-increment

---

🍃 MicroBlaze Options (MicroBlaze Options)

bestie said "tiny but mighty" and honestly? she ate 💅:

• -msoft-float
• -mhard-float
• -msmall-divides
• -mcpu=cpu
• -mmemcpy
• -mxl-soft-mul
• -mxl-soft-div
• -mxl-barrel-shift
• -mxl-pattern-compare
• -mxl-stack-check
• -mxl-gp-opt
• -mno-clearbss
• -mxl-multiply-high
• -mxl-float-convert
• -mxl-float-sqrt
• -mbig-endian
• -mlittle-endian
• -mxl-reorder
• -mxl-mode-app-model
• -mpic-data-is-text-relative

---

🌊 MIPS Options (MIPS Options)

okay this is the section, she has a million flags and she's not sorry, sit down, hydrate, we're going through all of them 🫶:

• -EL
• -EB
• -march=arch
• -mtune=arch
• -mips1
• -mips2
• -mips3
• -mips4
• -mips32
• -mips32r2
• -mips32r3
• -mips32r5
• -mips32r6
• -mips64
• -mips64r2
• -mips64r3
• -mips64r5
• -mips64r6
• -mips16
• -mno-mips16
• -mflip-mips16
• -minterlink-compressed
• -mno-interlink-compressed
• -minterlink-mips16
• -mno-interlink-mips16
• -mabi=abi
• -mabicalls
• -mno-abicalls
• -mshared
• -mno-shared
• -mplt
• -mno-plt
• -mxgot
• -mno-xgot
• -mgp32
• -mgp64
• -mfp32
• -mfpxx
• -mfp64
• -mhard-float
• -msoft-float
• -mno-float
• -msingle-float
• -mdouble-float
• -modd-spreg
• -mno-odd-spreg
• -mabs=mode
• -mnan=encoding
• -mdsp
• -mno-dsp
• -mdspr2
• -mno-dspr2
• -mmcu
• -mmno-mcu
• -meva
• -mno-eva
• -mvirt
• -mno-virt
• -mxpa
• -mno-xpa
• -mcrc
• -mno-crc
• -mginv
• -mno-ginv
• -mmicromips
• -mno-micromips
• -mmsa
• -mno-msa
• -mloongson-mmi
• -mno-loongson-mmi
• -mloongson-ext
• -mno-loongson-ext
• -mloongson-ext2
• -mno-loongson-ext2
• -mfpu=fpu-type
• -msmartmips
• -mno-smartmips
• -mpaired-single
• -mno-paired-single
• -mdmx
• -mno-mdmx
• -mips3d
• -mno-mips3d
• -mmt
• -mno-mt
• -mllsc
• -mno-llsc
• -mlong64
• -mlong32
• -msym32
• -mno-sym32
• -Gnum
• -mlocal-sdata
• -mno-local-sdata
• -mextern-sdata
• -mno-extern-sdata
• -mgpopt
• -mno-gopt
• -membedded-data
• -mno-embedded-data
• -muninit-const-in-rodata
• -mno-uninit-const-in-rodata
• -mcode-readable=setting
• -msplit-addresses
• -mno-split-addresses
• -mexplicit-relocs
• -mno-explicit-relocs
• -mexplicit-relocs=release
• -mcheck-zero-division
• -mno-check-zero-division
• -mdivide-traps
• -mdivide-breaks
• -mload-store-pairs
• -mno-load-store-pairs
• -mstrict-align
• -mno-strict-align
• -mno-unaligned-access
• -munaligned-access
• -mmemcpy
• -mno-memcpy
• -mlong-calls
• -mno-long-calls
• -mmad
• -mno-mad
• -mimadd
• -mno-imadd
• -mfused-madd
• -mno-fused-madd
• -nocpp
• -mfix-24k
• -mno-fix-24k
• -mfix-r4000
• -mno-fix-r4000
• -mfix-r4400
• -mno-fix-r4400
• -mfix-r5900
• -mno-fix-r5900
• -mfix-r10000
• -mno-fix-r10000
• -mfix-rm7000
• -mno-fix-rm7000
• -mfix-vr4120
• -mno-fix-vr4120
• -mfix-vr4130
• -mno-fix-vr4130
• -mfix-sb1
• -mno-fix-sb1
• -mflush-func=func
• -mno-flush-func
• -mbranch-cost=num
• -mbranch-likely
• -mno-branch-likely
• -mcompact-branches=policy
• -mfp-exceptions
• -mno-fp-exceptions
• -mvr4130-align
• -mno-vr4130-align
• -msynci
• -mno-synci
• -mlxc1-sxc1
• -mno-lxc1-sxc1
• -mmadd4
• -mno-madd4
• -mrelax-pic-calls
• -mno-relax-pic-calls
• -mmcount-ra-address
• -mframe-header-opt
• -mno-frame-header-opt

like. that's the whole list. no cap. 💀

---

📚 MMIX Options (MMIX Options)

• -mlibfuncs
• -mno-libfuncs
• -mepsilon
• -mno-epsilon
• -mabi=gnu
• -mabi=mmixware
• -mzero-extend
• -mknuthdiv
• -mtoplevel-symbols
• -melf
• -mbranch-predict
• -mno-branch-predict
• -mbase-addresses
• -mno-base-addresses
• -msingle-exit
• -mno-single-exit

---

🌸 MN10300 Options (MN10300 Options)

• -mmult-bug
• -mno-mult-bug
• -mno-am33
• -mam33
• -mam33-2
• -mam34
• -mtune=cpu-type
• -mreturn-pointer-on-d0
• -mno-crt0
• -mrelax
• -mliw
• -msetlb

---

🦋 Moxie Options (Moxie Options)

literally four options. she's tiny, she's cute, she's Moxie, obsessed 🎀:

• -meb
• -mel
• -mmul.x
• -mno-crt0

---

🧚 MSP430 Options (MSP430 Options)

• -msim
• -masm-hex
• -mmcu=
• -mcpu=
• -mlarge
• -msmall
• -mrelax
• -mwarn-mcu
• -mcode-region=
• -mdata-region=
• -msilicon-errata=
• -msilicon-errata-warn=
• -mhwmult=
• -minrt
• -mtiny-printf
• -mmax-inline-shift=

---

🧵 NDS32 Options (NDS32 Options)

• -mbig-endian
• -mlittle-endian
• -mreduced-regs
• -mfull-regs
• -mcmov
• -mno-cmov
• -mext-perf
• -mno-ext-perf
• -mext-perf2
• -mno-ext-perf2
• -mext-string
• -mno-ext-string
• -mv3push
• -mno-v3push
• -m16bit
• -mno-16bit
• -misr-vector-size=num
• -mcache-block-size=num
• -march=arch
• -mcmodel=code-model
• -mctor-dtor
• -mrelax

---

🎮 Nvidia PTX Options (Nvidia PTX Options)

three flags, main character energy, that's the whole vibe ✨:

• -m64
• -mmainkernel
• -moptimize

---

🌈 OpenRISC Options (OpenRISC Options)

• -mboard=name
• -mnewlib
• -mhard-mul
• -mhard-div
• -msoft-mul
• -msoft-div
• -msoft-float
• -mhard-float
• -mdouble-float
• -munordered-float
• -mcmov
• -mror
• -mrori
• -msext
• -msfimm
• -mshftimm
• -mcmodel=code-model

---

🕹️ PDP-11 Options (PDP-11 Options)

she's vintage, she's iconic, she's literally older than everyone here and she's still serving 👑:

• -mfpu
• -msoft-float
• -mac0
• -mno-ac0
• -m40
• -m45
• -m10
• -mint32
• -mno-int16
• -mint16
• -mno-int32
• -msplit
• -munix-asm
• -mdec-asm
• -mgnu-asm
• -mlra

---

💜 PowerPC Options (PowerPC Options)

PowerPC Options — See RS/6000 and PowerPC Options.

like, she literally points you to her sister's section. go scroll, bestie 🫶

---

⚙️ PRU Options (PRU Options)

• -mmcu=mcu
• -minrt
• -mno-relax
• -mloop
• -mabi=variant

---

🌺 RISC-V Options (RISC-V Options)

okay RISC-V is that girl right now and she knows it, and yes, there are a LOT of flags here too, all of them, zero skipped:

• -mbranch-cost=N-instruction
• -mplt
• -mno-plt
• -mabi=ABI-string
• -mfdiv
• -mno-fdiv
• -mfence-tso
• -mno-fence-tso
• -mdiv
• -mno-div
• -misa-spec=ISA-spec-string
• -march=ISA-string
• -mtune=processor-string
• -mpreferred-stack-boundary=num
• -msmall-data-limit=N-bytes
• -msave-restore
• -mno-save-restore
• -mshorten-memrefs
• -mno-shorten-memrefs
• -mstrict-align
• -mno-strict-align
• -mcmodel=medlow
• -mcmodel=medany
• -mcmodel=large
• -mexplicit-relocs
• -mno-explicit-relocs
• -mrelax
• -mno-relax
• -mriscv-attribute
• -mno-riscv-attribute
• -malign-data=type
• -mbig-endian
• -mlittle-endian
• -mstack-protector-guard=guard
• -mstack-protector-guard-reg=reg
• -mstack-protector-guard-offset=offset
• -mcsr-check
• -mno-csr-check
• -mmovcc
• -mno-movcc
• -minline-atomics
• -mno-inline-atomics
• -minline-strlen
• -mno-inline-strlen
• -minline-strcmp
• -mno-inline-strcmp
• -minline-strncmp
• -mno-inline-strncmp
• -mtls-dialect=desc
• -mtls-dialect=trad

---

🍓 RL78 Options (RL78 Options)

• -msim
• -mmul=none
• -mmul=g13
• -mmul=g14
• -mallregs
• -mcpu=g10
• -mcpu=g13
• -mcpu=g14
• -mg10
• -mg13
• -mg14
• -m64bit-doubles
• -m32bit-doubles
• -msave-mduc-in-interrupts

---

👑 RS/6000 and PowerPC Options (RS/6000 and PowerPC Options)

and here she is. the big one. the icon, the moment, the main character of machine-dependent options 💅 buckle up bestie, this is the longest list on the whole page and every single flag is coming with us:

• -mcpu=cpu-type
• -mtune=cpu-type
• -mcmodel=code-model
• -mpowerpc64
• -maltivec
• -mno-altivec
• -mpowerpc-gpopt
• -mno-powerpc-gpopt
• -mpowerpc-gfxopt
• -mno-powerpc-gfxopt
• -mmfcrf
• -mno-mfcrf
• -mpopcntb
• -mno-popcntb
• -mpopcntd
• -mno-popcntd
• -mfprnd
• -mno-fprnd
• -mcmpb
• -mno-cmpb
• -mhard-dfp
• -mno-hard-dfp
• -mfull-toc
• -mminimal-toc
• -mno-fp-in-toc
• -mno-sum-in-toc
• -m64
• -m32
• -mxl-compat
• -mno-xl-compat
• -mpe
• -malign-power
• -malign-natural
• -msoft-float
• -mhard-float
• -mmultiple
• -mno-multiple
• -mupdate
• -mno-update
• -mavoid-indexed-addresses
• -mno-avoid-indexed-addresses
• -mfused-madd
• -mno-fused-madd
• -mbit-align
• -mno-bit-align
• -mstrict-align
• -mno-strict-align
• -mrelocatable
• -mno-relocatable
• -mrelocatable-lib
• -mno-relocatable-lib
• -mtoc
• -mno-toc
• -mlittle
• -mlittle-endian
• -mbig
• -mbig-endian
• -mdynamic-no-pic
• -mswdiv
• -msingle-pic-base
• -mprioritize-restricted-insns=priority
• -msched-costly-dep=dependence_type
• -minsert-sched-nops=scheme
• -mcall-aixdesc
• -mcall-eabi
• -mcall-freebsd
• -mcall-linux
• -mcall-netbsd
• -mcall-openbsd
• -mcall-sysv
• -mcall-sysv-eabi
• -mcall-sysv-noeabi
• -mtraceback=traceback_type
• -maix-struct-return
• -msvr4-struct-return
• -mabi=abi-type
• -msecure-plt
• -mbss-plt
• -msplit-patch-nops
• -mlongcall
• -mno-longcall
• -mpltseq
• -mno-pltseq
• -mblock-move-inline-limit=num
• -mblock-compare-inline-limit=num
• -mblock-compare-inline-loop-limit=num
• -mno-block-ops-unaligned-vsx
• -mstring-compare-inline-limit=num
• -misel
• -mno-isel
• -mvrsave
• -mno-vrsave
• -mmulhw
• -mno-mulhw
• -mdlmzb
• -mno-dlmzb
• -mprototype
• -mno-prototype
• -msim
• -mmvme
• -mads
• -myellowknife
• -memb
• -msdata
• -msdata=opt
• -mreadonly-in-sdata
• -mvxworks
• -G num
• -mrecip
• -mrecip=opt
• -mno-recip
• -mrecip-precision
• -mno-recip-precision
• -mveclibabi=type
• -mfriz
• -mno-friz
• -mpointers-to-nested-functions
• -mno-pointers-to-nested-functions
• -msave-toc-indirect
• -mno-save-toc-indirect
• -mpower8-fusion
• -mno-mpower8-fusion
• -mcrypto
• -mno-crypto
• -mhtm
• -mno-htm
• -mquad-memory
• -mno-quad-memory
• -mquad-memory-atomic
• -mno-quad-memory-atomic
• -mcompat-align-parm
• -mno-compat-align-parm
• -mfloat128
• -mno-float128
• -mfloat128-hardware
• -mno-float128-hardware
• -mgnu-attribute
• -mno-gnu-attribute
• -mstack-protector-guard=guard
• -mstack-protector-guard-reg=reg
• -mstack-protector-guard-offset=offset
• -mprefixed
• -mno-prefixed
• -mpcrel
• -mno-pcrel
• -mmma
• -mno-mmma
• -mrop-protect
• -mno-rop-protect
• -mprivileged
• -mno-privileged

okay that one was a lot. drink some water bestie, you earned it 🍓

RX Options (RX Options)

Bestie, if you're targeting RX, these flags are your whole personality ✨

• -m64bit-doubles
• -m32bit-doubles
• -fpu
• -nofpu
• -mcpu=
• -mbig-endian-data
• -mlittle-endian-data
• -msmall-data
• -msim
• -mno-sim
• -mas100-syntax
• -mno-as100-syntax
• -mrelax
• -mmax-constant-size=
• -mint-register=
• -mpid
• -mallow-string-insns
• -mno-allow-string-insns
• -mjsr
• -mno-warn-multiple-fast-interrupts
• -msave-acc-in-interrupts

S/390 and zSeries Options (S/390 and zSeries Options)

S/390 and zSeries girlies, this one's for you 🦋

• -mtune=cpu-type
• -march=cpu-type
• -mhard-float
• -msoft-float
• -mhard-dfp
• -mno-hard-dfp
• -mlong-double-64
• -mlong-double-128
• -mbackchain
• -mno-backchain
• -mpacked-stack
• -mno-packed-stack
• -msmall-exec
• -mno-small-exec
• -mmvcle
• -mno-mvcle
• -m64
• -m31
• -mdebug
• -mno-debug
• -mesa
• -mzarch
• -mhtm
• -mvx
• -mzvector
• -mtpf-trace
• -mno-tpf-trace
• -mtpf-trace-skip
• -mno-tpf-trace-skip
• -mfused-madd
• -mno-fused-madd
• -mwarn-framesize
• -mwarn-dynamicstack
• -mstack-size
• -mstack-guard
• -mhotpatch=halfwords,halfwords

SH Options (SH Options)

SH crew, let's go 💅

• -m1
• -m2
• -m2e
• -m2a-nofpu
• -m2a-single-only
• -m2a-single
• -m2a
• -m3
• -m3e
• -m4-nofpu
• -m4-single-only
• -m4-single
• -m4
• -m4a-nofpu
• -m4a-single-only
• -m4a-single
• -m4a
• -m4al
• -mb
• -ml
• -mdalign
• -mrelax
• -mbigtable
• -mfmovd
• -mrenesas
• -mno-renesas
• -mnomacsave
• -mieee
• -mno-ieee
• -mbitops
• -misize
• -minline-ic_invalidate
• -mpadstruct
• -mprefergot
• -musermode
• -multcost=number
• -mdiv=strategy
• -mdivsi3_libfunc=name
• -mfixed-range=register-range
• -maccumulate-outgoing-args
• -matomic-model=atomic-model
• -mbranch-cost=num
• -mzdcbranch
• -mno-zdcbranch
• -mcbranch-force-delay-slot
• -mfused-madd
• -mno-fused-madd
• -mfsca
• -mno-fsca
• -mfsrra
• -mno-fsrra
• -mpretend-cmove
• -mtas

Solaris 2 Options (Solaris 2 Options)

Solaris 2 besties, we see you 🌸

• -mclear-hwcap
• -mno-clear-hwcap
• -mimpure-text
• -mno-impure-text
• -pthreads

SPARC Options (SPARC Options)

SPARC girlies, this is your moment ✨

• -mcpu=cpu-type
• -mtune=cpu-type
• -mcmodel=code-model
• -mmemory-model=mem-model
• -m32
• -m64
• -mapp-regs
• -mno-app-regs
• -mfaster-structs
• -mno-faster-structs
• -mflat
• -mno-flat
• -mfpu
• -mno-fpu
• -mhard-float
• -msoft-float
• -mhard-quad-float
• -msoft-quad-float
• -mstack-bias
• -mno-stack-bias
• -mstd-struct-return
• -mno-std-struct-return
• -munaligned-doubles
• -mno-unaligned-doubles
• -muser-mode
• -mno-user-mode
• -mv8plus
• -mno-v8plus
• -mvis
• -mno-vis
• -mvis2
• -mno-vis2
• -mvis3
• -mno-vis3
• -mvis3b
• -mno-vis3b
• -mvis4
• -mno-vis4
• -mvis4b
• -mno-vis4b
• -mcbcond
• -mno-cbcond
• -mfmaf
• -mno-fmaf
• -mfsmuld
• -mno-fsmuld
• -mpopc
• -mno-popc
• -msubxc
• -mno-subxc
• -mfix-at697f
• -mfix-ut699
• -mfix-ut700
• -mfix-gr712rc

System V Options (System V Options)

System V vibes, short and sweet 💜

• -Qy
• -Qn
• -YP,paths
• -Ym,dir

V850 Options (V850 Options)

V850 cuties, here's your flag collection 🎀

• -mlong-calls
• -mno-long-calls
• -mep
• -mno-ep
• -mprolog-function
• -mno-prolog-function
• -mspace
• -mtda=n
• -msda=n
• -mzda=n
• -mapp-regs
• -mno-app-regs
• -mdisable-callt
• -mno-disable-callt
• -mv850e2v3
• -mv850e2
• -mv850e1
• -mv850es
• -mv850e
• -mv850
• -mv850e3v5
• -mloop
• -mrelax
• -mlong-jumps
• -msoft-float
• -mhard-float
• -mgcc-abi
• -mrh850-abi
• -mbig-switch

VAX Options (VAX Options)

VAX girlies, let's go back in time 💖

• -munix
• -mgnu
• -md
• -md-float
• -mg
• -mg-float
• -mlra

Visium Options (Visium Options)

Visium besties, this is your list 🌈

• -mdebug
• -msim
• -mfpu
• -mno-fpu
• -mhard-float
• -msoft-float
• -mcpu=cpu-type
• -mtune=cpu-type
• -msv-mode
• -muser-mode

VMS Options (VMS Options)

VMS vibes, keep it moving ✨

• -mvms-return-codes
• -mdebug-main=prefix
• -mmalloc64
• -mpointer-size=size

VxWorks Options (VxWorks Options)

VxWorks girlies, real-time energy 🔥

• -mrtp
• -msmp
• -non-static
• -Bstatic
• -Bdynamic
• -Xbind-lazy
• -Xbind-now

x86 Options (x86 Options)

Okay x86 bestie, this is the big one — grab a snack 🍓

• -mtune=cpu-type
• -march=cpu-type
• -mtune-ctrl=feature-list
• -mdump-tune-features
• -mno-default
• -mfpmath=unit
• -masm=dialect
• -mno-fancy-math-387
• -mno-fp-ret-in-387
• -m80387
• -mhard-float
• -msoft-float
• -mno-wide-multiply
• -mrtd
• -malign-double
• -mpreferred-stack-boundary=num
• -mincoming-stack-boundary=num
• -mcld
• -mcx16
• -msahf
• -mmovbe
• -mcrc32
• -mmwait
• -mrecip
• -mrecip=opt
• -mvzeroupper
• -mprefer-avx128
• -mprefer-vector-width=opt
• -mpartial-vector-fp-math
• -mmove-max=bits
• -mstore-max=bits
• -mnoreturn-no-callee-saved-registers
• -mmmx
• -msse
• -msse2
• -msse3
• -mssse3
• -msse4.1
• -msse4.2
• -msse4
• -mavx
• -mavx2
• -mavx512f
• -mavx512cd
• -mavx512vl
• -mavx512bw
• -mavx512dq
• -mavx512ifma
• -mavx512vbmi
• -msha
• -maes
• -mpclmul
• -mfsgsbase
• -mrdrnd
• -mf16c
• -mfma
• -mpconfig
• -mwbnoinvd
• -mptwrite
• -mclflushopt
• -mclwb
• -mxsavec
• -mxsaves
• -msse4a
• -m3dnow
• -m3dnowa
• -mpopcnt
• -mabm
• -mbmi
• -mtbm
• -mfma4
• -mxop
• -madx
• -mlzcnt
• -mbmi2
• -mfxsr
• -mxsave
• -mxsaveopt
• -mrtm
• -mhle
• -mlwp
• -mmwaitx
• -mclzero
• -mpku
• -mthreads
• -mgfni
• -mvaes
• -mwaitpkg
• -mshstk
• -mmanual-endbr
• -mcet-switch
• -mforce-indirect-call
• -mavx512vbmi2
• -mavx512bf16
• -menqcmd
• -mvpclmulqdq
• -mavx512bitalg
• -mmovdiri
• -mmovdir64b
• -mavx512vpopcntdq
• -mavx512vnni
• -mprfchw
• -mrdpid
• -mrdseed
• -msgx
• -mavx512vp2intersect
• -mserialize
• -mtsxldtrk
• -mamx-tile
• -mamx-int8
• -mamx-bf16
• -muintr
• -mhreset
• -mavxvnni
• -mamx-fp8
• -mavx512fp16
• -mavxifma
• -mavxvnniint8
• -mavxneconvert
• -mcmpccxadd
• -mamx-fp16
• -mprefetchi
• -mraoint
• -mamx-complex
• -mavxvnniint16
• -msm3
• -msha512
• -msm4
• -mapxf
• -musermsr
• -mavx10.1
• -mavx10.1-256
• -mavx10.1-512
• -mevex512
• -mavx10.2
• -mamx-avx512
• -mamx-tf32
• -mamx-transpose
• -mmovrs
• -mamx-movrs
• -mcldemote
• -mms-bitfields
• -mno-align-stringops
• -minline-all-stringops
• -minline-stringops-dynamically
• -mstringop-strategy=alg
• -mkl
• -mwidekl
• -mmemcpy-strategy=strategy
• -mmemset-strategy=strategy
• -mpush-args
• -maccumulate-outgoing-args
• -m128bit-long-double
• -m96bit-long-double
• -mlong-double-64
• -mlong-double-80
• -mlong-double-128
• -mregparm=num
• -msseregparm
• -mveclibabi=type
• -mvect8-ret-in-mem
• -mpc32
• -mpc64
• -mpc80
• -mdaz-ftz
• -mstackrealign
• -momit-leaf-frame-pointer
• -mno-red-zone
• -mno-tls-direct-seg-refs
• -mcmodel=code-model
• -mabi=name
• -maddress-mode=mode
• -m32
• -m64
• -mx32
• -m16
• -miamcu
• -mlarge-data-threshold=num
• -msse2avx
• -mfentry
• -mrecord-mcount
• -mnop-mcount
• -m8bit-idiv
• -minstrument-return=type
• -mfentry-name=name
• -mfentry-section=name
• -mavx256-split-unaligned-load
• -mavx256-split-unaligned-store
• -malign-data=type
• -mstack-protector-guard=guard
• -mstack-protector-guard-reg=reg
• -mstack-protector-guard-offset=offset
• -mstack-protector-guard-symbol=symbol
• -mgeneral-regs-only
• -mcall-ms2sysv-xlogues
• -mrelax-cmpxchg-loop
• -mindirect-branch=choice
• -mfunction-return=choice
• -mindirect-branch-register
• -mharden-sls=choice
• -mindirect-branch-cs-prefix
• -mneeded
• -mno-direct-extern-access
• -munroll-only-small-loops
• -mlam=choice

x86 Windows Options

See Cygwin and MinGW Options 💖

Xstormy16 Options (Xstormy16 Options)

Xstormy16 besties, just one flag, slay 💅

• -msim

Xtensa Options (Xtensa Options)

Xtensa girlies, here's your set ✨

• -mconst16
• -mno-const16
• -mfused-madd
• -mno-fused-madd
• -mforce-no-pic
• -mserialize-volatile
• -mno-serialize-volatile
• -mtext-section-literals
• -mno-text-section-literals
• -mauto-litpools
• -mno-auto-litpools
• -mtarget-align
• -mno-target-align
• -mlongcalls
• -mno-longcalls
• -mabi=abi-type
• -mextra-l32r-costs=cycles
• -mstrict-align
• -mno-strict-align

zSeries Options

See S/390 and zSeries Options 🌸

---

Options Controlling the Kind of Output

Okay bestie, now we're talking about the ~vibes~ of output. Compilation can involve up to four stages: preprocessing, compilation proper, assembly, and linking — and they always happen in that exact order, no cap 💅. GCC can preprocess and compile several files either into several assembler input files, or into one assembler input file. Then each assembler input file produces an object file, and linking combines all the object files — the newly compiled ones and the ones you specified as input — into an executable file. It's giving pipeline realness ✨.

For any given input file, the file name suffix is literally the vibe check for what kind of compilation happens:

• file.c — C source code that must be preprocessed.
• file.i — C source code that should not be preprocessed.
• file.ii — C++ source code that should not be preprocessed.
• file.m — Objective-C source code. Note that you must link with the libobjc library to make an Objective-C program work.
• file.mi — Objective-C source code that should not be preprocessed.
• file.mm, file.M — Objective-C++ source code. Note that you must link with the libobjc library to make an Objective-C++ program work. Note that .M refers to a literal capital M.
• file.mii — Objective-C++ source code that should not be preprocessed.
• file.h — C, C++, Objective-C or Objective-C++ header file to be turned into a precompiled header (default), or C, C++ header file to be turned into an Ada spec (via the -fdump-ada-spec switch).
• file.cc, file.cp, file.cxx, file.cpp, file.CPP, file.c++, file.C — C++ source code that must be preprocessed. Note that in .cxx, the last two letters must both be literally x. Likewise, .C refers to a literal capital C.
• file.mm, file.M — Objective-C++ source code that must be preprocessed.
• file.mii — Objective-C++ source code that should not be preprocessed.
• file.hh, file.H, file.hp, file.hxx, file.hpp, file.HPP, file.h++, file.tcc — C++ header file to be turned into a precompiled header or Ada spec.
• file.f, file.for, file.ftn, file.fi — Fixed form Fortran source code that should not be preprocessed.
• file.F, file.FOR, file.fpp, file.FPP, file.FTN — Fixed form Fortran source code that must be preprocessed (with the traditional preprocessor).
• file.f90, file.f95, file.f03, file.f08, file.fii — Free form Fortran source code that should not be preprocessed.
• file.F90, file.F95, file.F03, file.F08 — Free form Fortran source code that must be preprocessed (with the traditional preprocessor).
• file.cob, file.COB, file.cbl, file.CBL — COBOL source code.
• file.go — Go source code.
• file.d — D source code.
• file.di — D interface file.
• file.dd — D documentation code (Ddoc).
• file.ads — Ada source code file that contains a library unit declaration (a declaration of a package, subprogram, or generic, or a generic instantiation), or a library unit renaming declaration (a package, generic, or subprogram renaming declaration). Such files are also called specs.
• file.adb — Ada source code file containing a library unit body (a subprogram or package body). Such files are also called bodies.
• file.s — Assembler code.
• file.S, file.sx — Assembler code that must be preprocessed.
• other — An object file to be fed straight into linking. Any file name with no recognized suffix is treated this way.

Okay bestie so gcc is scarily good at guessing what language your files are in just from the suffix, but you can absolutely override that with -x and it's giving main character energy:

-x language 💅

This one says, out loud, what language the input files that come after it are in — instead of letting the compiler pick a default from the file name suffix. It applies to every following input file until the next -x shows up. The possible values for language are literally:

c  c-header  cpp-output
c++  c++-header  c++-system-header c++-user-header c++-cpp-output
objective-c  objective-c-header  objective-c-cpp-output  objc-cpp-output
objective-c++  objective-c++-header  objective-c++-cpp-output
objc++-cpp-output
assembler  assembler-with-cpp
ada  adascil  adawhy
cobol
d
f77  f77-cpp-input  f95  f95-cpp-input
go
modula-2  modula-2-cpp-output
rust
lto

⚠️ Important, no cap: -x does not imply a particular language standard. For example -x f77 may also require -std=legacy for some older source codes. It's not a package deal, babes.

-x none 🙈

Turns off any specification of a language, so that all the subsequent files get handled according to their file name suffixes — exactly as if -x had not been used at all. It's the undo button, the "nvm bestie" of options.

If you only want some of the stages of compilation, you can use -x (or the file name suffixes) to tell gcc where to start, and then one of -c, -S, or -E to say where gcc is supposed to stop.

Heads up: some combinations — for example, -x cpp-output -E — instruct gcc to do nothing at all. Like, literally nothing. Iconic of it, honestly.

-c 🧊

Compile or assemble the source files, but do not link. The linking stage simply does not happen. The ultimate output is an object file for each source file.

By default, the object file name for a source file is made by replacing the suffix .c, .i, .s, etc., with .o.

Unrecognized input files, that don't require compilation or assembly, are just ignored. gcc said "not my problem" 💀

-S 📜

Stop right after the stage of compilation proper; do not assemble. The output is an assembler code file for each non-assembler input file you specified.

By default, the assembler file name for a source file is made by replacing the suffix .c, .i, etc., with .s.

Input files that don't require compilation are ignored.

-E 🫧

Stop after the preprocessing stage; do not run the compiler proper. The output is preprocessed source code, and it gets sent to standard output.

Input files that don't require preprocessing are ignored.

-o file ✨

Places the primary output in file. This applies to whatever sort of output is being produced — whether that's an executable file, an object file, an assembler file or preprocessed C code. One option, every vibe.

If -o is not specified, the defaults are:

• the executable file goes in a.out
• the object file for source.suffix goes in source.o
• its assembler file goes in source.s
• a precompiled header file goes in source.suffix.gch
• and all preprocessed C source goes on standard output

Now here's the part where it gets extra: even though -o names only the primary output, it also affects the naming of auxiliary and dump outputs. See the examples below. Unless overridden, both auxiliary outputs and dump outputs are placed in the same directory as the primary output.

• In auxiliary outputs, the suffix of the input file is replaced with that of the auxiliary output file type.
• In dump outputs, the suffix of the dump file is appended to the input file suffix.
• In compilation commands, the base name of both auxiliary and dump outputs is that of the primary output.
• In compile and link commands, the primary output name, minus the executable suffix, is combined with the input file name.
• If both share the same base name (disregarding the suffix), the result of the combination is that base name. Otherwise, they are concatenated, separated by a dash.

gcc -c foo.c ...

will use foo.o as the primary output, and place aux outputs and dumps next to it — e.g., aux file foo.dwo for -gsplit-dwarf, and dump file foo.c.???r.final for -fdump-rtl-final.

If a non-linker output file is explicitly specified, aux and dump files by default take the same base name:

gcc -c foo.c -o dir/foobar.o ...

will name aux outputs dir/foobar.* and dump outputs dir/foobar.c.*.

A linker output will instead prefix aux and dump outputs:

gcc foo.c bar.c -o dir/foobar ...

will generally name aux outputs dir/foobar-foo.* and dir/foobar-bar.*, and dump outputs dir/foobar-foo.c.* and dir/foobar-bar.c.*.

The one exception to all of the above is when the executable shares the base name with the single input:

gcc foo.c -o dir/foo ...

in which case aux outputs are named dir/foo.* and dump outputs are named dir/foo.c.*.

The location and the names of auxiliary and dump outputs can be adjusted by the options -dumpbase, -dumpbase-ext, -dumpdir, -save-temps=cwd, and -save-temps=obj.

---

-dumpbase dumpbase 🎀

This sets the base name for auxiliary and dump output files. It does not affect the name of the primary output file.

Intermediate outputs, when preserved, are not regarded as primary outputs — they're auxiliary outputs:

gcc -save-temps -S foo.c

saves the (no longer) temporary preprocessed file in foo.i, and then compiles to the (implied) output file foo.s. Whereas:

gcc -save-temps -dumpbase save-foo -c foo.c

preprocesses to save-foo.i, compiles to save-foo.s (now an intermediate, thus an auxiliary output), and then assembles to the (implied) output file foo.o.

Absent this option, dump and aux files take their names from the input file — or from the (non-linker) output file, if one is explicitly specified. Dump output files (e.g. those requested by -fdump-* options) get the input name suffix, and aux output files (those requested by other non-dump options, e.g. -save-temps, -gsplit-dwarf, -fcallgraph-info) get it without the suffix.

Similar suffix differentiation of dump and aux outputs can be attained for an explicitly-given -dumpbase basename.suf by also specifying -dumpbase-ext .suf.

If dumpbase is explicitly specified with any directory component, any dumppfx specification (e.g. -dumpdir or -save-temps=*) is ignored, and instead of appending to it, dumpbase fully overrides it:

gcc foo.c -c -o dir/foo.o -dumpbase alt/foo \
  -dumpdir pfx- -save-temps=cwd ...

creates auxiliary and dump outputs named alt/foo.*, disregarding dir/ in -o, the ./ prefix implied by -save-temps=cwd, and pfx- in -dumpdir. It really said "I don't see anyone else here" 💅

When -dumpbase is specified in a command that compiles multiple inputs, or that compiles and then links, it may be combined with dumppfx, as specified under -dumpdir. Then, each input file is compiled using the combined dumppfx, and default values for dumpbase and auxdropsuf are computed for each input file:

gcc foo.c bar.c -c -dumpbase main ...

creates foo.o and bar.o as primary outputs, and avoids overwriting the auxiliary and dump outputs by using the dumpbase as a prefix, creating auxiliary and dump outputs named main-foo.* and main-bar.*.

An empty string specified as dumpbase avoids the influence of the output basename in the naming of auxiliary and dump outputs during compilation, computing default values:

gcc -c foo.c -o dir/foobar.o -dumpbase " ...

will name aux outputs dir/foo.* and dump outputs dir/foo.c.*. Note how their basenames are taken from the input name, but the directory still defaults to that of the output. Sneaky but cute.

The empty-string dumpbase does not prevent the use of the output basename for outputs during linking:

gcc foo.c bar.c -o dir/foobar -dumpbase " -flto ...

The compilation of the source files will name auxiliary outputs dir/foo.* and dir/bar.*, and dump outputs dir/foo.c.* and dir/bar.c.*. LTO recompilation during linking will use dir/foobar. as the prefix for dumps and auxiliary files.

-dumpbase-ext auxdropsuf 🌸

When forming the name of an auxiliary (but not a dump) output file, drop trailing auxdropsuf from dumpbase before appending any suffixes.

If not specified, this option defaults to the suffix of a default dumpbase — i.e., the suffix of the input file when -dumpbase is not present in the command line, or when dumpbase is combined with dumppfx.

gcc foo.c -c -o dir/foo.o -dumpbase x-foo.c -dumpbase-ext .c ...

creates dir/foo.o as the main output, and generates auxiliary outputs in dir/x-foo.*, taking the location of the primary output and dropping the .c suffix from the dumpbase. Dump outputs retain the suffix: dir/x-foo.c.*.

This option is disregarded if it does not match the suffix of a specified dumpbase — except as an alternative to the executable suffix when appending the linker output base name to dumppfx, as specified below:

gcc foo.c bar.c -o main.out -dumpbase-ext .out ...

creates main.out as the primary output, and avoids overwriting the auxiliary and dump outputs by using the executable name minus auxdropsuf as a prefix, creating auxiliary outputs named main-foo.* and main-bar.* and dump outputs named main-foo.c.* and main-bar.c.*.

-dumpdir dumppfx 🦋

When forming the name of an auxiliary or dump output file, use dumppfx as a prefix:

gcc -dumpdir pfx- -c foo.c ...

creates foo.o as the primary output, and auxiliary outputs named pfx-foo.*, combining the given dumppfx with the default dumpbase derived from the default primary output, derived in turn from the input name. Dump outputs also take the input name suffix: pfx-foo.c.*.

If dumppfx is to be used as a directory name, it must end with a directory separator:

gcc -dumpdir dir/ -c foo.c -o obj/bar.o ...

creates obj/bar.o as the primary output, and auxiliary outputs named dir/bar.*, combining the given dumppfx with the default dumpbase derived from the primary output name. Dump outputs also take the input name suffix: dir/bar.c.*.

It defaults to the location of the output file, unless the output file is a special file like /dev/null. Options -save-temps=cwd and -save-temps=obj override this default, just like an explicit -dumpdir option. In case multiple such options are given, the last one prevails:

gcc -dumpdir pfx- -c foo.c -save-temps=obj ...

...so it hands you back foo.o, and the auxiliary outputs are named foo.* — because -save-temps=* overrides the dumppfx that the earlier -dumpdir option set up. And like, it genuinely does not matter that =obj is the default for -save-temps, or that the output directory is implicitly the current directory. Dump outputs? Those come out as foo.c.*. 💅

Now when you're compiling from multiple input files and -dumpbase is in the mix, bestie, here's the tea: dumpbase, minus a auxdropsuf suffix, and a dash get appended to (or they override it, if they contain any directory components) an explicit or defaulted dumppfx — so that every one of those multiple compilations gets its own differently-named aux and dump outputs. Slay, no collisions. ✨

gcc foo.c bar.c -c -dumpdir dir/pfx- -dumpbase main ...

That one outputs auxiliary dumps to dir/pfx-main-foo.* and dir/pfx-main-bar.*, appending dumpbase- to dumppfx. And the dump outputs keep the input file suffix: dir/pfx-main-foo.c.* and dir/pfx-main-bar.c.*, respectively. ✨

Now contrast that with the single-input compilation:

gcc foo.c -c -dumpdir dir/pfx- -dumpbase main ...

That one applies -dumpbase to a single source, so it does not compute and append a separate dumpbase per input file. Its auxiliary and dump outputs just go in dir/pfx-main.*.

---

When you're compiling and then linking from multiple input files, a defaulted or explicitly specified dumppfx also gets that whole dumpbase- transformation from above (think the compilation of foo.c and bar.c we just did, but without the -c). And if neither -dumpdir nor -dumpbase are given, then the linker output base name — minus auxdropsuf, if specified, or the executable suffix otherwise — plus a dash, gets appended to the default dumppfx instead. 💫

But heads up 🙈 — unlike the earlier cases of linking:

gcc foo.c bar.c -dumpdir dir/pfx- -o main ...

this does not append the output name main to dumppfx, because -dumpdir was explicitly specified. The whole goal here is that the explicitly-specified dumppfx may contain the specified output name as part of the prefix, if you want it that way. Only an explicitly-specified -dumpbase would be combined with it — and that's purely so we don't just toss out a meaningful option like it's nothing. 💀

And when you're compiling and then linking from a single input file, the linker output base name only gets appended to the default dumppfx (same as above) if it does not share the base name with that single input file name. This has been covered in the single-input linking cases above, obvi — but not with an explicit -dumpdir, because that inhibits the combination, even if it gets overridden by -save-temps=*:

gcc foo.c -dumpdir alt/pfx- -o dir/main.exe -save-temps=cwd ...

Auxiliary outputs are named foo.*, and dump outputs are foo.c.*, in the current working directory as ultimately requested by -save-temps=cwd. iconic behavior honestly

---

Summing it all up for an intuitive though slightly imprecise data flow: the primary output name is broken into a directory part and a basename part; dumppfx is set to the former, unless it's overridden by -dumpdir or -save-temps=*, and dumpbase is set to the latter, unless -dumpbase overrides it. If there are multiple inputs or linking, this dumpbase may be combined with dumppfx and taken from each input file. Auxiliary output names for each input are formed by combining dumppfx, dumpbase minus suffix, and the auxiliary output suffix; dump output names are only different in that the suffix from dumpbase is retained. Simple. Kinda. Not really. 🙈

When it comes to auxiliary and dump outputs created during LTO recompilation, a combination of dumppfx and dumpbase — as given, or as derived from the linker output name but not from inputs, even in cases in which this combination would not otherwise be used as such — is passed down with a trailing period replacing the compiler-added dash, if any, as a -dumpdir option to lto-wrapper. And because it's involved in linking, this program does not normally get any -dumpbase and -dumpbase-ext, and it ignores them. Rude but fair. ✨

When running sub-compilers, lto-wrapper appends LTO stage names to the received dumppfx, makes sure it contains a directory component so that it overrides any -dumpdir, and passes that as -dumpbase to sub-compilers.

---

• -v — Prints (on standard error output) the actual commands that got executed to run the stages of compilation. It also prints the version number of the compiler driver program, and of the preprocessor and the compiler proper. Basically: receipts. 🧾
• -### — Like -v, except the commands are not executed and arguments are quoted unless they contain only alphanumeric characters or "./-_". Super useful for shell scripts that want to capture the driver-generated command lines. chef's kiss for scripting 💖
• --help — Prints (on the standard output) a description of the command-line options understood by gcc. If -v is also specified, then --help is also passed on to the various processes invoked by gcc, so they can display the command-line options they accept too. And if -Wextra has also been specified (prior to the --help option), then command-line options that have no documentation associated with them are also displayed. Spicy. 🌶️
• --target-help — Prints (on the standard output) a description of target-specific command-line options for each tool. For some targets, extra target-specific information may also be printed. 🎯
• --help={class|[^]qualifier}[,...] — Prints (on the standard output) a description of the command-line options understood by the compiler that fit into all the specified classes and qualifiers. These are the supported classes:
  • optimizers — Display all of the optimization options supported by the compiler.
  • warnings — Display all of the options controlling warning messages produced by the compiler.
  • target — Display target-specific options. Unlike the --target-help option though, target-specific options of the linker and assembler are not displayed. This is because those tools do not currently support the extended --help= syntax.
  • params — Display the values recognized by the --param option.
  • language — Display the options supported for language, where language is the name of one of the languages supported in this version of GCC. If an option is supported by all languages, you need to select the common class instead.
  • common — Display the options that are common to all languages.

These are the supported qualifiers:

• undocumented — Display only those options that are undocumented.
• joined — Display options taking an argument that appears after an equal sign in the same continuous piece of text, such as: --help=target.
• separate — Display options taking an argument that appears as a separate word following the original option, such as: -o output-file.

So for example, to display all the undocumented target-specific switches supported by the compiler, use:

--help=target,undocumented

You can invert the sense of a qualifier by prefixing it with the ^ character — so for example, to display all binary warning options (i.e., ones that are either on or off and that do not take an argument) that have a description, use:

--help=warnings,^joined,^undocumented

The argument to --help= should not consist solely of inverted qualifiers. Don't do it bestie. 💀

Combining several classes is possible, although this usually restricts the output so much that there's literally nothing to display. One case where it does work, however, is when one of the classes is target. For example, to display all the target-specific optimization options, use:

--help=target,optimizers

The --help= option can be repeated on the command line. Each successive use displays its requested class of options, skipping those that have already been displayed. And if --help is also specified anywhere on the command line, then that takes precedence over any --help= option. 👑

If the -Q option appears on the command line before the --help= option, then the descriptive text displayed by --help= changes. Instead of describing the displayed options, an indication is given as to whether the option is enabled, disabled or set to a specific value (assuming that the compiler knows this at the point where the --help= option is used).

Here's a truncated example from the ARM port of gcc:

% gcc -Q -mabi=2 --help=target -c
The following options are target specific:
-mabi=                                2
-mabort-on-noreturn                   [disabled]
-mapcs                                [disabled]

The output is sensitive to the effects of previous command-line options, so for example it's possible to find out which optimizations are enabled at -O2 by using:

-Q -O2 --help=optimizers

Or alternatively, you can discover which binary optimizations are enabled by -O3 by using:

gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
diff /tmp/O2-opts /tmp/O3-opts | grep enabled

diff and grep, the original detective duo 🔍

• --version — Display the version number and copyrights of the invoked GCC. ✨
• -pass-exit-codes — Normally, the gcc program exits with the code of 1 if any phase of the compiler returns a non-success return code. But if you specify -pass-exit-codes, the gcc program instead returns with the numerically highest error produced by any phase returning an error indication. And FYI, the C, C++, and Fortran front ends return 4 if an internal compiler error is encountered. 📊
• -pipe — Use pipes rather than temporary files for communication between the various stages of compilation. This fails to work on some systems where the assembler is unable to read from a pipe — but the GNU assembler has no trouble at all. 💅
• -specs=file — Process file after the compiler reads in the standard specs file, in order to override the defaults which the gcc driver program uses when determining what switches to pass to cc1, cc1plus, as, ld, etc. You can specify more than one -specs=file on the command line, and they're processed in order, left to right. 📜
• -wrapper — Invoke all subcommands under a wrapper program. The name of the wrapper program and its parameters are passed as a comma separated list.

gcc -c t.c -wrapper gdb,--args

This invokes all subprograms of gcc under gdb --args, thus the invocation of cc1 is gdb --args cc1 .... debugging inception, we love to see it 🐛

• -ffile-prefix-map=old=new — okay so when you're compiling files that live in directory old, this makes the compiler record every reference to them in the output as if they actually lived in directory new instead 💫. Giving this one option is literally the same as handing over all the individual -f*-prefix-map options at once, no cap. It's the move for making reproducible builds that don't care where they were built. Little heads up tho: directories referenced by directives are not affected by these options 🙈. Peep also -fmacro-prefix-map, -fdebug-prefix-map, -fprofile-prefix-map and -fcanon-prefix-map.
• -fcanon-prefix-map — so for the -f*-prefix-map options, normally the comparison of the old prefix against the filename that'd normally show up in the result is done with plain textual comparison of the prefixes (or, on case-insensitive filesystems, ignoring letter case, and on DOS-based filesystems treating slashes and backslashes as equal). This option makes those comparisons happen on canonicalized paths of old and the referenced filename instead. Slay ✨.
• -fplugin=name.so — loads the plugin code sitting in file name.so, which is assumed to be a shared object that the compiler will dlopen 🫶. The base name of that shared object file is what identifies the plugin for argument parsing (see -fplugin-arg-name-key=value right below). Every plugin is expected to define the callback functions laid out in the Plugins API.
• -fplugin-arg-name-key=value — defines an argument called key with a value of value for the plugin called name 💅.
• -fdump-ada-spec[-slim] — for C and C++ source and include files, this generates the corresponding Ada specs 🌸.
• -fada-spec-parent=unit — tagging along with -fdump-ada-spec[-slim] up above, this generates Ada specs as child units of parent unit.
• -fdump-go-spec=file — for input files in any language, this generates the corresponding Go declarations into file 👑. Out come Go "const", "type", "var", and "func" declarations, which can be a genuinely useful way to start writing a Go interface to code written in some other language. Iconic, honestly.
• @file — reads command-line options from file 📄. The options that get read are inserted right in place of the original @file option. But lowkey important: if file doesn't exist, or can't be read, then the option gets treated literally and is not removed.

Options inside file are separated by whitespace 🌈. You can sneak a whitespace character into an option by wrapping the whole entire option in either single or double quotes. And literally any character (backslash included!) can be included by prefixing the character you want with a backslash. The file may even contain additional @file options of its own — any of those get processed recursively 🌀.

📚 Compiling C++ Programs 📚

C++ source files conventionally use one of these suffixes: .C, .cc, .cpp, .CPP, .c++, .cp, or .cxx ✨. C++ header files often use .hh, .hpp, .H, or (for shared template code) .tcc. Preprocessed C++ files use .ii, and C++20 module interface units sometimes use .ixx, .cppm, .cxxm, .c++m, or .ccm.

GCC recognizes files with these names and compiles them as C++ programs even if you call the compiler the exact same way you would for compiling C programs (usually with the name gcc) 💖.

However — and this is the part everybody trips on — using gcc does not add the C++ library 😭. g++ is a program that calls GCC and automatically specifies linking against the C++ library. It treats .c, .h and .i files as C++ source files instead of C source files unless -x is used. This program is also useful when precompiling a C header file with a .h extension for use in C++ compilations. On many systems, g++ is also installed under the name c++ 🎀.

When you compile C++ programs, you may specify many of the same command-line options that you use for compiling programs in any language; or command-line options meaningful for C and related languages; or options that are meaningful only for C++ programs 🦋.

---

🛠️ Options Controlling C Dialect 🛠️

The following options control the dialect of C (or languages derived from C, such as C++, Objective-C and Objective-C++) that the compiler accepts 💜:

-ansi

In C mode, this is equivalent to -std=c90. In C++ mode, it is equivalent to -std=c++98 ✨.

-std=

Determine the language standard 🎯. This option is currently only supported when compiling C or C++.

The compiler can accept several base standards, such as c90 or c++98, and GNU dialects of those standards, such as gnu90 or gnu++98 🌸. When a base standard is specified, the compiler accepts all programs following that standard plus those using GNU extensions that do not contradict it. For example, -std=c90 turns off certain features of GCC that are incompatible with ISO C90, such as the "asm" and "typeof" keywords — but not other GNU extensions that don't have a meaning in ISO C90, like omitting the middle term of a "?:": expression 👀. On the other hand, when a GNU dialect of a standard is specified, all features supported by the compiler are enabled, even when those features change the meaning of the base standard. As a result, some strict-conforming programs may be rejected 🚫. The particular standard is used by -Wpedantic to identify which features are GNU extensions given that version of the standard. For example -std=gnu90 -Wpedantic warns about C++ style // comments, while -std=gnu99 -Wpedantic does not 🧐.

A value for this option must be provided; possible values are:

• c90
• c89
• iso9899:1990
  • Support all ISO C90 programs (certain GNU extensions that conflict with ISO C90 are disabled). Same as -ansi for C code.
• iso9899:199409
  • ISO C90 as modified in amendment 1.
• c99
• c9x
• iso9899:1999
• iso9899:199x
  • ISO C99. This standard is substantially completely supported, modulo bugs and floating-point issues (mainly but not entirely relating to optional C99 features from Annexes F and G) 😭. See <https://gcc.gnu.org/c99status.html> for more information. The names c9x and iso9899:199x are deprecated.
• c11
• c1x
• iso9899:2011
  • ISO C11, the 2011 revision of the ISO C standard. This standard is substantially completely supported, modulo bugs, floating-point issues (mainly but not entirely relating to optional C11 features from Annexes F and G) and the optional Annexes K (Bounds-checking interfaces) and L (Analyzability). The name c1x is deprecated.
• c17
• c18
• iso9899:2017
• iso9899:2018
  • ISO C17, the 2017 revision of the ISO C standard (published in 2018). This standard is same as C11 except for corrections of defects (all of which are also applied with -std=c11) and a new value of "__STDC_VERSION__", and so is supported to the same extent as C11.
• c23
• c2x
• iso9899:2024
  • ISO C23, the 2023 revision of the ISO C standard (published in 2024). The name c2x is deprecated.
• c2y
  • The next version of the ISO C standard, still under development. The support for this version is experimental and incomplete 🧪.
• gnu90
• gnu89
  • GNU dialect of ISO C90 (including some C99 features).
• gnu99
• gnu9x
  • GNU dialect of ISO C99. The name gnu9x is deprecated.
• gnu11
• gnu1x
  • GNU dialect of ISO C11. The name gnu1x is deprecated.
• gnu17
• gnu18
  • GNU dialect of ISO C17.
• gnu23
• gnu2x
  • GNU dialect of ISO C23. This is the default for C code. The name gnu2x is deprecated.
• gnu2y
  • The next version of the ISO C standard, still under development, plus GNU extensions. The support for this version is experimental and incomplete. The name gnu2x is deprecated.
• c++98
• c++03
  • The 1998 ISO C++ standard plus the 2003 technical corrigendum and some additional defect reports. Same as -ansi for C++ code.
• gnu++98
• gnu++03
  • GNU dialect of -std=c++98.
• c++11
• c++0x
  • The 2011 ISO C++ standard plus amendments. The name c++0x is deprecated.
• gnu++11
• gnu++0x
  • GNU dialect of -std=c++11. The name gnu++0x is deprecated.
• c++14
• c++1y
  • The 2014 ISO C++ standard plus amendments. The name c++1y is deprecated.
• gnu++14
• gnu++1y
  • GNU dialect of -std=c++14. The name gnu++1y is deprecated.
• c++17
• c++1z
  • The 2017 ISO C++ standard plus amendments. The name c++1z is deprecated.
• gnu++17
• gnu++1z
  • GNU dialect of -std=c++17. This is the default for C++ code. The name gnu++1z is deprecated.
• c++20
• c++2a
  • The 2020 ISO C++ standard plus amendments. Support is experimental, and could change in incompatible ways in future releases 😬. The name c++2a is deprecated.
• gnu++20
• gnu++2a
  • GNU dialect of -std=c++20. Support is experimental, and could change in incompatible ways in future releases. The name gnu++2a is deprecated.
• c++23
• c++2b
  • The 2023 ISO C++ standard plus amendments (published in 2024). Support is experimental, and could change in incompatible ways in future releases. The name c++2b is deprecated.
• gnu++23
• gnu++2b
  • GNU dialect of -std=c++23. Support is experimental, and could change in incompatible ways in future releases. The name gnu++2b is deprecated.
• c++2c
• c++26
  • The next revision of the ISO C++ standard, planned for 2026. Support is highly experimental, and will almost certainly change in incompatible ways in future releases 💀.
• gnu++2c
• gnu++26
  • GNU dialect of -std=c++2c. Support is highly experimental, and will almost certainly change in incompatible ways in future releases 💀.

-aux-info filename

Outputs to the given filename prototyped declarations for all functions declared and/or defined in a translation unit, including those in header files 📝. Heads up bestie: this option is silently ignored in any language other than C 🙈.

...and bestie, that dump file is chatty — it doesn't just hand you declarations, it tells you where they came from. In comments, the file logs the origin of each declaration (source file and line), whether that declaration was implicit, prototyped or unprototyped (I, N for new or O for old, respectively, sitting in the first character after the line number and the colon 🕵️), and whether it came from a declaration or a definition (C or F, respectively, in the following character). And for function definitions? You also get a K&R-style list of arguments followed by their declarations, tucked inside comments, right after the declaration. Iconic little paper trail, honestly. 💅

• -fno-asm — stop recognizing asm, inline and typeof as keywords, so your code is free to use those words as identifiers. The keywords __asm__, __inline__ and __typeof__ still work if you need them. In C, -ansi implies -fno-asm.

Okay but in C++, inline is a standard keyword and this switch does not touch it. What you probably want there is -fno-gnu-keywords, which disables typeof but leaves asm and inline alone. In C99 mode (-std=c99 or -std=gnu99), this switch only affects the asm and typeof keywords, because inline is a standard keyword in ISO C99. And in C23 mode (-std=c23 or -std=gnu23), it only affects the asm keyword, since typeof is a standard keyword in ISO C23. The keyword list really said "evolve" and GCC evolved with it. 🦋

• -fno-builtin
• -fno-builtin-function — don't recognize built-in functions that do not begin with __builtin_ as prefix.

So here's the tea ☕: GCC normally generates special code to handle certain built-in functions more efficiently. Calls to alloca may become single instructions that adjust the stack directly, and calls to memcpy may become inline copy loops. The resulting code is often both smaller and faster — slay — but since the function calls no longer appear as such, you cannot set a breakpoint on those calls, nor can you change the behavior of the functions by linking with a different library. 🙈

Also, when a function is recognized as a built-in function, GCC may use information about that function to warn about problems with calls to that function, or to generate more efficient code, even if the resulting code still contains calls to that function. For example, warnings are given with -Wformat for bad calls to printf when printf is built in, and strlen is known not to modify global memory.

With the -fno-builtin-function option, only the built-in function function is disabled. But listen — function must not begin with __builtin_. If a function is named that is not built-in in this version of GCC, this option is ignored. And there is no corresponding -fbuiltin-function option; if you wish to enable built-in functions selectively when using -fno-builtin or -ffreestanding, you may define macros such as:

#define abs(n)          __builtin_abs ((n))
#define strcpy(d, s)    __builtin_strcpy ((d), (s))

• -fcond-mismatch — allow conditional expressions with mismatched types in the second and third arguments. The value of such an expression is void. Heads up: this option is not supported for C++. C++ said "not my genre." 💀
• -ffreestanding — assert that compilation targets a freestanding environment. This implies -fno-builtin. A freestanding environment is one in which the standard library may not exist, and program startup may not necessarily be at main. The most obvious example is an OS kernel 👑. This is equivalent to -fno-hosted.
• -fgimple — enable parsing of function definitions marked with __GIMPLE. This is an experimental feature that allows unit testing of GIMPLE passes. Experimental, bestie — you've been warned. 🧪
• -fgnu-tm — when the option -fgnu-tm is specified, the compiler generates code for the Linux variant of Intel's current Transactional Memory ABI specification document (Revision 1.1, May 6 2009). This is an experimental feature whose interface may change in future versions of GCC, as the official specification changes. Please note that not all architectures are supported for this feature.

For more information on GCC's support for transactional memory,

Note that the transactional memory feature is not supported with non-call exceptions (-fnon-call-exceptions). That pairing? Not compatible. Periodt. 🚫

• -fgnu89-inline — the option -fgnu89-inline tells GCC to use the traditional GNU semantics for inline functions when in C99 mode.

Using this option is roughly equivalent to adding the gnu_inline function attribute to all inline functions.

The option -fno-gnu89-inline explicitly tells GCC to use the C99 semantics for inline when in C99 or gnu99 mode (i.e., it specifies the default behavior). This option is not supported in -std=c90 or -std=gnu90 mode.

The preprocessor macros __GNUC_GNU_INLINE__ and __GNUC_STDC_INLINE__ may be used to check which semantics are in effect for inline functions. Vibe check your inline semantics, queen. ✨

• -fhosted — assert that compilation targets a hosted environment. This implies -fbuiltin. A hosted environment is one in which the entire standard library is available, and in which main has a return type of int. Examples are nearly everything except a kernel. This is equivalent to -fno-freestanding. Basically the default main character energy. 🌈
• -flax-vector-conversions — allow implicit conversions between vectors with differing numbers of elements and/or incompatible element types. > This option should not be used for new code. Genuinely, don't. It's giving legacy chaos. 😭
• -fms-extensions — accept some non-standard constructs used in Microsoft header files.

In C++ code, this allows member names in structures to be similar to previous types declarations.

typedef int UOW;
struct ABC {
  UOW UOW;
};

Some cases of unnamed fields in structures and unions are only accepted with this option.

Note that this option is off for all targets except for x86 targets using ms-abi.

• -fpermitted-flt-eval-methods=style — ISO/IEC TS 18661-3 defines new permissible values for FLT_EVAL_METHOD that indicate that operations and constants with a semantic type that is an interchange or extended format should be evaluated to the precision and range of that type. These new values are a superset of those permitted under C99/C11, which does not specify the meaning of other positive values of FLT_EVAL_METHOD. As such, code conforming to C11 may not have been written expecting the possibility of the new values.

-fpermitted-flt-eval-methods specifies whether the compiler should allow only the values of FLT_EVAL_METHOD specified in C99/C11, or the extended set of values specified in ISO/IEC TS 18661-3.

style is either c11 or ts-18661-3 as appropriate.

The default when in a standards compliant mode (-std=c11 or similar) is -fpermitted-flt-eval-methods=c11. The default when in a GNU dialect (-std=gnu11 or similar) is -fpermitted-flt-eval-methods=ts-18661-3.

---

The -fdeps-* options are used to extract structured dependency information for a source. This involves determining what resources provided by other source files will be required to compile the source as well as what resources are provided by the source. This information can be used to add required dependencies between compilation rules of dependent sources based on their contents rather than requiring such information be reflected within the build tools as well. Build systems could never. 🎀

• -fdeps-file=file — where to write structured dependency information.
• -fdeps-format=format — the format to use for structured dependency information. p1689r5 is the only supported format right now. Note that when this argument is specified, the output of -MF is stripped of some information (namely C++ modules) so that it does not use extended makefile syntax not understood by most tools.
• -fdeps-target=file — analogous to -MT but for structured dependency information. This indicates the target which will ultimately need any required resources and provide any resources extracted from the source that may be required by other sources.
• -fplan9-extensions — accept some non-standard constructs used in Plan 9 code.

This enables -fms-extensions, permits passing pointers to structures with anonymous fields to functions that expect pointers to elements of the type of the field, and permits referring to anonymous fields declared using a typedef. This is only supported for C, not C++. C++ once again sitting this one out. 🫶

• -fsigned-bitfields
• -funsigned-bitfields
• -fno-signed-bitfields
• -fno-unsigned-bitfields — these options control whether a bit-field is signed or unsigned, when the declaration does not use either signed or unsigned. By default, such a bit-field is signed, because this is consistent: the basic integer types such as int are signed types. Consistency is the vibe. 💅
• -fsigned-char — let the type char be signed, like signed char.

Note that this is equivalent to -fno-unsigned-char, which is the negative form of -funsigned-char. Likewise, the option -fno-signed-char is equivalent to -funsigned-char. Flip-flop city, bestie. 🔄

• -funsigned-char — let the type char be unsigned, like unsigned char.

Each kind of machine has a default for what char should be. It is either like unsigned char by default or like signed char by default.

Ideally, a portable program should always use signed char or unsigned char when it depends on the signedness of an object. But many programs have been written to use plain char and expect it to be signed, or expect it to be unsigned, depending on the machines they were written for. This option, and its inverse, let you make such a program work with the opposite default.

The type char is always a distinct type from each of signed char or unsigned char, even though its behavior is always just like one of those two. Distinct type, same behavior — it's giving main character with a secret twin. 👯

• -fstrict-flex-arrays (C and C++ only)
• -fstrict-flex-arrays=level (C and C++ only) — control when to treat the trailing array of a structure as a flexible array member for the purpose of accessing the elements of such an array. The value of level controls the level of strictness.

-fstrict-flex-arrays is equivalent to -fstrict-flex-arrays=3, which is the strictest; a trailing array is treated as a flexible array member only when it is declared as a flexible array member per C99 standard onwards.

The negative form -fno-strict-flex-arrays is equivalent to -fstrict-flex-arrays=0, which is the least strict. In this case all trailing arrays of structures are treated as flexible array members. From zero to three, the strictness spectrum is fully unlocked. 🌈

okay so there are two more sneak levels hanging out between 0 and 3, and they're there to support older code that's out here using the GCC zero-length array extension ([0]) or the one-element-array-as-flexible-array-member bit ([1]). 🫶

• when level is 1, the trailing array gets treated as a flexible array member when it's declared as [], [0], or [1] 🎀
• when level is 2, the trailing array gets treated as a flexible array member when it's declared as [], or [0] ✨

and like, you can control this behavior for one specific trailing array field of a structure by using the variable attribute strict_flex_array attribute. so targeted. so precise. 💅

ngl, the -fstrict_flex_arrays option interacts with the -Wstrict-flex-arrays option. they're besties. they talk.

---

-fsso-struct=endianness 🌈

Sets the default scalar storage order of structures and unions to whatever endianness you name. The accepted values are big-endian, little-endian and native — and native means the native endianness of the target, which is also the default.

🚨 not supported for C++ bestie, just so you know.

Warning: the -fsso-struct switch makes GCC generate code that is not binary compatible with code generated without it if the endianness you specified is not the native endianness of the target. like, it's giving chaos if you mix. 💀

---

✨ Options Controlling C++ Dialect ✨

This lil section describes the command-line options that are only meaningful for C++ programs. but plot twist — you can also use most of the GNU compiler options no matter what language your program is in. 🤝 for example, you might compile a file firstClass.C like this:

g++ -g -fstrict-enums -O -c firstClass.C

In this example, only -fstrict-enums is an option meant only for C++ programs; you can use the other options with any language supported by GCC. slay.

Some options for compiling C programs, such as -std, are also relevant for C++ programs. iykyk.

Here is a list of options that are only for compiling C++ programs: 👑

• -fabi-version=n — Use version n of the C++ ABI. The default is version 0. 🦋
  • Version 0 refers to the version conforming most closely to the C++ ABI specification. Therefore, the ABI obtained using version 0 will change in different versions of G++ as ABI bugs are fixed. (so it's the vibes-based one, basically)
  • Version 1 is the version of the C++ ABI that first appeared in G++ 3.2.
  • Version 2 is the version of the C++ ABI that first appeared in G++ 3.4, and was the default through G++ 4.9.
  • Version 3 corrects an error in mangling a constant address as a template argument.
  • Version 4, which first appeared in G++ 4.5, implements a standard mangling for vector types.
  • Version 5, which first appeared in G++ 4.6, corrects the mangling of attribute const/volatile on function pointer types, decltype of a plain decl, and use of a function parameter in the declaration of another parameter.
  • Version 6, which first appeared in G++ 4.7, corrects the promotion behavior of C++11 scoped enums and the mangling of template argument packs, const/static_cast, prefix ++ and --, and a class scope function used as a template argument.
  • Version 7, which first appeared in G++ 4.8, that treats nullptr_t as a builtin type and corrects the mangling of lambdas in default argument scope.
  • Version 8, which first appeared in G++ 4.9, corrects the substitution behavior of function types with function-cv-qualifiers.
  • Version 9, which first appeared in G++ 5.2, corrects the alignment of "nullptr_t".
  • Version 10, which first appeared in G++ 6.1, adds mangling of attributes that affect type identity, such as ia32 calling convention attributes (e.g. stdcall).
  • Version 11, which first appeared in G++ 7, corrects the mangling of sizeof... expressions and operator names. For multiple entities with the same name within a function, that are declared in different scopes, the mangling now changes starting with the twelfth occurrence. It also implies -fnew-inheriting-ctors.
  • Version 12, which first appeared in G++ 8, corrects the calling conventions for empty classes on the x86_64 target and for classes with only deleted copy/move constructors. It accidentally changes the calling convention for classes with a deleted copy constructor and a trivial move constructor. (oopsie 🙈)
  • Version 13, which first appeared in G++ 8.2, fixes the accidental change in version 12.
  • Version 14, which first appeared in G++ 10, corrects the mangling of the nullptr expression.
  • Version 15, which first appeared in G++ 10.3, corrects G++ 10 ABI tag regression.
  • Version 16, which first appeared in G++ 11, changes the mangling of "__alignof__" to be distinct from that of "alignof", and dependent operator names.
  • Version 17, which first appeared in G++ 12, fixes layout of classes that inherit from aggregate classes with default member initializers in C++14 and up.
  • Version 18, which first appeared in G++ 13, fixes manglings of lambdas that have additional context.
  • Version 19, which first appeared in G++ 14, fixes manglings of structured bindings to include ABI tags.
  • Version 20, which first appeared in G++ 15, fixes manglings of lambdas in static data member initializers.

See also -Wabi. 👀

• -fabi-compat-version=n — On targets that support strong aliases, G++ works around mangling changes by creating an alias with the correct mangled name when defining a symbol with an incorrect mangled name. This switch specifies which ABI version to use for the alias. 💅

With -fabi-version=0 (the default), this defaults to 13 (GCC 8.2 compatibility). If another ABI version is explicitly selected, this defaults to 0. For compatibility with GCC versions 3.2 through 4.9, use -fabi-compat-version=2.

If this option is not provided but -Wabi=n is, that version is used for compatibility aliases. If this option is provided along with -Wabi (without the version), the version from this option is used for the warning.

• -fno-access-control — Turn off all access checking. This switch is mainly useful for working around bugs in the access control code. like, it's the "let me in" of compiler flags. 🔓
• -faligned-new — Enable support for C++17 "new" of types that require more alignment than "void* ::operator new(std::size_t)" provides. A numeric argument such as "-faligned-new=32" can be used to specify how much alignment (in bytes) is provided by that function, but few users will need to override the default of alignof(std::max_align_t). 🌸

This flag is enabled by default for -std=c++17.

• -fno-assume-sane-operators-new — The C++ standard allows replacing the global "new", "new[]", "delete" and "delete[]" operators, though a lot of C++ programs don't replace them and just use the implementation provided version. Furthermore, the C++ standard allows omitting those calls if they are made from new or delete expressions (and by extension the same is assumed if "__builtin_operator_new" or "__builtin_operator_delete" functions are used). This option allows control over some optimizations around calls to those operators. 📦

With "-fassume-sane-operators-new-delete" option GCC may assume that calls to the replaceable global operators from new or delete expressions or from "__builtin_operator_new" or "__builtin_operator_delete" calls don't read or modify any global variables or variables whose address could escape to the operators (global state; except for "errno" for the "new" and "new[]" operators). This allows most optimizations across those calls and is something that the implementation provided operators satisfy unless "malloc" implementation details are observable in the code or unless "malloc" hooks are used, but might not be satisfied if a program replaces those operators. This behavior is enabled by default. 🔥

With "-fno-assume-sane-operators-new-delete" option GCC must assume all these calls (whether from new or delete expressions or called directly) may read and write global state unless proven otherwise (e.g. when GCC compiles their implementation). Use this option if those operators are or may be replaced and code needs to expect such behavior.

• -fchar8_t, -fno-char8_t — Enable support for "char8_t" as adopted for C++20. This includes the addition of a new "char8_t" fundamental type, changes to the types of UTF-8 string and character literals, new signatures for user-defined literals, associated standard library updates, and new "__cpp_char8_t" and "__cpp_lib_char8_t" feature test macros. 🧚

This option enables functions to be overloaded for ordinary and UTF-8 strings:

int f(const char *);    // #1
int f(const char8_t *); // #2
int v1 = f("text");     // Calls #1
int v2 = f(u8"text");   // Calls #2

and introduces new signatures for user-defined literals:

int operator""_udl1(char8_t);
int v3 = u8'x'_udl1;
int operator""_udl2(const char8_t*, std::size_t);
int v4 = u8"text"_udl2;
template<typename T, T...> int operator""_udl3();
int v5 = u8"text"_udl3;

The change to the types of UTF-8 string and character literals introduces incompatibilities with ISO C++11 and later standards. For example, the following code is well-formed under ISO C++11, but is ill-formed when -fchar8_t is specified. 😭

const char *cp = u8"xx";// error: invalid conversion from
                        //        `const char8_t*' to `const char*'
int f(const char*);
auto v = f(u8"xx");     // error: invalid conversion from
                        //        `const char8_t*' to `const char*'
std::string s{u8"xx"};  // error: no matching function for call to
                        //        `std::basic_string<char>::basic_string()'
using namespace std::literals;
s = u8"xx"s;            // error: conversion from
                        //        `basic_string<char8_t>' to non-scalar
                        //        type `basic_string<char>' requested

• -fcheck-new — okay so this one makes the compiler double-check that the pointer operator new hands back is actually not null before it goes and pokes at the storage it just allocated. 💅 ngl you almost never need this, because the C++ standard says operator new only returns 0 if it's declared throw(), and in that specific case the compiler checks the return value anyway, even if you don't pass this flag. In literally every other case, when operator new has a non-empty exception specification, running out of memory gets signalled by throwing std::bad_alloc. See also new (nothrow). ✨
• -fconcepts — turns on support for the C++ Concepts feature, aka constraining your template arguments so they stop accepting absolutely any type that wanders in 🎀. With -std=c++20 and above, Concepts are just part of the language standard, so -fconcepts defaults to on.

Some constructs that the earlier C++ Extensions for Concepts Technical Specification, ISO 19217 (2015), allowed but that didn't make it into the standard, could additionally be enabled with -fconcepts-ts. That option was deprecated in GCC 14 and removed in GCC 15, so bestie, the expectation is that you convert your code over to C++20 concepts. No cap. 🦋

• *-fconstexpr-depth=n* — sets the max nested evaluation depth for C++11 constexpr functions to n. You need a limit like this so endless recursion during constant expression evaluation actually gets caught instead of vibing forever 💀. The minimum the standard requires is 512.
• -fconstexpr-cache-depth=n — sets the maximum level of nested evaluation depth for C++11 constexpr functions that gets *cached to *n. It's a heuristic, so it's trading compilation speed (when the cache saves you from repeating calculations) against memory consumption (when the cache blows up huge from super recursive evaluations). Default is 8. Almost nobody is gonna want to touch this, but if your code does heavy constexpr math you might wanna experiment and find the value that slays for you. 🧚
• -fconstexpr-fp-except — Annex F of the C standard says IEC559 floating point exceptions hit at compile time should not stop compilation. C++ compilers have historically ignored that guidance and instead treated floating point division by zero as non-constant, even though it has a well defined value 🙈. This flag tells the compiler to let Annex F win over the other rules that say a given operation is undefined.

constexpr float inf = 1./0.; // OK with -fconstexpr-fp-except

• *-fconstexpr-loop-limit=n* — sets the maximum number of iterations for a loop inside a C++14 constexpr function to n. Again, the limit exists so infinite loops during constant expression evaluation get detected 🔥. Default is 262144 (1<<18).
• -fconstexpr-ops-limit=n — sets the maximum number of operations during a **single constexpr evaluation. Here's the thing: even when one loop's iteration count is capped by the limit above, if you've got several nested loops and each of them has a whole lot of iterations that are still under that limit, or if inside some loop — or even totally outside a loop — way too many expressions need evaluating, the constexpr evaluation can end up taking forever 😭. Default is 33554432 (1<<25).
• -fcontracts — enables experimental support for the C++ Contracts feature, the one that got briefly added to and then yanked out of the C++20 working paper (N4820). The implementation also bundles in the proposed enhancements from papers P1290, P1332, and P1429 🫶. This functionality is meant mostly for people who wanna experiment with refining the feature and getting it into shape for some future C++ standard.

When a checked contract gets violated, the violation handler is called. You can replace that handler by defining:

void
handle_contract_violation (const std::experimental::contract_violation&);

There are different sets of extra flags that can be used together to say which contracts get checked and how — one set for N4820 contracts, one for P1332 contracts, one for P1429 contracts — and these sets cannot be used together, pick one bestie 😤.

• -fcontract-mode=[on|off] — controls whether any contracts have any semantics at all. Defaults to on.
• -fcontract-assumption-mode=[on|off] — [N4820] controls whether contracts with level axiom should have the assume semantic. Defaults to on.
• -fcontract-build-level=[off|default|audit] — [N4820] specifies which level of contracts to generate checks for. Defaults to default.
• -fcontract-continuation-mode=[on|off] — [N4820] controls whether the program is allowed to keep executing after a contract violation. In other words: do checked contracts have the maybe semantic described below rather than the never semantic. Defaults to off.
• *-fcontract-role=<name>:<default>,<audit>,<axiom>* — [P1332] specifies the concrete semantics for each contract level of a particular contract role.
• *-fcontract-semantic=[default|audit|axiom]:<semantic>* — [P1429] specifies the concrete semantic for a particular contract level.
• -fcontract-strict-declarations=[on|off] — controls whether to reject adding contracts to a function after its first declaration. Defaults to off.

The possible concrete semantics you can specify with -fcontract-role or -fcontract-semantic are:

• "ignore" — this contract has no effect.
• "assume" — this contract is treated like C++23 [[assume]].
• "check_never_continue", "never", "abort" — this contract is checked. If it fails, the violation handler is called. If the handler returns, std::terminate is called.
• "check_maybe_continue", "maybe" — this contract is checked. If it fails, the violation handler is called. If the handler returns, execution continues normally.

• -fcoroutines — enables support for the C++ coroutines extension (experimental) ✨.
• -fdiagnostics-all-candidates — lets the C++ front end note all candidates when overload resolution fails, including when a deleted function is selected. So much nicer for actually figuring out what went wrong 🌸.
• -fno-elide-constructors — the C++ standard lets an implementation skip creating a temporary that's only used to initialize another object of the same type. Passing this option disables that optimization and forces G++ to call the copy constructor in all cases 💅. It also makes G++ call trivial member functions that would otherwise be expanded inline.

In C++17 the compiler is required to omit these temporaries, but this option still affects trivial member functions.

• -fno-enforce-eh-specs — don't generate code that checks for violation of exception specifications at run time. Heads up bestie, this option violates the C++ standard, but it may be useful for shrinking code size in production builds, kinda like defining NDEBUG. It does not give your user code permission to throw exceptions that violate the exception specifications — the compiler still optimizes based on those specs, so throwing an unexpected exception results in undefined behavior at run time. 💀
• -fextern-tls-init
• -fno-extern-tls-init — the C++11 and OpenMP standards allow thread_local and threadprivate variables to have dynamic (runtime) initialization. To support that, any use of such a variable goes through a wrapper function that does whatever initialization is needed. When the use and the definition of the variable live in the same translation unit, that overhead can be optimized away — but when the use is in a different translation unit there's significant overhead even if the variable never actually needs dynamic initialization 😭. If you the programmer can be sure that no use of the variable in a non-defining TU needs to trigger dynamic initialization (either because the variable is statically initialized, or because a use of the variable in the defining TU will run before any uses in another TU), you can dodge that overhead with -fno-extern-tls-init.

On targets that support symbol aliases, the default is -fextern-tls-init. On targets that don't support symbol aliases, the default is -fno-extern-tls-init.

• -ffold-simple-inlines
• -fno-fold-simple-inlines — permits the C++ frontend to fold calls to std::move, std::forward, std::addressof and std::as_const. Unlike inlining, this means no debug information gets generated for those calls. Since these functions are rarely interesting to debug, this flag is enabled by default unless -fno-inline is active. 🦋
• -fno-gnu-keywords — don't recognize typeof as a keyword, so your code can use that word as an identifier. You can use the keyword __typeof__ instead. This option is implied by the strict ISO C++ dialects: -ansi, -std=c++98, -std=c++11, etc. 🎀
• -fno-immediate-escalation — don't enable immediate function escalation, the thing where certain functions can get promoted to consteval, as specified in P2564R3. For example:

consteval int id(int i) { return i; }

constexpr int f(auto t)
{
  return t + id(t); // id causes f<int> to be promoted to consteval
}

void g(int i)
{
  f (3);
}

this compiles in C++20: f is an immediate-escalating function (because of the auto it's a function template, and it's declared constexpr), and id(t) is an immediate-escalating expression, so f gets promoted to consteval. As a result, the call to id(t) is in an immediate context, so it doesn't have to produce a constant — that's the mechanism that allows consteval function composition, iconic honestly 👑. But with -fno-immediate-escalation, f is not promoted to consteval, and since the call to the consteval function id(t) isn't a constant expression, the compiler rejects the code.

This option is turned on by default; it's only effective in C++20 mode or later.

• -fimplicit-constexpr — makes inline functions implicitly constexpr, if they satisfy the requirements for a constexpr function. You can use this in C++14 mode or later. It can result in initialization switching from dynamic to static, plus other optimizations ✨.
• -fno-implicit-templates — bestie, listen: this one says never emit code for non-inline templates that got instantiated implicitly (aka just because somebody used them 👀). Only code for explicit instantiations. And because of that, if you turn this on you must carefully structure your code to include every explicit instantiation it needs, otherwise you get undefined symbols at link time. 💀
• -fno-implicit-inline-templates — same vibe, but inline templates: don't emit code for implicit instantiations of those either. The default handles inlines differently on purpose, so that compiles with and without optimization need the exact same set of explicit instantiations. ✨
• -fno-implement-inlines — to save space, don't emit out-of-line copies of inline functions controlled by #pragma implementation. Heads up though: this causes linker errors if those functions aren't inlined everywhere they're called. 😭
• -fmodules, -fno-modules — turn support for C++20 modules on and off. You usually don't need -fno-modules because that's the default already. And even though modules are a C++20 feature, they are not currently implicitly enabled just by selecting that standard version. 🙈
• -fmodule-header, -fmodule-header=user, -fmodule-header=system — compile a header file to create an importable header unit. 🎀
• -fmodule-implicit-inline — in modular code, member functions defined right inside their class definitions are not implicitly inline. That's different from traditional C++ behavior, and for good reasons, bestie. It can make porting code a little rough though, so this option makes such function definitions implicitly inline anyway. It does generate an ABI incompatibility, so you have to use it everywhere or nowhere — no half measures. (Those kinds of definitions outside of a named module stay implicitly inline regardless. 😌)
• -fno-module-lazy — disable lazy module importing and module mapper creation.
• -fmodule-mapper=[hostname]:port[?ident], -fmodule-mapper=|program[?ident] args..., -fmodule-mapper==socket[?ident], -fmodule-mapper=<>[inout][?ident], -fmodule-mapper=<in>out[?ident], -fmodule-mapper=file[?ident] — this is the oracle 🧚 you query for module name to filename mappings. If you don't specify one, the CXX_MODULE_MAPPER environment variable is used, and if that's unset too, an in-process default is provided.
• -fmodule-only — only emit the Compiled Module Interface, inhibiting any object file. That's it, that's the whole job. 👑
• -fms-extensions — disable Wpedantic warnings about constructs used in MFC, like implicit int and getting a pointer to member function via non-standard syntax.
• -fnew-inheriting-ctors — enables the P0136 adjustment to the semantics of C++11 constructor inheritance. It's part of C++17 but also considered a Defect Report against C++11 and C++14. This flag is on by default unless you specify -fabi-version=10 or lower.
• -fnew-ttp-matching — enables the P0522 resolution to Core issue 150, about template template parameters and default arguments: it lets a template with default template arguments be used as an argument for a template template parameter that has fewer template parameters. This flag is on by default for -std=c++17. 🔥
• -fno-nonansi-builtins — disable built-in declarations of functions that aren't mandated by ANSI/ISO C. That's the crew of ffs, alloca, _exit, index, bzero, conjf, and other related functions.
• -fnothrow-opt — treat a throw() exception specification as if it were a noexcept specification, to reduce or eliminate the text size overhead compared to a function with no exception specification. Fun twist: if the function has local variables of types with non-trivial destructors, the exception specification actually makes the function smaller, because the EH cleanups for those variables can be optimized away. The semantic effect is that an exception thrown out of a function with such an exception specification results in a call to terminate rather than unexpected. 🫶
• -fno-operator-names — don't treat the operator name keywords and, bitand, bitor, compl, not, or and xor as synonyms as keywords.
• -fno-optional-diags — disable diagnostics that the standard says a compiler doesn't need to issue. Right now the only such diagnostic G++ issues is the one for a name having multiple meanings within a class. 💅
• -fno-pretty-templates — okay so normally, when an error message refers to a specialization of a function template, the compiler prints the signature of the template followed by the template arguments and any typedefs or typenames in the signature (e.g. void f(T) [with T = int] rather than void f(int)) so it's clear which template is involved. And when an error message refers to a specialization of a class template, the compiler omits any template arguments that match the default template arguments for that template. If either of those behaviors makes the error message harder to understand instead of easier, you can use -fno-pretty-templates to disable them. 🦋
• -frange-for-ext-temps — enables lifetime extension of C++ range based for temporaries. With -std=c++23 and above this is part of the language standard, so the lifetime of the temporaries is extended until the end of the loop by default. This option lets you enable that same behavior in earlier versions of the standard too.
• -fno-rtti — disable generation of information about every class with virtual functions, the info used by C++ run-time type identification features (dynamic_cast and typeid). If you don't use those parts of the language, this flag saves you some space. Note that exception handling uses the same information, but G++ generates it as needed. The dynamic_cast operator can still be used for casts that don't require run-time type information, i.e. casts to void * or to unambiguous base classes.

Careful, bestie: mixing code compiled with -frtti and code compiled with -fno-rtti may not work. For example, programs may fail to link if a class compiled with -fno-rtti is used as a base for a class compiled with -frtti. 💀

• -fsized-deallocation — enable the built-in global declarations

void operator delete (void *, std::size_t) noexcept;
void operator delete[] (void *, std::size_t) noexcept;

as introduced in C++14. This is useful for user-defined replacement deallocation functions that, for example, use the size of the object to make deallocation faster. It's enabled by default under -std=c++14 and above. And the flag -Wsized-deallocation warns about places that might want to add a definition. 🌸

• -fstrict-enums — let the compiler optimize using the assumption that a value of enumerated type can only be one of the values of the enumeration (as defined in the C++ standard; basically, a value that can be represented in the minimum number of bits needed to represent all the enumerators). That assumption may not be valid if the program uses a cast to convert an arbitrary integer value to the enumerated type. Also good to know: this option has no effect for an enumeration type with a fixed underlying type. ✨
• -fstrong-eval-order, -fstrong-eval-order=kind — evaluate member access, array subscripting, and shift expressions in left-to-right order, and evaluate assignment in right-to-left order, as adopted for C++17. -fstrong-eval-order is equivalent to -fstrong-eval-order=all, and it's enabled by default with -std=c++17 or later.

-fstrong-eval-order=some enables just the ordering of member access and shift expressions, and is the default for C++ dialects prior to C++17.

-fstrong-eval-order=none is equivalent to -fno-strong-eval-order.

• -ftemplate-backtrace-limit=n — set the maximum number of template instantiation notes for a single warning or error to n. The default value is 10. 🔢
• -ftemplate-depth=n — set the maximum instantiation depth for template classes to n. A limit on the template instantiation depth is needed to detect endless recursions during template class instantiation. ANSI/ISO C++ conforming programs must not rely on a maximum depth greater than 17 (changed to 1024 in C++11). The default value is 900, since the compiler can run out of stack space before hitting 1024 in some situations. 😭
• -fno-threadsafe-statics — don't emit the extra code that uses the routines specified in the C++ ABI for thread-safe initialization of local statics. You can use this option to reduce code size slightly in code that doesn't need to be thread-safe. 🧚
• -fuse-cxa-atexit — register destructors for objects with static storage duration with the __cxa_atexit function rather than the atexit function. This option is required for fully standards-compliant handling of static destructors, but it only works if your C library supports __cxa_atexit. 🫶
• -fno-use-cxa-get-exception-ptr — don't use the __cxa_get_exception_ptr runtime routine. This causes std::uncaught_exception to be incorrect, but it's necessary if the runtime routine isn't available.
• -fvisibility-inlines-hidden — this switch declares that the user does not attempt to compare pointers to inline functions or methods where the addresses of the two functions are taken in different shared objects. 👑

The effect of this is that GCC may, effectively, mark inline methods with __attribute__ ((visibility ("hidden"))) so that they don't appear in the export table of a DSO and don't require a PLT indirection when used within the DSO. Enabling this option can have a dramatic effect on load and link times of a DSO, because it massively reduces the size of the dynamic export table when the library makes heavy use of templates. 🔥

The behavior of this switch is not quite the same as marking the methods as hidden directly, because it does not affect static variables local to the function or cause the compiler to deduce that the function is defined in only one shared object.

You may mark a method as having a visibility explicitly to negate the effect of the switch for that method. For example, if you do want to compare pointers to a particular inline method, you might mark it as having default visibility. Marking the enclosing class with explicit visibility has no effect. 💅

Explicitly instantiated inline methods are unaffected by this option, as their linkage might otherwise cross a shared library boundary. 🌈

• -fvisibility-ms-compat — bestie this flag is a whole vibe translator 💅 it tries to use visibility settings to make GCC's C++ linkage model compatible with the one Microsoft Visual Studio uses.

The flag makes these changes to GCC's linkage model:

1. It sets the default visibility to "hidden", like -fvisibility=hidden.
2. Types, but not their members, are not hidden by default.
3. The One Definition Rule is relaxed for types without explicit visibility specifications that are defined in more than one shared object: those declarations are permitted if they are permitted when this option is not used.

In new code it is better to use -fvisibility=hidden and export those classes that are intended to be externally visible. Unfortunately it is possible for code to rely, perhaps accidentally, on the Visual Studio behavior. 🙈

Among the consequences of these changes are that static data members of the same type with the same name but defined in different shared objects are different, so changing one does not change the other; and that pointers to function members defined in different shared objects may not compare equal. When this flag is given, it is a violation of the ODR to define types with the same name differently. not me casually breaking the One Definition Rule 😭

• -fno-weak — do NOT use weak symbol support, even if the linker is out here offering it. By default, G++ uses weak symbols if they are available. This option exists only for testing, and should not be used by end-users; it results in inferior code and has no benefits. This option may be removed in a future release of G++. so like… it's giving deprecated-testing-tool energy, pls leave it alone 🫶
• -fext-numeric-literals <span>(C++ and Objective-C++ only)</span> — accept imaginary, fixed-point, or machine-defined literal number suffixes as GNU extensions. When this option is turned off these suffixes are treated as C++11 user-defined literal numeric suffixes. This is on by default for all pre-C++11 dialects and all GNU dialects: -std=c++98, -std=gnu++98, -std=gnu++11, -std=gnu++14. This option is off by default for ISO C++11 onwards (-std=c++11, ...).
• -nostdinc++ — do not search for header files in the standard directories specific to C++, but do still search the other standard directories. (This option is used when building the C++ library.) library-builders only besties ✨
• -flang-info-include-translate
• -flang-info-include-translate-not
• -flang-info-include-translate=header — inform of include translation events. 💬 The first will note accepted include translations, the second will note declined include translations. The header form will inform of include translations relating to that specific header. If header is of the form "user" or "<system>" it will be resolved to a specific user or system header using the include path.
• -flang-info-module-cmi
• -flang-info-module-cmi=module — inform of Compiled Module Interface pathnames. 🗂️ The first will note all read CMI pathnames. The module form will not reading a specific module's CMI. module may be a named module or a header-unit (the latter indicated by either being a pathname containing directory separators or enclosed in "<>" or "").
• -stdlib=libstdc++,libc++ — when G++ is configured to support this option, it allows specification of alternate C++ runtime libraries. two lovely options on the menu 🌈 libstdc++ (the default, native C++ runtime for G++) and libc++ which is the C++ runtime installed on some operating systems (e.g. Darwin versions from Darwin11 onwards). The option switches G++ to use the headers from the specified library and to emit "-lstdc++" or "-lc++" respectively, when a C++ runtime is required for linking. 🔗

---

In addition, these warning options have meanings only for C++ programs: 👑

• -Wabi-tag <span>(C++ and Objective-C++ only)</span> — warn when a type with an ABI tag is used in a context that does not have that ABI tag. See C++ Attributes for more information about ABI tags. 🏷️
• -Wcomma-subscript <span>(C++ and Objective-C++ only)</span> — warn about uses of a comma expression within a subscripting expression. This usage was deprecated in C++20 and is going to be removed in C++23. 😱 However, a comma expression wrapped in "( )" is not deprecated. Example:

void f(int *a, int b, int c) {
    a[b,c];     // deprecated in C++20, invalid in C++23
    a[(b,c)];   // OK
}

In C++23 it is valid to have comma separated expressions in a subscript when an overloaded subscript operator is found and supports the right number and types of arguments. G++ will accept the formerly valid syntax for code that is not valid in C++23 but used to be valid but deprecated in C++20 with a pedantic warning that can be disabled with -Wno-comma-subscript.

Enabled by default with -std=c++20 unless -Wno-deprecated, and after -std=c++23 regardless of -Wno-deprecated. Before -std=c++20, enabled with explicit -Wdeprecated.

This warning is upgraded to an error by -pedantic-errors in C++23 mode or later. so serious 💀

• -Wctad-maybe-unsupported <span>(C++ and Objective-C++ only)</span> — warn when performing class template argument deduction (CTAD) on a type with no explicitly written deduction guides. This warning will point out cases where CTAD succeeded only because the compiler synthesized the implicit deduction guides, which might not be what the programmer intended. Certain style guides allow CTAD only on types that specifically "opt-in"; i.e., on types that are designed to support CTAD. This warning can be suppressed with the following pattern:

struct allow_ctad_t; // any name works
template <typename T> struct S {
  S(T) { }
};
// Guide with incomplete parameter type will never be considered.
S(allow_ctad_t) -> S<void>;

• -Wctor-dtor-privacy <span>(C++ and Objective-C++ only)</span> — warn when a class seems unusable because all the constructors or destructors in that class are private, and it has neither friends nor public static member functions. screaming "let me IN" at the compiler 🔓 Also warn if there are no non-private methods, and there's at least one private member function that isn't a constructor or destructor.
• -Wdangling-reference <span>(C++ and Objective-C++ only)</span> — warn when a reference is bound to a temporary whose lifetime has ended. For example:

int n = 1;
const int& r = std::max(n - 1, n + 1); // r is dangling

In the example above, two temporaries are created, one for each argument, and a reference to one of the temporaries is returned. However, both temporaries are destroyed at the end of the full expression, so the reference "r" is dangling. ghost reference fr fr 👻 This warning also detects dangling references in member initializer lists:

const int& f(const int& i) { return i; }
struct S {
  const int &r; // r is dangling
  S() : r(f(10)) { }
};

Member functions are checked as well, but only their object argument:

struct S {
   const S& self () { return *this; }
};
const S& s = S().self(); // s is dangling

Certain functions are safe in this respect, for example "std::use_facet": they take and return a reference, but they don't return one of its arguments, which can fool the warning. Such functions can be excluded from the warning by wrapping them in a "#pragma":

#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wdangling-reference"
const T& foo (const T&) { ... }
#pragma GCC diagnostic pop

The "#pragma" can also surround the class; in that case, the warning will be disabled for all the member functions.

-Wdangling-reference also warns about code like

auto p = std::minmax(1, 2);

where "std::minmax" returns "std::pair<const int&, const int&>", and both references dangle after the end of the full expression that contains the call to "std::minmax".

The warning does not warn for "std::span"-like classes. We consider classes of the form:

template<typename T>
struct Span {
  T* data_;
  std::size len_;
};

as "std::span"-like; that is, the class is a non-union class that has a pointer data member and a trivial destructor.

The warning can be disabled by using the "gnu::no_dangling" attribute.

This warning is enabled by -Wextra. ✨

• -Wdelete-non-virtual-dtor <span>(C++ and Objective-C++ only)</span> — warn when "delete" is used to destroy an instance of a class that has virtual functions and non-virtual destructor. It is unsafe to delete an instance of a derived class through a pointer to a base class if the base class does not have a virtual destructor. this one's genuinely dangerous bestie, listen to it 😭 This warning is enabled by -Wall.
• -Wdeprecated-copy <span>(C++ and Objective-C++ only)</span> — warn that the implicit declaration of a copy constructor or copy assignment operator is deprecated if the class has a user-provided copy constructor or copy assignment operator, in C++11 and up. This warning is enabled by -Wextra. With -Wdeprecated-copy-dtor, also deprecate if the class has a user-provided destructor.
• -Wno-deprecated-enum-enum-conversion <span>(C++ and Objective-C++ only)</span> — disable the warning about the case when the usual arithmetic conversions are applied on operands where one is of enumeration type and the other is of a different enumeration type. This conversion was deprecated in C++20. For example:

enum E1 { e };
enum E2 { f };
int k = f - e;

-Wdeprecated-enum-enum-conversion is enabled by default with -std=c++20. In pre-C++20 dialects, this warning can be enabled by -Wenum-conversion or -Wdeprecated.

• -Wno-deprecated-enum-float-conversion <span>(C++ and Objective-C++ only)</span> — disable the warning about the case when the usual arithmetic conversions are applied on operands where one is of enumeration type and the other is of a floating-point type. This conversion was deprecated in C++20. For example:

enum E1 { e };
enum E2 { f };
bool b = e <= 3.7;

-Wdeprecated-enum-float-conversion is enabled by default with -std=c++20. In pre-C++20 dialects, this warning can be enabled by -Wenum-conversion or -Wdeprecated. 💖

• -Wdeprecated-literal-operator (C++ and Objective-C++ only) — bestie, this one taps you on the shoulder when someone declares a user-defined literal operator with a lil space before the suffix, which is officially deprecated. It's on by default in C++23, or whenever you pass explicit -Wdeprecated. ✨

string operator "" _i18n(const char*, std::size_t); // deprecated
string operator ""_i18n(const char*, std::size_t); // preferred

• -Wdeprecated-variadic-comma-omission (C++ and Objective-C++ only) — this is the "where's the comma bestie 😭" warning. It flags omitting a comma before the varargs ... at the end of a function parameter list, because that is deprecated. Enabled by default in C++26, or with explicit -Wdeprecated.

void f1(int...); // deprecated
void f1(int, ...); // preferred
template <typename ...T>
void f2(T...); // ok
template <typename ...T>
void f3(T......); // deprecated

• -Wno-elaborated-enum-base — for C++11 and above, this warns if an (invalid) additional enum-base shows up inside an elaborated-type-specifier. Translation: if an enum with a given underlying type and no enumerator list gets used in a declaration that isn't just a standalone declaration of the enum. It's enabled by default, and with -pedantic-errors this warning gets upgraded to a full-on error 💀.
• -Wno-init-list-lifetime (C++ and Objective-C++ only) — this one tells the compiler not to warn about uses of std::initializer_list that are probably gonna leave you with dangling pointers. Why does that happen? Because the underlying array for an initializer_list is handled like a totally normal C++ temporary object, so it's so easy to accidentally keep a pointer to that array past the end of its lifetime. For example:
  • if a function returns a temporary initializer_list, or a local initializer_list variable, the array's lifetime ends at the end of the return statement — so the value you returned is holding a dangling pointer.
  • if a new-expression creates an initializer_list, the array only lives until the end of the enclosing full-expression, so the initializer_list chilling in the heap has a dangling pointer.
  • when an initializer_list variable is assigned from a brace-enclosed initializer list, the temporary array made for the right side of the assignment only lives until the end of the full-expression, so by the next statement the initializer_list variable is sitting there with a dangling pointer 😭

// li's initial underlying array lives as long as li
std::initializer_list<int> li = { 1,2,3 };
// assignment changes li to point to a temporary array
li = { 4, 5 };
// now the temporary is gone and li has a dangling pointer
int i = li.begin()[0] // undefined behavior

• when a list constructor stores the begin pointer from the initializer_list argument, that does not extend the lifetime of the array — so if a class variable is constructed from a temporary initializer_list, that pointer is left dangling by the end of the variable declaration statement. 💀

• -Winvalid-constexpr — warns when a function never produces a constant expression. In C++20 and earlier, for every constexpr function and function template, there had to be at least one set of function arguments in at least one instantiation such that an invocation of the function or constructor could be an evaluated subexpression of a core constant expression. C++23 removed this restriction, so it's possible to have a function or function template marked constexpr where no invocation actually satisfies the requirements of a core constant expression.

This warning is enabled as a pedantic warning by default in C++20 and earlier. In C++23, -Winvalid-constexpr can be turned on, and in that case it'll be an ordinary warning. Example time:

void f (int& i);
constexpr void
g (int& i)
{
  // Warns by default in C++20, in C++23 only with -Winvalid-constexpr.
  f(i);
}

• -Winvalid-imported-macros — verifies that all imported macro definitions are valid at the end of compilation. This is not enabled by default, since it needs extra processing to figure out. It can be useful when you're preparing sets of header-units to make sure the macros stay consistent. 🧚
• -Wno-literal-suffix (C++ and Objective-C++ only) — tells it not to warn when a string or character literal is followed by a ud-suffix that does not start with an underscore. As a conforming extension, GCC treats those suffixes as separate preprocessing tokens, all to keep backwards compatibility with code that uses formatting macros from <inttypes.h>. Like:

#define __STDC_FORMAT_MACROS
#include <inttypes.h>
#include <stdio.h>

int main() {
  int64_t i64 = 123;
  printf("My int64: %" PRId64"\n", i64);
}

In this case, PRId64 is treated as a separate preprocessing token.

This option also controls the warnings when a user-defined literal operator is declared with a literal suffix identifier that doesn't begin with an underscore. Heads up: literal suffix identifiers that don't begin with an underscore are reserved for future standardization. These warnings are enabled by default. 💅

• -Wno-narrowing (C++ and Objective-C++ only) — for C++11 and later standards, narrowing conversions get diagnosed by default, as the standard demands. A narrowing conversion from a constant produces an error, and a narrowing conversion from a non-constant produces a warning, but -Wno-narrowing suppresses that diagnostic. Ngl, this does not affect the meaning of well-formed code — narrowing conversions are still considered ill-formed in SFINAE contexts.

With -Wnarrowing in C++98, it warns when a narrowing conversion prohibited by C++11 happens inside { }, like:

int i = { 2.2 }; // error: narrowing from double to int

This flag is included in -Wall and -Wc++11-compat. ✨

• -Wnoexcept (C++ and Objective-C++ only) — warns when a noexcept-expression evaluates to false because of a call to a function that doesn't have a non-throwing exception specification (i.e. throw() or noexcept) but is known by the compiler to never throw an exception.
• -Wnoexcept-type (C++ and Objective-C++ only) — warns if the C++17 feature that makes noexcept part of a function type changes the mangled name of a symbol compared to C++14. Enabled by -Wabi and -Wc++17-compat. Example bestie:

template <class T> void f(T t) { t(); };
void g() noexcept;
void h() { f(g); }

In C++14, f calls f<void(*)()>, but in C++17 it calls f<void(*)()noexcept>. 🦋

• -Wclass-memaccess (C++ and Objective-C++ only) — warns when the destination of a call to a raw memory function like memset or memcpy is an object of class type, and when writing into that object might bypass the class's non-trivial or deleted constructor or copy assignment, violate const-correctness or encapsulation, or corrupt virtual table pointers. Modifying the representation of such objects may violate invariants that member functions of the class are out here maintaining. For instance, the memset call below is undefined because it modifies a non-trivial class object, so it gets diagnosed. The safe way to initialize or clear the storage of objects like these is to use the appropriate constructor or assignment operator, if one's available.

std::string str = "abc";
memset (&str, 0, sizeof str);

The -Wclass-memaccess option is enabled by -Wall. Explicitly casting the pointer to the class object to void *, or to a type that can be safely accessed by the raw memory function, suppresses the warning. 💅

• -Wnon-virtual-dtor (C++ and Objective-C++ only) — warns when a class has virtual functions and an accessible non-virtual destructor, either itself or in an accessible polymorphic base class — in which case it's possible but unsafe to delete an instance of a derived class through a pointer to the class itself or the base class. This warning is automatically enabled if -Weffc++ is specified. The -Wdelete-non-virtual-dtor option (enabled by -Wall) should be preferred, because it warns about the unsafe cases without false positives. 👑
• -Wregister (C++ and Objective-C++ only) — warns on uses of the register storage class specifier, except when it's part of the GNU Explicit Register Variables extension. Using the register keyword as a storage class specifier has been deprecated in C++11 and removed in C++17. Enabled by default with -std=c++17.
• -Wreorder (C++ and Objective-C++ only) — warns when the order of member initializers written in the code doesn't match the order they actually have to be executed in. Like:

struct A {
  int i;
  int j;
  A(): j (0), i (1) { }
};

The compiler rearranges the member initializers for i and j to match the declaration order of the members, and emits a warning to that effect. This warning is enabled by -Wall. 🌸

• -Wno-pessimizing-move (C++ and Objective-C++ only) — this one warns you when a call to std::move is actually preventing copy elision (the audacity 😭). A typical scenario where copy elision can occur is returning in a function with a class return type, when the expression being returned is the name of a non-volatile automatic object, and isn't a function parameter, and has the same type as the function return type.

struct T {
...
};
T fn()
{
  T t;
  ...
  return std::move (t);
}

But in this example, that std::move call prevents copy elision. This warning is enabled by -Wall.

• -Wno-redundant-move (C++ and Objective-C++ only) — this one warns about redundant calls to std::move; meaning, when a move operation would've happened even without the std::move call. That's because the compiler is forced to treat the object as if it were an rvalue in certain situations, like returning a local variable, where copy elision isn't applicable. Consider:

struct T {
...
};
T fn(T t)
{
  ...
  return std::move (t);
}

okay so here the std::move call is literally redundant bestie, it's doing nothing 💀. And because G++ implements Core Issue 1579, here's another little example:

struct T { // convertible to U
...
};
struct U {
...
};
U fn()
{
  T t;
  ...
  return std::move (t);
}

in this one, copy elision isn't even applicable, because the type of the expression being returned and the function's return type are different — but G++ still treats the return value as if it were designated by an rvalue. she's built different fr fr ✨

This warning is enabled by -Wextra.

---

• -Wrange-loop-construct (C++ and Objective-C++ only) — this one calls out a C++ range-based for-loop when it's making an unnecessary copy 🙈. That can happen when the range declaration isn't a reference, but like… probably should be? example time:

struct S { char arr[128]; };
void fn () {
  S arr[5];
  for (const auto x : arr) { ... }
}

It does not warn when the type being copied is a trivially-copyable type whose size is less than 64 bytes.

It also warns when a loop variable in a range-based for-loop is initialized with a value of a different type, which results in a copy. like:

void fn() {
  int arr[10];
  for (const double &x : arr) { ... }
}

In the example above, in every single iteration of the loop a temporary value of type double gets created and destroyed, and then the reference const double & is bound to it. every loop. every time. exhausting honestly 💅

This warning is enabled by -Wall.

• -Wredundant-tags (C++ and Objective-C++ only) — warns about redundant class-key and enum-key in references to class types and enumerated types in contexts where the key can be eliminated without causing an ambiguity. example:

struct foo;
struct foo *p;   // warn that keyword struct can be eliminated

but on the other hand, in this example there's no warning:

struct foo;
void foo ();   // "hides" struct foo
void bar (struct foo&);  // no warning, keyword struct is necessary

see? the keyword was necessary, she's not gonna come for you for that 🫶

• -Wno-subobject-linkage (C++ and Objective-C++ only) — do not warn if a class type has a base or a field whose type uses the anonymous namespace, or depends on a type with no linkage. If a type A depends on a type B with no or internal linkage, defining it in multiple translation units would be an ODR violation because the meaning of B is different in each translation unit 😭. If A only appears in a single translation unit, the best way to silence the warning is to give it internal linkage by putting it in an anonymous namespace as well. The compiler doesn't give this warning for types defined in the main .C file, as those are unlikely to have multiple definitions. -Wsubobject-linkage is enabled by default.
• -Weffc++ (C++ and Objective-C++ only) — warns about violations of the following style guidelines from Scott Meyers' Effective C++ series of books 👑:
  • Define a copy constructor and an assignment operator for classes with dynamically-allocated memory.
  • Prefer initialization to assignment in constructors.
  • Have operator= return a reference to *this.
  • Don't try to return a reference when you must return an object.
  • Distinguish between prefix and postfix forms of increment and decrement operators.
  • Never overload &&, ||, or ,.

This option also enables -Wnon-virtual-dtor, which is also one of the effective C++ recommendations. However, the check is extended to warn about the lack of virtual destructor in accessible non-polymorphic bases classes too.

bestie if you select this option, be aware that the standard library headers do not obey all of these guidelines 🌸 — use grep -v to filter out those warnings. yw 💖

• -Wno-exceptions (C++ and Objective-C++ only) — disables the warning about the case when an exception handler is shadowed by another handler, which can point out a wrong ordering of exception handlers. (ordering matters!! 👀)
• -Wstrict-null-sentinel (C++ and Objective-C++ only) — warns about the use of an uncasted NULL as sentinel. When compiling only with GCC this is a valid sentinel, since NULL is defined to __null. Although it's a null pointer constant rather than a null pointer, it's guaranteed to be of the same size as a pointer. But this use is not portable across different compilers. so like… careful 🫠
• -Wno-non-template-friend (C++ and Objective-C++ only) — disables warnings when non-template friend functions are declared within a template. In very old versions of GCC that predate implementation of the ISO standard, declarations such as friend int foo(int), where the name of the friend is an unqualified-id, could be interpreted as a particular specialization of a template function. The warning exists to diagnose compatibility problems, and it's enabled by default. vintage drama ✨
• -Wold-style-cast (C++ and Objective-C++ only) — warns if an old-style (C-style) cast to a non-void type is used within a C++ program. The new-style casts (dynamic_cast, static_cast, reinterpret_cast, and const_cast) are less vulnerable to unintended effects and much easier to search for. serve the new style, retire the old one 💅
• -Woverloaded-virtual, -Woverloaded-virtual=n (C++ and Objective-C++ only) — warns when a function declaration hides virtual functions from a base class. for example, in:

struct A {
  virtual void f();
};

struct B: public A {
  void f(int); // does not override
};

the A class version of f is hidden in B, and code like:

B* b;
b->f();

fails to compile 💀. In cases where the different signatures are not an accident, the simplest solution is to add a using-declaration to the derived class to un-hide the base function, e.g. add using A::f; to B.

The optional level suffix controls the behavior when all the declarations in the derived class override virtual functions in the base class, even if not all of the base functions are overridden:

struct C {
  virtual void f();
  virtual void f(int);
};

struct D: public C {
  void f(int); // does override
}

This pattern is less likely to be a mistake; if D is only used virtually, the user might have decided that the base class semantics for some of the overloads are fine. At level 1, this case does not warn; at level 2, it does. -Woverloaded-virtual by itself selects level 2. Level 1 is included in -Wall.

• -Wno-pmf-conversions (C++ and Objective-C++ only) — disables the diagnostic for converting a bound pointer to member function to a plain pointer.
• -Wsign-promo (C++ and Objective-C++ only) — warns when overload resolution chooses a promotion from unsigned or enumerated type to a signed type, over a conversion to an unsigned type of the same size. Previous versions of G++ tried to preserve unsignedness, but the standard mandates the current behavior. the standard said no 🚫
• -Wtemplates (C++ and Objective-C++ only) — warns when a primary template declaration is encountered. Some coding rules disallow templates, and this may be used to enforce that rule. The warning is inactive inside a system header file, such as the STL, so one can still use the STL. One may also instantiate or specialize templates.
• -Wmismatched-new-delete (C++ and Objective-C++ only) — warns for mismatches between calls to operator new or operator delete and the corresponding call to the allocation or deallocation function 🔥. This includes invocations of C++ operator delete with pointers returned from either mismatched forms of operator new, or from other functions that allocate objects for which the operator delete isn't a suitable deallocator, as well as calls to other deallocation functions with pointers returned from operator new for which the deallocation function isn't suitable.

For example, the delete expression in the function below is diagnosed because it doesn't match the array form of the new expression the pointer argument was returned from. Similarly, the call to free is also diagnosed:

void f ()
{
  int *a = new int[n];
  delete a;   // warning: mismatch in array forms of expressions

  char *p = new char[n];
  free (p);   // warning: mismatch between new and free
}

The related option -Wmismatched-dealloc diagnoses mismatches involving allocation and deallocation functions other than operator new and operator delete.

-Wmismatched-new-delete is included in -Wall.

• -Wmismatched-tags (C++ and Objective-C++ only) — warns for declarations of structs, classes, and class templates and their specializations with a class-key that does not match either the definition or the first declaration if no definition is provided.

For example, the declaration of struct Object in the argument list of draw triggers the warning. To avoid it, either remove the redundant class-key struct or replace it with class to match its definition.

class Object {
public:
  virtual ~Object () = 0;
};
void draw (struct Object*);

It is not wrong to declare a class with the class-key struct, as the example above shows. The -Wmismatched-tags option is intended to help achieve a consistent style of class declarations. In code that is intended to be portable to Windows-based compilers, the warning helps prevent unresolved references due to the difference in the mangling of symbols declared with different class-keys. The option can be used either on its own or in conjunction with -Wredundant-tags. consistency is the moment bestie 🎀

• -Wmultiple-inheritance (C++ and Objective-C++ only) — bestie, this one yells at you when a class is defined with multiple direct base classes 💅 Some coding rules straight-up ban multiple inheritance, and this is how you enforce that rule. The warning is inactive inside a system header file, like the STL, so you can still use the STL. You can also define classes that indirectly use multiple inheritance, no problem.
• -Wvirtual-inheritance — warns when a class is defined with a virtual direct base class. Same vibe as above: some coding rules are anti-multiple-inheritance and this enforces it. Also inactive inside a system header file (hi STL 🫶), and you may also define classes that indirectly use virtual inheritance.
• -Wno-virtual-move-assign — hush-hush mode for warnings about inheriting from a virtual base with a non-trivial C++11 move assignment operator. Lowkey this is dangerous 😭 because if the virtual base is reachable along more than one path, it gets moved multiple times, which can mean both objects end up in the moved-from state. If the move assignment operator is written to avoid moving from a moved-from object, then go ahead and disable this warning.
• -Wnamespaces — warns when a namespace definition is opened. Coding rules that disallow namespaces? This is how you enforce them. Inactive inside a system header file like the STL (STL stays winning ✨), and using directives and qualified names are still totally fine.
• -Wno-template-body (C++ and Objective-C++ only) — turns off diagnosing errors while parsing a template, and instead issues an error only upon instantiation of the template. You can also use this flag to downgrade such errors into warnings with Wno-error= or -fpermissive.
• -Wno-template-id-cdtor (C++ and Objective-C++ only) — disables the warning about using simple-template-id as the declarator-id of a constructor or destructor, which became invalid in C++20 via DR 2237. Like, look:

template<typename T> struct S {
  S<T>(); // should be S();
  ~S<T>();  // should be ~S();
};

-Wtemplate-id-cdtor is enabled by default with -std=c++20; it is also enabled by -Wc++20-compat.

• -Wtemplate-names-tu-local — warns when a template body hides an exposure of a translation-unit-local entity. In most cases, referring to a translation-unit-local entity (like an internal linkage declaration) inside an entity that is emitted into a module's CMI is an error. BUT — within the initializer of a variable, or in the body of a non-inline function, that's not an exposure and no error gets emitted 🙈

This can cause variable or function templates to accidentally become unusable if they reference such an entity, because other translation units that import the template will never be able to instantiate it. This warning tries to sniff out cases where that might happen. The presence of an explicit instantiation silences the warning.

This flag is enabled by -Wextra.

• -Wno-terminate (C++ and Objective-C++ only) — disables the warning about a throw-expression that's about to immediately result in a call to terminate.
• -Wno-vexing-parse (C++ and Objective-C++ only) — the most vexing parse, the syntactic ambiguity of the century 😭 warns about cases when a declaration looks like a variable definition, but the C++ language requires it to be interpreted as a function declaration. Example:

void f(double a) {
  int i();        // extern int i (void);
  int n(int(a));  // extern int n (int);
}

Another example:

struct S { S(int); };
void f(double a) {
  S x(int(a));   // extern struct S x (int);
  S y(int());    // extern struct S y (int (*) (void));
  S z();         // extern struct S z (void);
}

The warning will suggest options for how to deal with such an ambiguity; e.g., it can suggest removing the parentheses or using braces instead 🎀

This warning is enabled by default.

• -Wno-class-conversion (C++ and Objective-C++ only) — don't warn when a conversion function converts an object to the same type, to a base class of that type, or to void; such a conversion function will never be called anyway.
• -Wvolatile (C++ and Objective-C++ only) — warns about deprecated uses of the volatile qualifier. This includes postfix and prefix ++ and -- expressions of volatile-qualified types, using simple assignments where the left operand is a volatile-qualified non-class type for their value, compound assignments where the left operand is a volatile-qualified non-class type, volatile-qualified function return type, volatile-qualified parameter type, and structured bindings of a volatile-qualified type. This usage was deprecated in C++20.

Enabled by default with -std=c++20. Before -std=c++20, enabled with explicit -Wdeprecated.

• -Waligned-new — warns about a new-expression of a type that requires greater alignment than the alignof(std::max_align_t) but uses an allocation function without an explicit alignment parameter. This option is enabled by -Wall.

Normally this only warns about global allocation functions, but -Waligned-new=all also warns about class member allocation functions 👑

• -Wno-placement-new
• -Wplacement-new=n — warns about placement new expressions with undefined behavior, like constructing an object in a buffer that's smaller than the type of the object. For example, the placement new expression below is diagnosed because it attempts to construct an array of 64 integers in a buffer only 64 bytes large 💀

char buf [64];
new (buf) int[64];

This warning is enabled by default.

• -Wplacement-new=1 — this is the default warning level of -Wplacement-new. At this level the warning is not issued for some strictly undefined constructs that GCC allows as extensions for compatibility with legacy code. For example, the following new expression is not diagnosed at this level even though it has undefined behavior according to the C++ standard, because it writes past the end of the one-element array.

struct S { int n, a[1]; };
S *s = (S *)malloc (sizeof *s + 31 * sizeof s->a[0]);
new (s->a)int [32]();

• -Wplacement-new=2 — at this level, in addition to diagnosing all the same constructs as at level 1, a diagnostic is also issued for placement new expressions that construct an object in the last member of a structure whose type is an array of a single element and whose size is less than the size of the object being constructed. While the previous example would be diagnosed, the following construct makes use of the flexible member array extension to avoid the warning at level 2 ✨

struct S { int n, a[]; };
S *s = (S *)malloc (sizeof *s + 32 * sizeof s->a[0]);
new (s->a)int [32]();

• -Wcatch-value
• -Wcatch-value=n (C++ and Objective-C++ only) — warns about catch handlers that do not catch via reference. With -Wcatch-value=1 (or just -Wcatch-value for short) warn about polymorphic class types that are caught by value. With -Wcatch-value=2 warn about all class types caught by value. With -Wcatch-value=3 warn about all types that are not caught by reference. -Wcatch-value is enabled by -Wall.
• -Wconditionally-supported (C++ and Objective-C++ only) — warns for conditionally-supported (C++11 [intro.defs]) constructs.
• -Wno-defaulted-function-deleted (C++ and Objective-C++ only) — warns when an explicitly defaulted function is deleted by the compiler. That can happen when the function's declared type doesn't match the type of the function that would have been implicitly declared. This warning is enabled by default.
• -Wno-delete-incomplete (C++ and Objective-C++ only) — don't warn when deleting a pointer to incomplete type, which may cause undefined behavior at runtime 🙈 This warning is enabled by default.
• -Wextra-semi (C++, Objective-C++ only) — warns about redundant semicolons. There are various contexts where an extra semicolon can show up. One is a semicolon after in-class function definitions, which is valid in all C++ dialects (and is never a pedwarn):

struct S {
  void foo () {};
};

Another is an extra semicolon at namespace scope, which has been allowed since C++11 (therefore is a pedwarn in C++98):

struct S {
};
;

And yet another is an extra semicolon in class definitions, which has been allowed since C++11 (therefore is a pedwarn in C++98):

struct S {
  int a;
  ;
};

• -Wno-global-module (C++ and Objective-C++ only) — disables the diagnostic for when the global module fragment of a module unit doesn't consist only of preprocessor directives.
• -Wno-inaccessible-base (C++, Objective-C++ only) — controls warnings when a base class is inaccessible in a class derived from it due to ambiguity. The warning is enabled by default. Note that the warning for ambiguous virtual bases is enabled by the -Wextra option 🌈

struct A { int a; };

struct B : A { };

struct C : B, A { };

• -Wno-inherited-variadic-ctor — suppresses warnings about use of C++11 inheriting constructors when the base class inherited from has a C variadic constructor; the warning is on by default because the ellipsis is not inherited, no cap 🫶
• -Wno-invalid-offsetof (C++ and Objective-C++ only) — suppresses warnings from applying the offsetof macro to a non-POD type. According to the 2014 ISO C++ standard, applying offsetof to a non-standard-layout type is undefined. In existing C++ implementations, however, offsetof typically gives meaningful results. This flag is for users who are aware that they're writing nonportable code and who have deliberately chosen to ignore the warning about it 💅

The restrictions on offsetof may be relaxed in a future version of the C++ standard.

• -Wsized-deallocation (C++ and Objective-C++ only) — bestie, this one's about keeping your delete in balance. It yells at you when you define an unsized deallocation function

void operator delete (void *) noexcept;
void operator delete[] (void *) noexcept;

but then don't define the matching sized one

void operator delete (void *, std::size_t) noexcept;
void operator delete[] (void *, std::size_t) noexcept;

or the other way around. Either direction, it's gonna notice. It's turned on by -Wextra together with -fsized-deallocation. 👑

• -Wsuggest-final-types — not me snitching on your types 🫣. This warns about types that have virtual methods where your code quality would genuinely be better if you slapped the C++11 final specifier on them, or (if that's not possible) declared them in an anonymous namespace. Why do we care? Because it lets GCC devirtualize polymorphic calls way more aggressively — and devirtualization is the glow-up ✨. Heads up: this warning hits harder with link-time optimization, since then the info about the class hierarchy graph is way more complete. iykyk.
• -Wsuggest-final-methods — same energy, but for virtual methods instead of types 💅. It warns about virtual methods where things would be better if the method got declared with the C++11 final specifier, or if possible, its type got declared in an anonymous namespace or with the final specifier. Also more effective with link-time optimization, since the class hierarchy graph info is more complete there. Pro tip from the manpage itself: it's recommended to consider the suggestions of -Wsuggest-final-types first, then rebuild with your new annotations. Order matters, bestie 🎀.
• -Wsuggest-override — the "you forgot the vibe marker" warning 🚨. It warns about overriding virtual functions that aren't marked with the override keyword. Just add the word. It's free. It's iconic. Do it.
• -Wno-conversion-null (C++ and Objective-C++ only) — turns off the yelling about conversions between NULL and non-pointer types. FYI -Wconversion-null is enabled by default, so this is you actively choosing peace 🙈.

---

🌈 Options Controlling Objective-C and Objective-C++ Dialects 🌈

(NOTE: This manual does not describe the Objective-C and Objective-C++ languages themselves.)

Okay so this section is all about the command-line options that are only meaningful for Objective-C and Objective-C++ programs. You can still use most of the language-independent GNU compiler options too, don't worry 💖. Like, you might compile a file some_class.m like this:

gcc -g -fgnu-runtime -O -c some_class.m

In this example, -fgnu-runtime is an option meant only for Objective-C and Objective-C++ programs; the other options you can use with any language GCC supports, no gatekeeping 🫶.

Also, since Objective-C is an extension of the C language, Objective-C compilations may also use options specific to the C front-end (e.g. -Wtraditional). Similarly, Objective-C++ compilations may use C++-specific options (e.g. -Wabi). Twinsies energy ✨.

Here's the list of options that are only for compiling Objective-C and Objective-C++ programs:

• -fconstant-string-class=class-name — use class-name as the name of the class to instantiate for each literal string specified with the syntax @"...". The default class name is NXConstantString if the GNU runtime is being used, and NSConstantString if the NeXT runtime is being used (see below). On Darwin / macOS platforms, the -fconstant-cfstrings option, if also present, **overrides the -fconstant-string-class setting and causes @"..." literals to be laid out as constant CoreFoundation strings. Note that -fconstant-cfstrings is an alias for the target-specific -mconstant-cfstrings equivalent. 👑
• -fgnu-runtime — generate object code compatible with the standard GNU Objective-C runtime. This is the default for most types of systems 🌸.
• -fnext-runtime — generate output compatible with the NeXT runtime. This is the default for NeXT-based systems, including Darwin / macOS. The macro __NEXT_RUNTIME__ is predefined if (and only if) this option is used 🦋.
• -fno-nil-receivers — assume that all Objective-C message dispatches ([receiver message:arg]) in this translation unit ensure that the receiver is not nil. This allows for more efficient entry points in the runtime to be used. Heads up: this option is only available in conjunction with the NeXT runtime and ABI version 0 or 1 — no mixing, bestie 💅.
• *-fobjc-abi-version=n* — use version n of the Objective-C ABI for the selected runtime. Currently this option is only supported for the NeXT runtime. In that case:
  • Version 0 is the traditional (32-bit) ABI without support for properties and other Objective-C 2.0 additions.
  • Version 1 is the traditional (32-bit) ABI with support for properties and other Objective-C 2.0 additions.
  • Version 2 is the modern (64-bit) ABI.

If nothing is specified, the default is Version 0 on 32-bit target machines, and Version 2 on 64-bit target machines. Slay, automatic.

• -fobjc-call-cxx-cdtors — for each Objective-C class, check if any of its instance variables is a C++ object with a non-trivial default constructor. If so, synthesize a special - (id) .cxx_construct instance method which runs non-trivial default constructors on any such instance variables, in order, and then returns self. Similarly, check if any instance variable is a C++ object with a non-trivial destructor, and if so, synthesize a special - (void) .cxx_destruct method which runs all such default destructors, in reverse order 🔄.

The - (id) .cxx_construct and - (void) .cxx_destruct methods thusly generated only operate on instance variables declared in the current Objective-C class, and not those inherited from superclasses. It is the responsibility of the Objective-C runtime to invoke all such methods in an object's inheritance hierarchy. The - (id) .cxx_construct methods are invoked by the runtime immediately after a new object instance is allocated; the - (void) .cxx_destruct methods are invoked immediately before the runtime deallocates an object instance.

As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has support for invoking the - (id) .cxx_construct and - (void) .cxx_destruct methods 🫶.

• -fobjc-direct-dispatch — allow fast jumps to the message dispatcher. On Darwin this is accomplished via the comm page 🚀.
• -fobjc-exceptions — enable syntactic support for structured exception handling in Objective-C, similar to what C++ offers. This option is required to use the Objective-C keywords @try, @throw, @catch, @finally and @synchronized. It's available with both the GNU runtime and the NeXT runtime (but not available in conjunction with the NeXT runtime on Mac OS X 10.2 and earlier — sorry, ancient Macs 😭).
• -fobjc-gc — enable garbage collection (GC) in Objective-C and Objective-C++ programs. This option is only available with the NeXT runtime; the GNU runtime has a different garbage collection implementation that does not require special compiler flags. No cap 🧚.
• -fobjc-nilcheck — for the NeXT runtime with version 2 of the ABI, check for a nil receiver in method invocations before doing the actual method call. This is the default, and can be disabled using -fno-objc-nilcheck. Important caveat: class methods and super calls are never checked for nil in this way no matter what this flag is set to 😤. Currently this flag does nothing when the GNU runtime, or an older version of the NeXT runtime ABI, is used.
• -fobjc-std=objc1 — conform to the language syntax of Objective-C 1.0, the language recognized by GCC 4.0. This only affects the Objective-C additions to the C/C++ language; it does not affect conformance to C/C++ standards, which is controlled by the separate C/C++ dialect option flags. When this option is used with the Objective-C or Objective-C++ compiler, any Objective-C syntax that isn't recognized by GCC 4.0 is rejected. This is useful if you need to make sure your Objective-C code can be compiled with older versions of GCC — time-travel compatibility, lowkey iconic ⭐.
• -freplace-objc-classes — emit a special marker instructing ld(1) not to statically link in the resulting object file, and allow dyld(1) to load it in at run time instead. This is used in conjunction with the Fix-and-Continue debugging mode, where the object file in question may be recompiled and dynamically reloaded in the course of program execution, without the need to restart the program itself. Currently, Fix-and-Continue functionality is only available in conjunction with the NeXT runtime on Mac OS X 10.3 and later 🔥.
• -fzero-link — when compiling for the NeXT runtime, the compiler ordinarily replaces calls to objc_getClass("...") (when the name of the class is known at compile time) with static class references that get initialized at load time, which improves run-time performance. Specifying the -fzero-link flag suppresses this behavior and causes calls to objc_getClass("...") to be retained. This is useful in Zero-Link debugging mode, since it allows individual class implementations to be modified during program execution. Meanwhile, the GNU runtime currently always retains calls to objc_get_class("...") regardless of command-line options. 💀
• -fno-local-ivars — by default, instance variables in Objective-C can be accessed as if they were local variables from within the methods of the class they're declared in. This can lead to shadowing between instance variables and other variables declared either locally inside a class method or globally with the same name 😵‍💫. Specifying the -fno-local-ivars flag disables this behavior, thus avoiding variable shadowing issues. Name collisions? Not on my watch.
• -fivar-visibility=[public|protected|private|package] — set the default instance variable visibility to the specified option, so that instance variables declared outside the scope of any access modifier directives default to the specified visibility. Pick your fighter 💅.
• -gen-decls — dump interface declarations for all classes seen in the source file to a file named sourcename.decl. Very "here's everything, laid out" energy 📝.
• -Wassign-intercept (Objective-C and Objective-C++ only) — warn whenever an Objective-C assignment is being intercepted by the garbage collector 👀.
• -Wno-property-assign-default (Objective-C and Objective-C++ only) — do not warn if a property for an Objective-C object has no assign semantics specified. Silence is golden 🎀.
• -Wno-protocol (Objective-C and Objective-C++ only) — bestie, if a class says it implements a protocol, GCC's default vibe is to yell at you for every single method in that protocol the class doesn't implement itself. Even if the implementation is inherited from the superclass 😭. So flip on -Wno-protocol and suddenly inherited methods count as implemented, no warning for them. Peace. ✨
• -Wobjc-root-class (Objective-C and Objective-C++ only) — warns when a class interface has no superclass. Most classes are gonna inherit from something like NSObject (or Object), that's the norm. But if you're intentionally declaring root classes, you can silence the warning by marking those interfaces with __attribute__((objc_root_class)). Slay, do it on purpose. 👑
• -Wselector (Objective-C and Objective-C++ only) — warns you if multiple methods of different types for the same selector show up during compilation. And the check happens on the method list at the final stage of compilation. Plus, extra check: for every selector in a @selector(...) expression, it makes sure a matching method was found during compilation. ⚠️ Important vibe check: since these checks only scan the method table at the very end, you get no warnings if that final stage never happens — like if an error kills the compilation early, or if you're using -fsyntax-only. No end stage, no warnings, that's just how it is.
• -Wstrict-selector-match (Objective-C and Objective-C++ only) — warns if there are multiple methods with differing argument and/or return types for a given selector, when you're trying to send a message with that selector to a receiver of type id or Class. When this flag is off (and off is the default, bestie), the compiler skips those warnings whenever the differences it found are confined to types that share the same size and alignment. iykyk. 🧚
• -Wundeclared-selector (Objective-C and Objective-C++ only) — warns if you've got a @selector(...) expression referring to an undeclared selector. A selector counts as undeclared when no method with that name was declared before that @selector(...) expression — not explicitly in an @interface or @protocol declaration, and not implicitly in an @implementation section. This option does its checks the moment a @selector(...) expression appears, while -Wselector waits until the final stage of compilation. It also enforces that whole coding-style convention that methods and selectors must be declared before you use them. Order matters, periodt. 💅
• -print-objc-runtime-info — generates a C header describing the largest structure that gets passed by value, if there even is one.

---

🌈 Options Controlling OpenMP and OpenACC 🌈

GCC is out here supporting OpenMP extensions for the C, C++, and Fortran languages with -fopenmp. And same energy for OpenACC extensions, supported in all three languages with -fopenacc. Iconic trio, fr fr. ✨

• -foffload=disable, -foffload=default, -foffload=target-list — these say which OpenMP and OpenACC offload targets code should be generated for. The default behavior — which is the same as -foffload=default — generates code for all supported offload targets. The -foffload=disable form generates code only for the host fallback. And -foffload=target-list generates code only for the comma-separated list of offload targets you name. 🎯

Offload targets get specified in GCC's internal target-triplet format. Wanna peek at the list of configured offload targets? Run the compiler with -v and look under OFFLOAD_TARGET_NAMES. Literally that easy.

• -foffload-options=options, -foffload-options=target-triplet-list=options — with -foffload-options=options, GCC passes those specified options to the compilers for all enabled offloading targets. Wanna be picky and target just one (or a few)? Use the -foffload-options=target-list=options form. The target-list is a comma-separated list in the same format as for the -foffload= option. 🫶

Typical command lines look like this:

-foffload-options='-fno-math-errno -ffinite-math-only' \
        -foffload-options=nvptx-none=-latomic
-foffload-options=amdgcn-amdhsa=-march=gfx906

• -fopenacc — enables handling of OpenACC directives #pragma acc in C/C++ and !$acc in free-form Fortran, plus !$acc, c$acc and *$acc in fixed-form Fortran. Heads up: this option implies -pthread, so it's only supported on targets that actually have support for -pthread. No pthread, no party. 🎀
• -fopenacc-dim=geom — sets the default compute dimensions for parallel offload regions that don't explicitly specify them. The geom value is a triple of :-separated sizes, in order: gang, worker, and vector. You can omit a size and it'll use a target-specific default value. 💫
• -fopenmp — enables handling of OpenMP directives #pragma omp, [[omp::directive(...)]], [[omp::decl(...)]], and [[omp::sequence(...)]] in C/C++. In Fortran it enables !$omp and the conditional compilation sentinel !$. And in fixed source form Fortran, those sentinels can also start with c or *. ✨

Same deal as before: this option implies -pthread, so it's only supported on targets that have support for -pthread. And -fopenmp implies -fopenmp-simd. Chain reaction, love that for us. 🔥

• -fopenmp-simd — enables handling of OpenMP's simd, declare simd, declare reduction, assume, ordered, scan and loop directive, plus combined or composite directives with simd as a constituent — via #pragma omp, [[omp::directive(...)]], [[omp::sequence(...)]] and [[omp::decl(...)]] in C/C++, and !$omp in Fortran. It also enables the conditional compilation sentinel !$ in Fortran, and in fixed source form Fortran the sentinels can also start with c or *. Every other OpenMP directive? Ignored. 💀 And unless -fopenmp is additionally specified, the loop region binds to the current task region, independent of whatever bind clause you specified. Noted.
• -fopenmp-target-simd-clone, -fopenmp-target-simd-clone=device-type — on top of generating SIMD clones for functions marked with the declare simd directive, GCC will also generate clones for functions marked with the OpenMP declare target directive that are suitable for vectorization when this option is in effect. The device-type can be one of none, host, nohost, and any — which correspond to keywords for the device_type clause of the declare target directive — and clones get generated for the intersection of devices specified. ✨

So: -fopenmp-target-simd-clone is equivalent to -fopenmp-target-simd-clone=any, and -fno-openmp-target-simd-clone is equivalent to -fopenmp-target-simd-clone=none.

At -O2 and higher (but not -Os or -Og) this optimization defaults to -fopenmp-target-simd-clone=nohost; otherwise it's disabled by default.

---

💅 Options to Control Diagnostic Messages Formatting 💅

Historically, diagnostic messages have been formatted with zero regard for the output device's aspect (like its width, etc.) — rude, honestly. So the options below let you control the formatting algorithm for diagnostic messages, like how many characters per line, and how often source location info gets reported. 🎀

Fair warning, bestie: some language front ends may simply not honor these options. It's giving "suggestions, not guarantees". 🙈

• -fmessage-length=n — tries to format error messages so they fit on lines of about n characters. If n is zero, no line-wrapping happens at all; each error message shows up on a single line. Fun fact: that's the default for all front ends. ✨

Note — this option also affects the display of the #error and #warning pre-processor directives, plus the deprecated function/type/variable attribute. It does not, however, affect the pragma GCC warning and pragma GCC error pragmas. Okay so those two are just built different.

• -fdiagnostics-plain-output — asks for diagnostic output to look as plain as humanly possible. Super useful when you're running dejagnu or other utilities that need to parse diagnostics output and prefer it stay more stable over time. Right now, -fdiagnostics-plain-output is equivalent to all of these options together: -fno-diagnostics-show-caret -fno-diagnostics-show-line-numbers -fdiagnostics-color=never -fdiagnostics-urls=never -fdiagnostics-path-format=separate-events -fdiagnostics-text-art-charset=none -fno-diagnostics-show-event-links. And in the future, if GCC changes the default appearance of its diagnostics, the corresponding option to disable that new behavior gets added to this list. Future-proofing, we love to see it. 🫶
• -fdiagnostics-show-location=once — only meaningful in line-wrapping mode. Tells the diagnostic messages reporter to emit source location info once — so if a message is too long to fit on a single physical line and has to be wrapped, the source location won't get emitted (as a prefix) again, over and over, on all the continuation lines. That's the default behavior, btw. 💖
• -fdiagnostics-show-location=every-line — only meaningful in line-wrapping mode. Tells the reporter to emit the same source location information (as a prefix) for every physical line that results from breaking up a message that's too long to fit on a single line. Consistent. ✨
• -fdiagnostics-color[=WHEN], -fno-diagnostics-color — use color in diagnostics, bestie! WHEN is never, always, or auto. The default depends on how the compiler was configured — it can be any of those WHEN options, or also never if the GCC_COLORS environment variable isn't present in the environment, and auto otherwise. auto makes GCC use color only when standard error is a terminal, and when it's not executing inside an emacs shell. And the shortcut forms: -fdiagnostics-color and -fno-diagnostics-color are aliases for -fdiagnostics-color=always and -fdiagnostics-color=never, respectively. Colorful and concise, we stan. 🌈

The colors are defined by the environment variable GCC_COLORS. Its value is a colon-separated list of capabilities and Select Graphic Rendition (SGR) substrings. SGR commands are interpreted by the terminal or terminal emulator. (See the section in the documentation of your text terminal for permitted values and their meanings as character attributes.) These substring values are integers in decimal representation and can be concatenated with semicolons. Common values to concatenate include 1 for bold, 4 for underline, 5 for blink, 7 for inverse, 39 for default foreground color, 30 to 37 for foreground colors, 90 to 97 for 16-color mode foreground colors, 38;5;0 to 38;5;255 for 88-color and 256-color modes foreground colors, 49 for default background color, 40 to 47 for background colors, 100 to 107 for 16-color mode background colors, and 48;5;0 to 48;5;255 for 88-color and 256-color modes background colors.

The default GCC_COLORS is

error=01;31:warning=01;35:note=01;36:range1=32:range2=34:locus=01:\
quote=01:path=01;36:fixit-insert=32:fixit-delete=31:\
diff-filename=01:diff-hunk=32:diff-delete=31:diff-insert=32:\
type-diff=01;32:fnname=01;32:targs=35:valid=01;31:invalid=01;32\
highlight-a=01;32:highlight-b=01;34

where 01;31 is bold red, 01;35 is bold magenta, 01;36 is bold cyan, 32 is green, 34 is blue, 01 is bold, and 31 is red. Setting GCC_COLORS to the empty string disables colors. Supported capabilities are as follows. 🎨

• "error=" — the SGR substring for error: markers.
• "warning=" — the SGR substring for warning: markers.
• "note=" — the SGR substring for note: markers.
• "path=" — the SGR substring for colorizing paths of control-flow events as printed via -fdiagnostics-path-format=, like the identifiers of individual events and lines indicating interprocedural calls and returns.
• "range1=" — the SGR substring for the first additional range.
• "range2=" — the SGR substring for the second additional range.
• "locus=" — the SGR substring for location information, file:line or file:line:column etc.
• "quote=" — the SGR substring for information printed within quotes.
• "fnname=" — the SGR substring for names of C++ functions.
• "targs=" — the SGR substring for C++ function template parameter bindings.
• "fixit-insert=" — the SGR substring for fix-it hints suggesting text to be inserted or replaced.
• "fixit-delete=" — the SGR substring for fix-it hints suggesting text to be deleted.
• "diff-filename=" — the SGR substring for filename headers within generated patches.
• "diff-hunk=" — the SGR substring for the starts of hunks within generated patches.
• "diff-delete=" — the SGR substring for deleted lines within generated patches.
• "diff-insert=" — the SGR substring for inserted lines within generated patches.
• "type-diff=" — the SGR substring for highlighting mismatching types within template arguments in the C++ frontend.
• "valid=" — the SGR substring for highlighting valid elements within text art diagrams.
• "invalid=" — the SGR substring for highlighting invalid elements within text art diagrams.
• "highlight-a=", "highlight-b=" — SGR substrings for contrasting two different things within diagnostics, like a pair of mismatching types. See -fdiagnostics-show-highlight-colors.

---

✨ -fdiagnostics-urls[=WHEN] ✨

Use escape sequences to embed URLs in diagnostics. 🔗 Like, when -fdiagnostics-show-option emits text showing the command-line option controlling a diagnostic, it embeds a URL for the documentation of that option. Bestie. So helpful.

WHEN is never, always, or auto. auto makes GCC use URL escape sequences only when the standard error is a terminal, and when not executing in an emacs shell or any graphical terminal which is known to be incompatible with this feature, see below.

The default depends on how the compiler has been configured. It can be any of the above WHEN options. 👑

GCC can also be configured (via the --with-diagnostics-urls=auto-if-env configure-time option) so that the default is affected by environment variables. Under such a configuration, GCC defaults to using auto if either GCC_URLS or TERM_URLS environment variables are present and non-empty in the environment of the compiler, or never if neither are.

However, even with -fdiagnostics-urls=always the behavior is dependent on those environment variables: If GCC_URLS is set to empty or no, do not embed URLs in diagnostics. If set to st, URLs use ST escape sequences. If set to bel, the default, URLs use BEL escape sequences. Any other non-empty value enables the feature. If GCC_URLS is not set, use TERM_URLS as a fallback. Note: ST is an ANSI escape sequence, string terminator ESC \, BEL is an ASCII character, CTRL+G that usually sounds like a beep. 🛎️

At this time GCC tries to detect also a few terminals that are known to not implement the URL feature, and have bugs or at least had bugs in some versions that are still in use, where the URL escapes are likely to misbehave, i.e. print garbage on the screen. That list is currently xfce4-terminal, certain known to be buggy gnome-terminal versions, the linux console, and mingw. This check can be skipped with the -fdiagnostics-urls=always.

ngl those terminals are NOT passing the vibe check 💀 skip the check with -fdiagnostics-urls=always if you must.

---

🌸 -fno-diagnostics-show-option 🌸

By default, each diagnostic emitted includes text indicating the command-line option that directly controls the diagnostic (if such an option is known to the diagnostic machinery). Specifying the -fno-diagnostics-show-option flag suppresses that behavior.

🎀 -fno-diagnostics-show-caret 🎀

By default, each diagnostic emitted includes the original source line and a caret ^ indicating the column. This option suppresses this information. The source line is truncated to n characters, if the -fmessage-length=n option is given. When the output is done to the terminal, the width is limited to the width given by the COLUMNS environment variable or, if not set, to the terminal width.

🦋 -fno-diagnostics-show-labels 🦋

By default, when printing source code (via -fdiagnostics-show-caret), diagnostics can label ranges of source code with pertinent information, such as the types of expressions:

            printf ("foo %s bar", long_i + long_j);
                         ~^       ~~~~~~~~~~~~~~~
                          |              |
                          char *         long int

This option suppresses the printing of these labels (in the example above, the vertical bars and the "char *" and "long int" text).

💅 -fno-diagnostics-show-event-links 💅

By default, when printing execution paths (via -fdiagnostics-path-format=inline-events), GCC will print lines connecting related events, such as the line connecting events 1 and 2 in:

            3 |   if (p)
              |       ^
              |       |
              |       (1) following `false' branch (when `p' is NULL)... ->-+
              |                                                             |
              |                                                             |
              |+------------------------------------------------------------+
            4 ||    return 0;
            5 ||  return *p;
              ||         ~
              ||         |
              |+-------->(2) ...to here
              |          (3) dereference of NULL `p'

This option suppresses the printing of such connector lines.

🌈 -fno-diagnostics-show-cwe 🌈

Diagnostic messages can optionally have an associated CWE (https://cwe.mitre.org/index.html) identifier. GCC itself only provides such metadata for some of the -fanalyzer diagnostics. GCC plugins may also provide diagnostics with such metadata. By default, if this information is present, it will be printed with the diagnostic. This option suppresses the printing of this metadata.

⭐ -fno-diagnostics-show-rules ⭐

Diagnostic messages can optionally have rules associated with them, such as from a coding standard, or a specification. GCC itself does not do this for any of its diagnostics, but plugins may do so. By default, if this information is present, it will be printed with the diagnostic. This option suppresses the printing of this metadata.

👑 -fno-diagnostics-show-highlight-colors 👑

GCC can use color for emphasis and contrast when printing diagnostic messages and quoting the user's source. 🎨

For example, in

        demo.c: In function `test_bad_format_string_args':
        ../../src/demo.c:25:18: warning: format `%i' expects argument of type `int', but argument 2 has type `const char *' [-Wformat=]
           25 |   printf("hello %i", msg);
              |                 ~^   ~~~
              |                  |   |
              |                  int const char *
              |                 %s

• the %i and "int" in the message and the "int" in the quoted source are colored using "highlight-a" (bold green by default), and
• the "const char *" in the message and in the quoted source are both colored using "highlight-b" (bold blue by default).

The intent is to draw the reader's eyes to the relationships between the various aspects of the diagnostic message and the source, using color to group related elements and distinguish between mismatching ones. aesthetic AND functional, we love to see it 💖

This additional colorization is enabled by default if color printing is enabled (as per -fdiagnostics-color=), but it can be separately disabled via -fno-diagnostics-show-highlight-colors.

🍓 -fno-diagnostics-show-line-numbers 🍓

By default, when printing source code (via -fdiagnostics-show-caret), a left margin is printed, showing line numbers. This option suppresses this left margin.

🧚 -fdiagnostics-minimum-margin-width=width 🧚

This option controls the minimum width of the left margin printed by -fdiagnostics-show-line-numbers. It defaults to 6.

• -fdiagnostics-parseable-fixits — bestie the compiler is about to start handing out fix-it hints in a machine-parseable format, literally perfect for your IDE to gobble up. For each fix-it, a line gets printed right after the relevant diagnostic, kicking off with the string "fix-it:". Like so:

fix-it:"test.c":{45:3-45:21}:"gtk_widget_show_all"

The location is expressed as a half-open range, expressed as a count of bytes, starting at byte 1 for the initial column. In the example above, bytes 3 through 20 of line 45 of "test.c" are to be replaced with the given string:

00000000011111111112222222222
12345678901234567890123456789
  gtk_widget_showall (dlg);
  ^^^^^^^^^^^^^^^^^^
  gtk_widget_show_all

The filename and replacement string escape backslash as "\\", tab as "\t", newline as "\n", double quotes as "\"", and non-printable characters as octal (e.g. vertical tab as "\013").

An empty replacement string indicates that the given range is to be removed. An empty range (e.g. "45:3-45:3") indicates that the string is to be inserted at the given position. 🎀

• -fdiagnostics-generate-patch — print fix-it hints to stderr in unified diff format, after any diagnostics are printed. She's giving patch energy, fr fr:

--- test.c
+++ test.c
@ -42,5 +42,5 @

 void show_cb(GtkDialog *dlg)
 {
-  gtk_widget_showall(dlg);
+  gtk_widget_show_all(dlg);
 }

The diff may or may not be colorized, following the same rules as for diagnostics (see -fdiagnostics-color). 🌈

• -fdiagnostics-show-template-tree — okay so in the C++ frontend, when printing diagnostics showing mismatching template types, like:

  could not convert 'std::map<int, std::vector<double> >()'
    from 'map<[...],vector<double>>' to 'map<[...],vector<float>>

the -fdiagnostics-show-template-tree flag enables printing a tree-like structure showing the common and differing parts of the types, like:

  map<
    [...],
    vector<
      [double != float]>>

The parts that differ are highlighted with color ("double" and "float" in this case). Obsessed with how much easier that is to read. 👑

• -fno-elide-type — so by default when the C++ frontend prints diagnostics showing mismatching template types, common parts of the types get printed as "[...]" to simplify the error message. For example:

  could not convert 'std::map<int, std::vector<double> >()'
    from 'map<[...],vector<double>>' to 'map<[...],vector<float>>

Specifying the -fno-elide-type flag suppresses that behavior. This flag also affects the output of the -fdiagnostics-show-template-tree flag. No more hiding, show me everything bestie 💅

• -fdiagnostics-path-format=KIND — specify how to print paths of control-flow events for diagnostics that have such a path associated with them.

KIND is none, separate-events, or inline-events, the default.

none means to not print diagnostic paths.

separate-events means to print a separate "note" diagnostic for each event within the diagnostic. For example:

test.c:29:5: error: passing NULL as argument 1 to 'PyList_Append' which requires a non-NULL parameter
test.c:25:10: note: (1) when 'PyList_New' fails, returning NULL
test.c:27:3: note: (2) when 'i < count'
test.c:29:5: note: (3) when calling 'PyList_Append', passing NULL from (1) as argument 1

inline-events means to print the events "inline" within the source code. This view attempts to consolidate the events into runs of sufficiently-close events, printing them as labelled ranges within the source.

For example, the same events as above might be printed as:

  'test': events 1-3
   25 |   list = PyList_New(0);
      |          ^~~~~~~~~~~~~
      |          |
      |          (1) when 'PyList_New' fails, returning NULL
   26 |
   27 |   for (i = 0; i < count; i++) {
      |   ~~~
      |   |
      |   (2) when 'i < count'
   28 |     item = PyLong_FromLong(random());
   29 |     PyList_Append(list, item);
      |     ~~~~~~~~~~~~~~~~~~~~~~~~~
      |     |
      |     (3) when calling 'PyList_Append', passing NULL from (1) as argument 1

Interprocedural control flow is shown by grouping the events by stack frame, and using indentation to show how stack frames are nested, pushed, and popped.

For example:

  'test': events 1-2
    |
    |  133 | {
    |      | ^
    |      | |
    |      | (1) entering 'test'
    |  134 |   boxed_int *obj = make_boxed_int (i);
    |      |                    ~~~~~~~~~~~~~~~~~~
    |      |                    |
    |      |                    (2) calling 'make_boxed_int'
    |
    +--> 'make_boxed_int': events 3-4
           |
           |  120 | {
           |      | ^
           |      | |
           |      | (3) entering 'make_boxed_int'
           |  121 |   boxed_int *result = (boxed_int *)wrapped_malloc (sizeof (boxed_int));
           |      |                                    ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
           |      |                                    |
           |      |                                    (4) calling 'wrapped_malloc'
           |
           +--> 'wrapped_malloc': events 5-6
                  |
                  |    7 | {
                  |      | ^
                  |      | |
                  |      | (5) entering 'wrapped_malloc'
                  |    8 |   return malloc (size);
                  |      |          ~~~~~~~~~~~~~
                  |      |          |
                  |      |          (6) calling 'malloc'
                  |
    <-------------+
    |
 'test': event 7
    |
    |  138 |   free_boxed_int (obj);
    |      |   ^~~~~~~~~~~~~~~~~~~~
    |      |   |
    |      |   (7) calling 'free_boxed_int'
    |
(etc)

• -fdiagnostics-show-path-depths — this option provides extra deets when printing control-flow paths associated with a diagnostic. 🫶

If this is option is provided then the stack depth will be printed for each run of events within -fdiagnostics-path-format=inline-events. If provided with -fdiagnostics-path-format=separate-events, then the stack depth and function declaration will be appended when printing each event.

This is intended for use by GCC developers and plugin developers when debugging diagnostics that report interprocedural control flow. It's giving dev-tools only, iykyk.

• -fno-show-column — do not print column numbers in diagnostics. This may be necessary if diagnostics are being scanned by a program that does not understand the column numbers, such as dejagnu.
• -fdiagnostics-column-unit=UNIT — select the units for the column number. This affects traditional diagnostics (in the absence of -fno-show-column), as well as JSON format diagnostics if requested.

The default UNIT, display, considers the number of display columns occupied by each character. This may be larger than the number of bytes required to encode the character, in the case of tab characters, or it may be smaller, in the case of multibyte characters. For example, the character "GREEK SMALL LETTER PI (U+03C0)" occupies one display column, and its UTF-8 encoding requires two bytes; the character "SLIGHTLY SMILING FACE (U+1F642)" occupies two display columns, and its UTF-8 encoding requires four bytes. 😭 the way unicode really said "sike" to your byte math

Setting UNIT to byte changes the column number to the raw byte count in all cases, as was traditionally output by GCC prior to version 11.1.0.

• -fdiagnostics-column-origin=ORIGIN — select the origin for column numbers, i.e. the column number assigned to the first column. The default value of 1 corresponds to traditional GCC behavior and to the GNU style guide. Some utilities may perform better with an origin of 0; any non-negative value may be specified. 🌸
• -fdiagnostics-escape-format=FORMAT — when GCC prints pertinent source lines for a diagnostic it normally attempts to print the source bytes directly. However, some diagnostics relate to encoding issues in the source file, such as malformed UTF-8, or issues with Unicode normalization. These diagnostics are flagged so that GCC will escape bytes that are not printable ASCII when printing their pertinent source lines.

This option controls how such bytes should be escaped.

The default FORMAT, unicode displays Unicode characters that are not printable ASCII in the form <U+XXXX>, and bytes that do not correspond to a Unicode character validly-encoded in UTF-8-encoded will be displayed as hexadecimal in the form <XX>.

For example, a source line containing the string before followed by the Unicode character U+03C0 ("GREEK SMALL LETTER PI", with UTF-8 encoding 0xCF 0x80) followed by the byte 0xBF (a stray UTF-8 trailing byte), followed by the string after will be printed for such a diagnostic as:

 before<U+03C0><BF>after

Setting FORMAT to bytes will display all non-printable-ASCII bytes in the form <XX>, thus showing the underlying encoding of non-ASCII Unicode characters. For the example above, the following will be printed:

 before<CF><80><BF>after

• -fdiagnostics-text-art-charset=CHARSET — some diagnostics can contain "text art" diagrams: visualizations created from text, intended to be viewed in a monospaced font. 🎨

This option selects which characters should be used for printing such diagrams, if any. CHARSET is none, ascii, unicode, or emoji. Not the emoji charset, bestie — actual emoji in your diagnostics, iconic beyond words 💖

So, bestie, about those diagrams... The none value? It just ~suppresses~ the printing of such diagrams, periodt. The ascii value makes sure those diagrams are pure ASCII ("ASCII art") — old school, but iconic. The unicode value lets you use unicode drawing characters conservatively (like box-drawing characters ✨). And the emoji value? Oh honey, it further adds the possibility of emoji in the output (like emitting U+26A0 WARNING SIGN followed by U+FE0F VARIATION SELECTOR-16 to select the emoji variant of the character). The default is emoji, except when the environment variable LANG is set to C, in which case the default is ascii. No cap.

• -fdiagnostics-format=FORMAT — Pick a whole different vibe for printing diagnostics. FORMAT can be text, sarif-stderr, sarif-file, json, json-stderr, or json-file. Using this option replaces any additional "output sinks" added by -fdiagnostics-add-output=, or that set by -fdiagnostics-set-output=. The default is text. The sarif-stderr and sarif-file formats both emit diagnostics in SARIF Version 2.1.0 format, either to stderr, or to a file named source.sarif, respectively. The various json, json-stderr, and json-file values are deprecated 💀 and refer to a legacy JSON-based output format. The json format is a synonym for json-stderr. The json-stderr and json-file formats are identical, apart from where the JSON is emitted to. With json-stderr, the JSON is emitted to stderr, whereas with json-file it is written to source.gcc.json.
• -fdiagnostics-add-output=DIAGNOSTICS-OUTPUT-SPEC — Add an additional "output sink" for emitting diagnostics. DIAGNOSTICS-OUTPUT-SPEC should specify a scheme, optionally followed by ":" and one or more KEY=VALUE pairs, in this form:

<SCHEME>
<SCHEME>:<KEY>=<VALUE>
<SCHEME>:<KEY>=<VALUE>,<KEY2>=<VALUE2>

etc.

Schemes, keys, or values with a name prefixed "experimental" may change or be removed without notice. 🚨

SCHEME can be:

• text — Emit diagnostics to stderr using GCC's classic text output format. Supported keys are:
  • color=[yes|no] — Override colorization settings from -fdiagnostics-color for this text output.
  • experimental-nesting=[yes|no] — Enable an experimental mode that emphasizes hierarchical relationships within diagnostics messages, displaying location information on separate lines.
  • experimental-nesting-show-locations=[yes|no] — If "experimental-nesting=yes", then by default locations are shown; set this key to "no" to disable printing such locations. This exists for use by GCC developers, for writing DejaGnu test cases.
  • experimental-nesting-show-levels=[yes|no] — This is a debugging option for use with "experimental-nesting=yes". Set this key to "yes" to print explicit nesting levels in the output. This exists for use by GCC developers.
• sarif — Emit diagnostics to a file in SARIF format. Supported keys are:
  • file=FILENAME — Specify the filename to write the SARIF output to, potentially with a leading absolute or relative path. If not specified, it defaults to source.sarif.
  • version=[2.1|2.2-prerelease] — Specify the version of SARIF to use for the output. If not specified, defaults to 2.1. "2.2-prerelease" uses an unofficial draft of the future SARIF 2.2 specification and should only be used for experimentation in this release.

For example,

-fdiagnostics-add-output=sarif:version=2.1,file=foo.2.1.sarif
-fdiagnostics-add-output=sarif:version=2.2-prerelease,file=foo.2.2.sarif

would add a pair of outputs, each writing to a different file, using versions 2.1 and 2.2 of the SARIF standard respectively.

In EBNF:

<diagnostics-output-specifier> = <diagnostics-output-name>
                                   | <diagnostics-output-name>, ":", <key-value-pairs>;

<diagnostics-output-name> = "text" | "sarif";

<key-value-pairs> = <key-value-pair>
                      | <key-value-pair> "," <key-value-pairs>;

<key-value-pair> = <key> "=" <value>;

<key> = ? string without a '=' ? ;
<value> = ? string without a ',' ? ;

• -fdiagnostics-set-output=DIAGNOSTICS-OUTPUT-SPEC — This works in a similar way to -fdiagnostics-add-output= except that instead of adding an additional "output sink" for diagnostics, it replaces all existing output sinks, such as from -fdiagnostics-format=, -fdiagnostics-add-output=, or a prior call to -fdiagnostics-set-output=.
• -fno-diagnostics-json-formatting — By default, when JSON is emitted for diagnostics (via -fdiagnostics-format=sarif-stderr, -fdiagnostics-format=sarif-file, -fdiagnostics-format=json, -fdiagnostics-format=json-stderr, -fdiagnostics-format=json-file), GCC will add newlines and indentation to visually emphasize the hierarchical structure of the JSON. Use -fno-diagnostics-json-formatting to suppress this whitespace. It must be passed before the option it is to affect. This is intended for compatibility with tools that do not expect the output to contain newlines, such as that emitted by older GCC releases.

---

🚨 Options to Request or Suppress Warnings 🚨

Warnings are diagnostic messages that report constructions that are not inherently erroneous but that are risky or suggest there may have been an error. 👀

The following language-independent options do not enable specific warnings but control the kinds of diagnostics produced by GCC.

• -fsyntax-only — Check the code for syntax errors, but don't do anything beyond that. Just a vibe check, no actual compiling. 💅
• -fmax-errors=n — Limits the maximum number of error messages to n, at which point GCC bails out rather than attempting to continue processing the source code. If n is 0 (the default), there is no limit on the number of error messages produced. If -Wfatal-errors is also specified, then -Wfatal-errors takes precedence over this option.
• -w — Inhibit all warning messages. Like, mute button for warnings. 🙈
• -Werror — Turn all warnings into errors. No more Mr. Nice Compiler. 🔥
• -Werror= — Turn the specified warning into an error. The specifier for a warning is appended; for example -Werror=switch turns the warnings controlled by -Wswitch into errors. This switch takes a negative form, to be used to negate -Werror for specific warnings; for example -Wno-error=switch makes -Wswitch warnings not be errors, even when -Werror is in effect. The warning message for each controllable warning includes the option that controls the warning. That option can then be used with -Werror= and -Wno-error= as described above. (Printing of the option in the warning message can be disabled using the -fno-diagnostics-show-option flag.) Note that specifying -Werror=foo automatically implies -Wfoo. However, -Wno-error=foo does not imply anything.
• -Wfatal-errors — This option causes the compiler to abort compilation on the first error occurred rather than trying to keep going and printing further error messages. Like, it just gives up immediately. 💀

You can request many specific warnings with options beginning with -W, for example -Wunused-variable to request warnings on declarations of variables that are never used. Each of these specific warning options also has a negative form beginning with -Wno- to turn off warnings; for example, -Wno-unused-variable. This manual lists only one of the two forms, whichever is not the default. For further language-specific options also refer to C++ Dialect Options and Objective-C and Objective-C++ Dialect Options. Additional warnings can be produced by enabling the static analyzer;

Some options, such as -Wall and -Wextra, turn on other options, such as -Wunused, which may turn on further options, such as -Wunused-variable. The combined effect of positive and negative forms is that more specific options have priority over less specific ones, independently of their position in the command line. For options of the same specificity, the last one takes effect. Options enabled or disabled via pragmas take effect as if they appeared at the end of the command line.

When an unrecognized warning option is requested (e.g., -Wunknown-warning), GCC gives an error stating that the option is not recognized. However, if the -Wno- form is used, the behavior is slightly different: no diagnostic is produced for -Wno-unknown-warning unless other diagnostics are being produced. This allows the use of new -Wno- options with old compilers, but if something goes wrong, the compiler warns that an unrecognized option is present.

The effectiveness of some warnings depends on optimizations also being enabled. For example, -Wsuggest-final-types is more effective with link-time optimization. Some other warnings may not be issued at all unless optimization is enabled. While optimization in general improves the efficacy of warnings about control and data-flow problems, in some cases it may also cause false positives.

• -Wpedantic, -pedantic — Issue all the warnings demanded by strict ISO C and ISO C++; diagnose all programs that use forbidden extensions, and some other programs that do not follow ISO C and ISO C++. This follows the version of the ISO C or C++ standard specified by any -std option used. Valid ISO C and ISO C++ programs should compile properly with or without this option (though a rare few require -ansi or a -std option specifying the version of the standard). However, without this option, certain GNU extensions and traditional C and C++ features are supported as well. With this option, they are diagnosed (or rejected with -pedantic-errors). -Wpedantic does not cause warning messages for use of the alternate keywords whose names begin and end with __. This alternate format can also be used to disable warnings for non-ISO __intN types, i.e. __intN__. Pedantic warnings are also disabled in the expression that follows "__extension__". However, only system header files should use these escape routes; application programs should avoid them. Some warnings about non-conforming programs are controlled by options other than -Wpedantic; in many cases they are implied by -Wpedantic but can be disabled separately by their specific option, e.g. -Wpedantic -Wno-pointer-sign.

Okay so real quick about base standards, because this trips people up: when you pick a standard with -std that's actually a GNU extended dialect of C — like gnu90 or gnu99 — there's a whole ✨base standard✨ hiding underneath it, aka the version of ISO C that the GNU dialect is built on top of. -Wpedantic warnings get handed out wherever that base standard says they're required. (And nope, it wouldn't make sense to only warn about stuff that isn't in the GNU C dialect you picked, because GNU dialects of C include literally every feature the compiler supports with that option, so there'd be nothing left to warn about. Math isn't mathing otherwise. 🙈)

• -pedantic-errors — throws a full-on error whenever the base standard (peep -Wpedantic) requires a diagnostic — including some cases where there's undefined behavior at compile-time, and some other cases that don't even stop programs that are valid per the standard from compiling. 🚨 Heads up bestie: this is not the same thing as -Werror=pedantic. That other option is honestly unlikely to be useful, because it only turns into errors the diagnostics controlled by -Wpedantic, while this one also covers required diagnostics that are always enabled or that get controlled by options other than -Wpedantic. Iconic of it tbh.

So if you want the required diagnostics that are warnings by default to be errors instead, but you do not want the -Wpedantic diagnostics switched on too, you can say -pedantic-errors -Wno-pedantic (or -pedantic-errors -Wno-error=pedantic if you want them enabled but as warnings only).

Some required diagnostics are errors by default, but you can downgrade them to warnings using -fpermissive or their specific warning option, e.g. -Wno-error=narrowing.

And some diagnostics for non-ISO practices are controlled by specific warning options other than -Wpedantic, but -pedantic-errors still promotes them to errors. Like, for instance:

-Wattributes (for standard attributes) · -Wchanges-meaning (C++) · -Wcomma-subscript (C++23 or later) · -Wdeclaration-after-statement (C90 or earlier) · -Welaborated-enum-base (C++11 or later) · -Wimplicit-int (C99 or later) · -Wimplicit-function-declaration (C99 or later) · -Wincompatible-pointer-types · -Wint-conversion · -Wlong-long (C90 or earlier) · -Wmain · -Wnarrowing (C++11 or later) · -Wpointer-arith · -Wpointer-sign · -Wincompatible-pointer-types · -Wregister (C++17 or later) · -Wvla (C90 or earlier) · -Wwrite-strings (C++11 or later)

• -fpermissive — downgrades some required diagnostics about nonconformant code from errors to warnings. So yeah, with -fpermissive some nonconforming code will just... compile. 💅 Some C++ diagnostics are controlled only by this flag, but it also downgrades some C and C++ diagnostics that have their own flag:

-Wdeclaration-missing-parameter-type (C and Objective-C only) · -Wimplicit-function-declaration (C and Objective-C only) · -Wimplicit-int (C and Objective-C only) · -Wincompatible-pointer-types (C and Objective-C only) · -Wint-conversion (C and Objective-C only) · -Wnarrowing (C++ and Objective-C++ only) · -Wreturn-mismatch (C and Objective-C only) · -Wtemplate-body (C++ and Objective-C++ only)

Oh and btw, -fpermissive is the default for historic C language modes: -std=c89, -std=gnu89, -std=c90, -std=gnu90. Old school era behavior, keeping it cute and legacy. 🫶

• -Wall — turns on all the warnings about constructions that some users consider questionable, and that are easy to avoid (or easy to tweak so the warning stops), even when macros are in the mix. It also switches on some language-specific warnings described over in C++ Dialect Options and Objective-C and Objective-C++ Dialect Options.

-Wall turns on the following warning flags:

-Waddress -Waligned-new (C++ and Objective-C++ only) -Warray-bounds=1 (only with -O2) -Warray-compare -Warray-parameter=2 -Wbool-compare -Wbool-operation -Wc++11-compat -Wc++14-compat -Wc++17compat -Wc++20compat -Wcatch-value (C++ and Objective-C++ only) -Wchar-subscripts -Wclass-memaccess (C++ and Objective-C++ only) -Wcomment -Wdangling-else -Wdangling-pointer=2 -Wdelete-non-virtual-dtor (C++ and Objective-C++ only) -Wduplicate-decl-specifier (C and Objective-C only) -Wenum-compare (in C/ObjC; this is on by default in C++) -Wenum-int-mismatch (C and Objective-C only) -Wformat=1 -Wformat-contains-nul -Wformat-diag -Wformat-extra-args -Wformat-overflow=1 -Wformat-truncation=1 -Wformat-zero-length -Wframe-address -Wimplicit (C and Objective-C only) -Wimplicit-function-declaration (C and Objective-C only) -Wimplicit-int (C and Objective-C only) -Winfinite-recursion -Winit-self (C++ and Objective-C++ only) -Wint-in-bool-context -Wlogical-not-parentheses -Wmain (only for C/ObjC and unless -ffreestanding) -Wmaybe-uninitialized -Wmemset-elt-size -Wmemset-transposed-args -Wmisleading-indentation (only for C/C++) -Wmismatched-dealloc -Wmismatched-new-delete (C++ and Objective-C++ only) -Wmissing-attributes -Wmissing-braces (only for C/ObjC) -Wmultistatement-macros -Wnarrowing (C++ and Objective-C++ only) -Wnonnull -Wnonnull-compare -Wopenmp-simd (C and C++ only) -Woverloaded-virtual=1 (C++ and Objective-C++ only) -Wpacked-not-aligned -Wparentheses -Wpessimizing-move (C++ and Objective-C++ only) -Wpointer-sign (only for C/ObjC) -Wrange-loop-construct (C++ and Objective-C++ only) -Wreorder (C++ and Objective-C++ only) -Wrestrict -Wreturn-type -Wself-move (C++ and Objective-C++ only) -Wsequence-point -Wsign-compare (C++ and Objective-C++ only) -Wsizeof-array-div -Wsizeof-pointer-div -Wsizeof-pointer-memaccess -Wstrict-aliasing -Wstrict-overflow=1 -Wswitch -Wtautological-compare -Wtrigraphs -Wuninitialized -Wunknown-pragmas -Wunused -Wunused-but-set-variable -Wunused-const-variable=1 (only for C/ObjC) -Wunused-function -Wunused-label -Wunused-local-typedefs -Wunused-value -Wunused-variable -Wuse-after-free=2 -Wvla-parameter -Wvolatile-register-var -Wzero-length-bounds

Lowkey important: some warning flags are not implied by -Wall. Some of them flag constructions that users generally don't consider questionable, but that you might occasionally wanna check for; others flag constructions that are necessary or hard to avoid in some cases, where there's just no simple way to edit the code and shut the warning up. A few of those get enabled by -Wextra, but many of them you have to turn on individually. No shortcuts, bestie. 😭

• -Wextra — enables some extra warning flags that -Wall doesn't. (Fun fact: this option used to be called -W. The older name is still supported, but the newer name is more descriptive. Glow-up complete. ✨)

-Wabsolute-value (only for C/ObjC) -Walloc-size -Wcalloc-transposed-args -Wcast-function-type -Wclobbered -Wdangling-reference (C++ only) -Wdeprecated-copy (C++ and Objective-C++ only) -Wempty-body -Wenum-conversion (only for C/ObjC) -Wexpansion-to-defined -Wignored-qualifiers (only for C/C++) -Wimplicit-fallthrough=3 -Wmaybe-uninitialized -Wmissing-field-initializers -Wmissing-parameter-name (C/ObjC only) -Wmissing-parameter-type (C/ObjC only) -Wold-style-declaration (C/ObjC only) -Woverride-init (C/ObjC only) -Wredundant-move (C++ and Objective-C++ only) -Wshift-negative-value (in C++11 to C++17 and in C99 and newer) -Wsign-compare (C++ and Objective-C++ only) -Wsized-deallocation (C++ and Objective-C++ only) -Wstring-compare -Wtype-limits -Wuninitialized -Wunterminated-string-initialization -Wunused-parameter (only with -Wunused or -Wall) -Wunused-but-set-parameter (only with -Wunused or -Wall)

The option -Wextra also prints warning messages for the following cases:

• A pointer is compared against integer zero with "<", "<=", ">", or ">=".
• (C++ only) An enumerator and a non-enumerator both appear in a conditional expression.
• (C++ only) Ambiguous virtual bases.
• (C++ only) Subscripting an array that has been declared "register".
• (C++ only) Taking the address of a variable that has been declared "register".
• (C++ only) A base class is not initialized in the copy constructor of a derived class.

• -Wabi (C, Objective-C, C++ and Objective-C++ only) — warns about code affected by ABI changes. This covers code that may not be compatible with the vendor-neutral C++ ABI and the psABI for the particular target. Those latter warnings also get controlled separately by -Wpsabi, which is implied by -Wabi. 👑

So here's the tea: since G++ now defaults to updating the ABI with each major release, normally -Wabi only warns about C++ ABI compatibility problems if there's a check that got added later in a release series for an ABI issue discovered after the initial release. But -Wabi warns about more things if you select an older ABI version (with -fabi-version=n).

-Wabi can also be used with an explicit version number to warn about C++ ABI compatibility with a particular -fabi-version level, e.g. -Wabi=2 to warn about changes relative to -fabi-version=2.

If you hand it an explicit version number and -fabi-compat-version isn't specified, the version number from this option gets used for compatibility aliases. If you don't provide an explicit version number with this option, but -fabi-compat-version is specified, then that version number is used for the C++ ABI warnings.

And now the honest part 🙈 — even though effort has been made to warn about all such cases, there are probably some cases that are not warned about, even though G++ is generating incompatible code. There may also be cases where warnings get emitted even though the code that gets generated is compatible.

You should rewrite your code to avoid these warnings if you're concerned about the fact that code generated by G++ may not be binary compatible with code generated by other compilers. Fr fr, this one's not a vibe.

Known incompatibilities in -fabi-version=2 (which was the default from GCC 3.4 to 4.9) include:

• A template with a non-type template parameter of reference type was mangled incorrectly:

extern int N;
template <int &> struct S {};
void n (S<N>) {2}

This was fixed in -fabi-version=3.

• SIMD vector types declared using "__attribute ((vector_size))" were mangled in a non-standard way that does not allow for overloading of functions taking vectors of different sizes.

The mangling was changed in -fabi-version=4.

• "__attribute ((const))" and "noreturn" were mangled as type qualifiers, and "decltype" of a plain declaration was folded away.

These mangling issues were fixed in -fabi-version=5.

• Scoped enumerators getting passed as arguments to a variadic function get promoted just like unscoped ones, which makes va_arg throw a lil tantrum. Ngl on most targets this doesn't actually touch the parameter-passing ABI, because like… there's literally no way to pass an argument smaller than int. So it's giving drama for no reason on those. 💀

Also bestie, the ABI switched up the mangling of template argument packs, const_cast, static_cast, prefix increment/decrement, and a class scope function used as a template argument.

All of this was fixed in -fabi-version=6. Slay.

• Lambdas in default argument scope were being mangled wrong, and the ABI changed the mangling of nullptr_t.

These got fixed in -fabi-version=7.

• When mangling a function type that had function-cv-qualifiers, the un-qualified function type was incorrectly treated as a substitution candidate. Rude.

Fixed in -fabi-version=8, which is the default for GCC 5.1. 💅

• So decltype(nullptr) incorrectly had an alignment of 1, which led to unaligned accesses. Note that this did not affect the ABI of a function with a nullptr_t parameter, because parameters have a minimum alignment. So like, breathe.

This was fixed in -fabi-version=9, the default for GCC 5.2.

• Target-specific attributes that affect the identity of a type — like ia32 calling conventions on a function type (stdcall, regparm, etc.) — did not affect the mangled name, which led to name collisions when function pointers were used as template arguments. Chaos, honestly.

This was fixed in -fabi-version=10, the default for GCC 6.1.

• -Wpsabi (C, Objective-C, C++ and Objective-C++ only) — ok so -Wpsabi turns on warnings about processor-specific ABI changes, like changes in alignment requirements or how function arguments get passed. On several targets — including AArch64, ARM, x86, MIPS, RS6000/PowerPC, and S/390 — these details have changed between different versions of GCC and/or different versions of the C or C++ language standards in ways that affect binary compatibility of compiled code. With -Wpsabi, GCC can detect potentially incompatible usages and warn you about them. Look out bestie, it's a vibe check for your ABI. 🦋

-Wpsabi is enabled by default, and is also implied by -Wabi.

• -Wno-changes-meaning (C++ and Objective-C++ only) — C++ requires that unqualified uses of a name within a class have the same meaning across the complete scope of the class, so declaring the name after using it is ill-formed. Periodt. Like:

struct A;
struct B1 { A a; typedef A A; }; // warning, 'A' changes meaning
struct B2 { A a; struct A { }; }; // error, 'A' changes meaning

By default, the B1 case is only a warning because the two declarations have the same type, while the B2 case is an error. Both diagnostics can be disabled with -Wno-changes-meaning. Or, alternately, the error case can be downgraded to a warning with -Wno-error=changes-meaning or -fpermissive.

Both diagnostics are also suppressed by -fms-extensions.

• -Wchar-subscripts — warns you if an array subscript has type char. This is a super common cause of error, because programmers always forget that this type is signed on some machines. 😭 This warning is enabled by -Wall.
• -Wno-coverage-mismatch — warns you if feedback profiles don't match when you're using the -fprofile-use option. If a source file gets changed between compiling with -fprofile-generate and compiling with -fprofile-use, the files with the profile feedback can fail to match the source file and GCC cannot use the profile feedback information. By default this warning is enabled and treated as an error. -Wno-coverage-mismatch can be used to disable the warning, or -Wno-error=coverage-mismatch can be used to disable the error. Disabling the error for this warning can result in poorly optimized code and is useful only in the case of very minor changes, like bug fixes to an existing code-base. Completely disabling the warning is not recommended. 🙈
• -Wno-coverage-too-many-conditions — warns if -fcondition-coverage is used and an expression has too many terms and GCC gives up on coverage. Coverage gets given up when there are more terms in the conditional than there are bits in a gcov_type_unsigned. This warning is enabled by default.
• -Wno-coverage-too-many-paths — warns if -fpath-coverage is used and a function has too many paths, so GCC gives up on coverage. When it gives up is controlled by -fpath-coverage-limit. This warning is enabled by default.
• -Wno-coverage-invalid-line-number — warns in case a function ends earlier than it begins, because of an invalid linenum macro. The warning is only emitted when --coverage is enabled.

By default, this warning is enabled and is treated as an error. -Wno-coverage-invalid-line-number can be used to disable the warning, or -Wno-error=coverage-invalid-line-number can be used to disable the error.

• -Wno-cpp (C, Objective-C, C++, Objective-C++ and Fortran only) — suppresses warning messages emitted by #warning directives. Silencing the haters fr fr. 🤫
• -Wdouble-promotion (C, C++, Objective-C and Objective-C++ only) — gives you a warning when a value of type float gets implicitly promoted to double. CPUs with a 32-bit "single-precision" floating-point unit do float in hardware, but emulate double in software. On a machine like that, doing computations with double values is wayyyyy more expensive because of the overhead needed for software emulation. 💸

And ngl it's super easy to accidentally do computations with double, because floating-point literals are implicitly of type double. For example, in:

float area(float radius)
{
   return 3.14159 * radius * radius;
}

the compiler does the entire computation with double, because the floating-point literal is a double.

• -Wduplicate-decl-specifier (C and Objective-C only) — warns if a declaration has a duplicate const, volatile, restrict or _Atomic specifier. This warning is enabled by -Wall.
• -Wformat
• -Wformat=n — checks calls to printf and scanf, etc., to make sure the arguments you supplied have types appropriate to the format string you specified, and that the conversions specified in the format string actually make sense. This includes standard functions, plus others specified by format attributes, in the printf, scanf, strftime and strfmon (an X/Open extension, not in the C standard) families — or other target-specific families. Which functions get checked without format attributes having been specified depends on the standard version you selected, and those checks of functions without the attribute specified get disabled by -ffreestanding or -fno-builtin.

The formats are checked against the format features supported by GNU libc version 2.2. These include all ISO C90 and C99 features, plus features from the Single Unix Specification and some BSD and GNU extensions. Other library implementations may not support all these features; GCC does not support warning about features that go beyond a particular library's limitations. However, if -Wpedantic is used with -Wformat, warnings get given about format features not in the selected standard version (but not for strfmon formats, since those aren't in any version of the C standard).

• -Wformat=1
• -Wformat — Option -Wformat is equivalent to -Wformat=1, and -Wno-format is equivalent to -Wformat=0. Since -Wformat also checks for null format arguments for several functions, -Wformat also implies -Wnonnull. Some aspects of this level of format checking can be disabled by these options: -Wno-format-contains-nul, -Wno-format-extra-args, and -Wno-format-zero-length. -Wformat is enabled by -Wall. ✨
• -Wformat=2 — enables -Wformat plus additional format checks. Currently equivalent to -Wformat -Wformat-nonliteral -Wformat-security -Wformat-y2k.

• -Wno-format-contains-nul — if -Wformat is specified, do not warn about format strings that contain NUL bytes.
• -Wno-format-extra-args — if -Wformat is specified, do not warn about excess arguments to a printf or scanf format function. The C standard specifies that such arguments are ignored.

Where the unused arguments lie between used arguments that are specified with $ operand number specifications, normally warnings are still given, since the implementation could not know what type to pass to va_arg to skip the unused arguments. However, in the case of scanf formats, this option suppresses the warning if the unused arguments are all pointers, since the Single Unix Specification says that such unused arguments are allowed.

• -Wformat-overflow
• -Wformat-overflow=level — warns about calls to formatted input/output functions such as sprintf and vsprintf that might overflow the destination buffer. When the exact number of bytes written by a format directive can't be determined at compile-time, it gets estimated based on heuristics that depend on the level argument and on optimization. While enabling optimization will in most cases improve the accuracy of the warning, it may also result in false positives. 🙃
  • -Wformat-overflow
  • -Wformat-overflow=1 — Level 1 of -Wformat-overflow, enabled by -Wformat, employs a conservative approach that warns only about calls that most likely overflow the buffer. At this level, numeric arguments to format directives with unknown values are assumed to have the value of one, and strings of unknown length are assumed to be empty. Numeric arguments that are known to be bounded to a subrange of their type, or string arguments whose output is bounded either by their directive's precision or by a finite set of string literals, are assumed to take on the value within the range that results in the most bytes on output. For example, the call to sprintf below is diagnosed because even with both a and b equal to zero, the terminating NUL character ('\0') appended by the function to the destination buffer will be written past its end. Increasing the size of the buffer by a single byte is sufficient to avoid the warning, though it may not be sufficient to avoid the overflow.

---

                       void f (int a, int b)
                       {
                         char buf [13];
                         sprintf (buf, "a = %i, b = %i\n", a, b);
                       }

• -Wformat-overflow=2 — bestie, level 2 is the thorough one 💅. It warns about calls that might overflow the destination buffer when you hand it an argument of sufficient length or magnitude. At level 2, unknown numeric arguments are assumed to have the minimum representable value for signed types with a precision greater than 1, and the maximum representable value otherwise. And unknown string arguments whose length can't be assumed bounded — not by the directive's precision, not by a finite set of string literals they may evaluate to, not by the character array they may point to — are assumed to be 1 character long. 👑

At level 2 the call in the example above is again diagnosed, but this time because with a equal to a 32-bit INT_MIN, the first %i directive will write some of its digits beyond the end of the destination buffer. To make the call safe regardless of the values of the two variables, the size of the destination buffer must be increased to at least 34 bytes. GCC includes the minimum size of the buffer in an informational note following the warning — so helpful of it 🫶.

An alternative to increasing the size of the destination buffer is to constrain the range of formatted values. The maximum length of string arguments can be bounded by specifying the precision in the format directive. And when numeric arguments of format directives can be assumed to be bounded by less than the precision of their type, choosing an appropriate length modifier to the format specifier will reduce the required buffer size. For example, if a and b in the example above can be assumed to be within the precision of the short int type, then using either the %hi format directive or casting the argument to short reduces the maximum required size of the buffer to 24 bytes. math ate, no crumbs ✨

                       void f (int a, int b)
                       {
                         char buf [23];
                         sprintf (buf, "a = %hi, b = %i\n", a, (short)b);
                       }

• -Wno-format-zero-length — if -Wformat is specified, do not warn about zero-length formats. The C standard literally specifies that zero-length formats are allowed, so bestie, sometimes nothing is a whole vibe 🌸.
• -Wformat-nonliteral — if -Wformat is specified, also warn if the format string is not a string literal and so cannot be checked — unless the format function takes its format arguments as a va_list. iykyk.
• -Wformat-security — if -Wformat is specified, also warn about uses of format functions that represent possible security problems. At present, this warns about calls to printf and scanf functions where the format string is not a string literal and there are no format arguments, as in printf (foo);. This may be a security hole if the format string came from untrusted input and contains %n. 🚨 (This is currently a subset of what -Wformat-nonliteral warns about, but in future warnings may be added to -Wformat-security that are not included in -Wformat-nonliteral.) So like, it may grow. Stay tuned 💀
• -Wformat-signedness — if -Wformat is specified, also warn if the format string requires an unsigned argument and the argument is signed and vice versa. mismatched energy, we don't love it 😭
• -Wformat-truncation
• -Wformat-truncation=level — warn about calls to formatted input/output functions such as snprintf and vsnprintf that might result in output truncation. When the exact number of bytes written by a format directive cannot be determined at compile-time, it is estimated based on heuristics that depend on the level argument and on optimization. While enabling optimization will in most cases improve the accuracy of the warning, it may also result in false positives. Except as noted otherwise, the option uses the same logic as -Wformat-overflow. 🦋
  • -Wformat-truncation
  • -Wformat-truncation=1 — level 1 of -Wformat-truncation, enabled by -Wformat, employs a conservative approach that warns only about calls to bounded functions whose return value is unused and that will most likely result in output truncation. lowkey chill about it.
  • -Wformat-truncation=2 — level 2 warns also about calls to bounded functions whose return value is used and that might result in truncation given an argument of sufficient length or magnitude. highkey thorough, we love a level 2 🔥
• -Wformat-y2k — if -Wformat is specified, also warn about strftime formats that may yield only a two-digit year. it's giving y2k, and not in the cute way 🎀
• -Wnonnull — warn about passing a null pointer for arguments marked as requiring a non-null value by the nonnull function attribute. 🚫

-Wnonnull is included in -Wall and -Wformat. It can be disabled with the -Wno-nonnull option.

• -Wnonnull-compare — warn when comparing an argument marked with the nonnull function attribute against null inside the function. bestie the compiler said it cannot be null, why are we checking 👀

-Wnonnull-compare is included in -Wall. It can be disabled with the -Wno-nonnull-compare option.

• -Wnull-dereference — warn if the compiler detects paths that trigger erroneous or undefined behavior due to dereferencing a null pointer. This option is only active when -fdelete-null-pointer-checks is active, which is enabled by optimizations in most targets. The precision of the warnings depends on the optimization options used. 💀
• -Wno-musttail-local-addr — do not warn about passing a pointer (or in C++, a reference) to a local variable or label to an argument of a musttail call. Those variables go out of scope before the tail call instruction. gone before the party even starts 🫶
• -Wmaybe-musttail-local-addr — warn when the address of a local variable can escape to a musttail call, unless it goes out of scope already before the musttail call.

                   int foo (int *);

                   int
                   bar (int *x)
                   {
                     if (x[0] == 1)
                       {
                         int a = 42;
                         foo (&a);
                         /* Without the C<musttail> attribute this call would not
                            be tail called, because address of the C<a> variable escapes
                            and the second foo call could dereference it.  With the attribute
                            the local variables are assumed to go out of scope immediately
                            before the tail call instruction and the compiler warns about
                            this.  */
                         [[gnu::musttail]] return foo (nullptr);
                       }
                     else
                       {
                         {
                           int a = 42;
                           foo (&a);
                         }
                         /* The C<a> variable isn't already in scope, so even when it
                            escaped, even without C<musttail> attribute it would be
                            undefined behavior to dereference it and the compiler could
                            turn this into a tail call.  No warning is diagnosed here.  */
                         [[gnu::musttail]] return foo (nullptr);
                       }
                   }

This warning is enabled by -Wextra. ✨

• -Wnrvo (C++ and Objective-C++ only) — warn if the compiler does not elide the copy from a local variable to the return value of a function in a context where it is allowed by [class.copy.elision]. This elision is commonly known as the Named Return Value Optimization. For instance, in the example below the compiler cannot elide copies from both v1 and v2, so it elides neither. 😭

                   std::vector<int> f()
                   {
                     std::vector<int> v1, v2;
                     // ...
                     if (cond) return v1;
                     else return v2; // warning: not eliding copy
                   }

• -Winfinite-recursion — warn about infinitely recursive calls. The warning is effective at all optimization levels but requires optimization in order to detect infinite recursion in calls between two or more functions. -Winfinite-recursion is included in -Wall. 🔁

Compare with -Wanalyzer-infinite-recursion which provides a similar diagnostic, but is implemented in a different way (as part of -fanalyzer). two icons, same vibe, different methods 💅

• -Winit-self (C, C++, Objective-C and Objective-C++ only) — warn about uninitialized variables that are initialized with themselves. Note this option can only be used with the -Wuninitialized option. like, girl, int i = i; is not the self-love we meant 🙈

For example, GCC warns about i being uninitialized in the following snippet only when -Winit-self has been specified:

                   int f()
                   {
                     int i = i;
                     return i;
                   }

This warning is enabled by -Wall in C++.

• -Wno-implicit-int (C and Objective-C only) — this option controls warnings when a declaration does not specify a type. This warning is enabled by default, as an error, in C99 and later dialects of C, and also by -Wall. The error can be downgraded to a warning using -fpermissive (along with certain other errors), or for this error alone, with -Wno-error=implicit-int. ✨

This warning is upgraded to an error by -pedantic-errors.

• -Wno-implicit-function-declaration (C and Objective-C only) — this option controls warnings when a function is used before being declared. This warning is enabled by default, as an error, in C99 and later dialects of C, and also by -Wall. The error can be downgraded to a warning using -fpermissive (along with certain other errors), or for this error alone, with -Wno-error=implicit-function-declaration. 🦋

This warning is upgraded to an error by -pedantic-errors.

• -Wimplicit (C and Objective-C only) — same as -Wimplicit-int and -Wimplicit-function-declaration. This warning is enabled by -Wall. two-for-one, we stan a bundle 🎀
• -Whardened — warn when -fhardened did not enable an option from its set (for which see -fhardened). For instance, using -fhardened and -fstack-protector at the same time on the command line causes -Whardened to warn because -fstack-protector-strong will not be enabled by -fhardened. ngl, GCC is being very "I told you so" about it.

This warning is enabled by default and has effect only when -fhardened is enabled.

• -Wimplicit-fallthrough — -Wimplicit-fallthrough is the same as -Wimplicit-fallthrough=3 and -Wno-implicit-fallthrough is the same as -Wimplicit-fallthrough=0.
• -Wimplicit-fallthrough=n — warn when a switch case falls through. For example:

switch (cond)
  {
  case 1:
    a = 1;
    break;
  case 2:
    a = 2;
  case 3:
    a = 3;
    break;
  }

Okay bestie so here's the tea 🍵 — this warning is NOT gonna yell at you when the last statement of a case literally can't fall through. Like, if there's a return chillin' there, or a call to a function declared with the noreturn attribute? It's giving "we're done here," and GCC knows 💅.

Also -Wimplicit-fallthrough= takes control flow statements into account too — stuff like ifs — and it only warns when appropriate. Like so:

switch (cond)
  {
  case 1:
    if (i > 3) {
      bar (5);
      break;
    } else if (i < 1) {
      bar (0);
    } else
      return;
  default:
    ...
  }

Now, since sometimes a switch case fall through is genuinely the vibe you're going for 🦋, GCC hands you an attribute — __attribute__ ((fallthrough)) — and you use it right alongside a null statement to suppress the warning that would otherwise show up:

switch (cond)
  {
  case 1:
    bar (0);
    __attribute__ ((fallthrough));
  default:
    ...
  }

C++17 and C23 have a standard way to do this too — suppress the -Wimplicit-fallthrough warning with [[fallthrough]]; instead of the GNU attribute ✨. And if you're on C++11 or C++14, [[gnu::fallthrough]]; works, which is a GNU extension, iykyk 🧚.

But plot twist — instead of these attributes, you can also just drop a fallthrough comment to silence the warning. The whole body of the C or C++ style comment needs to match the regular expressions listed below. The option argument n tells you which kind of comments get accepted:

• -Wimplicit-fallthrough=0 — disables the warning altogether. GONE. nothing. 💀
• -Wimplicit-fallthrough=1 — matches the ".*" regular expression, meaning any comment counts as a fallthrough comment. chaos mode, we love her.
• -Wimplicit-fallthrough=2 — case insensitively matches the ".*falls?[ \t-]*thr(ough|u).*" regular expression.
• -Wimplicit-fallthrough=3 — case sensitively matches one of the following regular expressions:
  • "-fallthrough"
  • "@fallthrough@"
  • "lint -fallthrough[ \t]*"
  • "[ \t.!]*(ELSE,? |INTENTIONAL(LY)? )?FALL(S | |-)?THR(OUGH|U)[\t.!]*(-[^\n\r]*)?"
  • "[ \t.!]*(Else,? |Intentional(ly)? )?Fall((s | |-)[Tt]|t)hr(ough|u)[ \t.!]*(-[^\n\r]*)?"
  • "[ \t.!]*([Ee]lse,? |[Ii]ntentional(ly)? )?fall(s | |-)?thr(ough|u)[ \t.!]*(-[^\n\r]*)?"
• -Wimplicit-fallthrough=4 — case sensitively matches one of the following regular expressions:
  • "-fallthrough"
  • "@fallthrough@"
  • "lint -fallthrough[ \t]*"
  • "[ \t]*FALLTHR(OUGH|U)[ \t]*"
• -Wimplicit-fallthrough=5 — straight up doesn't recognize any comments as fallthrough comments. Only attributes can disable the warning here. periodt. 👑

The comment needs to be followed — after optional whitespace and other comments — by case or default keywords, or by a user label that precedes some case or default label. Like:

switch (cond)
  {
  case 1:
    bar (0);
    /* FALLTHRU */
  default:
    ...
  }

And heads up: the -Wimplicit-fallthrough=3 warning gets enabled by -Wextra 🌸.

---

• -Wno-if-not-aligned (C, C++, Objective-C and Objective-C++ only) — this one controls whether warnings triggered by the warn_if_not_aligned attribute should be issued at all. And ngl, these warnings are enabled by default 🫶.
• -Wignored-qualifiers (C and C++ only) — warns you if the return type of a function has a type qualifier like const. For ISO C that kind of type qualifier literally does nothing, because the value a function returns isn't an lvalue — it's giving decorative tbh. For C++ the warning only fires for scalar types or void. And ISO C actually prohibits qualified void return types on function definitions, so those return types always get a warning even without this option. no cap.

This warning is also enabled by -Wextra ✨.

• -Wno-ignored-attributes (C and C++ only) — controls warnings for when an attribute gets ignored. This is different from the -Wattributes option because it warns whenever the compiler decides to drop an attribute, not because the attribute is unknown or used in the wrong place or whatever. This warning is enabled by default 💖.
• -Wmain — warns if the type of main is suspicious. 👀 main should be a function with external linkage, returning int, taking either zero arguments, two, or three arguments of appropriate types. This warning is enabled by default in C++, and it gets enabled by either -Wall or -Wpedantic.

Also: this warning gets upgraded to an error by -pedantic-errors. so if it fires you're in trouble fr fr 😭.

• -Wmisleading-indentation (C and C++ only) — warns when your indentation does not reflect the block structure. Specifically, a warning gets issued for if, else, while, and for clauses that have a guarded statement without braces, followed by an unguarded statement at the same indentation. it's giving "the code is lying to you."

In the following example, that call to bar is misleadingly indented as if the if conditional were guarding it:

if (some_condition ())
  foo ();
  bar ();  /* Gotcha: this is not guarded by the "if".  */

When tabs and spaces are mixed together, the warning uses the -ftabstop= option to figure out whether the statements line up — and that defaults to 8.

The warning is not issued for code involving multiline preprocessor logic, like this example here:

if (flagA)
  foo (0);
#if SOME_CONDITION_THAT_DOES_NOT_HOLD
  if (flagB)
#endif
  foo (1);

Also it's not issued after a #line directive, because that typically means the code is autogenerated, and we can't make any assumptions about the layout of the file that directive points to. 🧚

This warning is enabled by -Wall in C and C++ 💅.

• -Wmissing-attributes — warns when a declaration of a function is missing one or more attributes that a related function was declared with, and where that absence could adversely affect the correctness or efficiency of the generated code. For example, the warning fires for declarations of aliases that use attributes to specify less restrictive requirements than those of their targets. This is typically a potential optimization opportunity. By contrast, the -Wattribute-alias=2 option controls warnings issued when the alias is more restrictive than the target — which could lead to incorrect code generation, so that one's giving actual danger 😳.

The attributes considered here are alloc_align, alloc_size, cold, const, hot, leaf, malloc, nonnull, noreturn, nothrow, pure, returns_nonnull, and returns_twice.

In C++, the warning is issued when an explicit specialization of a primary template declared with attribute alloc_align, alloc_size, assume_aligned, format, format_arg, malloc, or nonnull is declared without it. The attributes deprecated, error, and warning suppress the warning.

You can use the copy attribute to apply the same set of attributes to a declaration as another declaration has — without explicitly listing out all the attributes yourself. This attribute can be applied to declarations of functions, variables, or types. iconic behavior honestly 💖

-Wmissing-attributes is enabled by -Wall.

For example, since the declaration of the primary function template below makes use of both the malloc attribute and the alloc_size attribute, the declaration of the explicit specialization of the template gets diagnosed because it's missing one of the attributes:

template <class T>
T* __attribute__ ((malloc, alloc_size (1)))
allocate (size_t);

template <>
void* __attribute__ ((malloc))   // missing alloc_size
allocate<void> (size_t);

• -Wmissing-braces — warns if an aggregate or union initializer isn't fully bracketed. In the example below, the initializer for a is not fully bracketed, but the one for b is fully bracketed ✨:

int a[2][2] = { 0, 1, 2, 3 };
int b[2][2] = { { 0, 1 }, { 2, 3 } };

This warning is enabled by -Wall.

• -Wmissing-include-dirs (C, C++, Objective-C, Objective-C++ and Fortran only) — warns if a user-supplied include directory doesn't exist. This option is disabled by default for C, C++, Objective-C and Objective-C++. For Fortran, it's partially enabled by default — it only warns for -I and -J. 🙈
• -Wno-missing-profile — this option controls warnings about missing feedback profiles when you're using the -fprofile-use option. It diagnoses those cases where a new function or a new file got added between compiling with -fprofile-generate and with -fprofile-use, without regenerating the profiles. In those cases the profile feedback data files don't contain any profile feedback info for the newly added function or file respectively. Also, if profile count data (.gcda) files get removed, GCC can't use any profile feedback information at all. In all these cases, warnings get issued to let you know a profile generation step is due 📣.

Ignoring the warning can result in poorly optimized code. -Wno-missing-profile can be used to disable the warning, but this is not recommended and should only be done when non-existent profile data is justified. bestie please just regenerate 😭

• -Wmismatched-dealloc — warns for calls to deallocation functions with pointer arguments that came back from allocation functions for which the former isn't a suitable deallocator. A pair of functions can be associated as matching allocators and deallocators through the malloc attribute. Unless it's disabled by the -fno-builtin option, the standard functions calloc, malloc, realloc, and free — plus the corresponding forms of C++ operator new and operator delete — are implicitly associated as matching allocators and deallocators 🫶.

In the example below, mydealloc is the deallocator for pointers returned from myalloc:

void mydealloc (void*);

---

__attribute__ ((malloc (mydealloc, 1))) void*
myalloc (size_t);

void f (void)
{
  void *p = myalloc (32);
  // ...use p...
  free (p);   // warning: not a matching deallocator for myalloc
  mydealloc (p);   // ok
}

So in the C++ world, the sibling option -Wmismatched-new-delete is the one that clocks mismatches involving either operator new or operator delete. 💅

And okay bestie, important: -Wmismatched-dealloc is included in -Wall. periodt.

---

-Wmultistatement-macros

This one yells at you about unsafe multiple statement macros that look like they're guarded by a clause like if, else, for, switch, or while — but plot twist, only the first statement actually gets guarded once the macro is expanded. 🙈

Like, observe:

#define DOIT x++; y++
if (c)
  DOIT;

That increments y unconditionally, not just when c holds. not the vibe. 😭

Usually you fix it by wrapping the macro in a do-while loop:

#define DOIT do { x++; y++; } while (0)
if (c)
  DOIT;

This warning is enabled by -Wall in C and C++. ✨

---

-Wparentheses

Warns if parentheses get omitted in certain contexts — you know, like when there's an assignment chilling in a spot where a truth value is expected, or when operators are nested in ways that people's precedence comprehension simply cannot handle. 💀

Also warns if a comparison like x<=y<=z shows up; that's literally equivalent to (x<=y ? 1 : 0) <= z, which is a completely different interpretation than ordinary mathematical notation. math girlies stay losing.

Also warns for dangerous uses of the GNU extension to ?: with the middle operand omitted. When the condition in the ? : operator is a boolean expression, the omitted value is always 1. Often programmers expect it to be a value computed inside the conditional expression instead. 🫠

For C++ it also warns about some cases of unnecessary parentheses in declarations, which can hint at an attempt at a function call instead of a declaration:

{
  // Declares a local variable called mymutex.
  std::unique_lock<std::mutex> (mymutex);
  // User meant std::unique_lock<std::mutex> lock (mymutex);
}

This warning is enabled by -Wall.

---

-Wno-self-move (C++ and Objective-C++ only)

This warning calls out when a value gets moved to itself with std::move. A std::move like that typically has no effect whatsoever. it's giving "girl what are you doing". 🙈

struct T {
...
};
void fn()
{
  T t;
  ...
  t = std::move (t);
}

This warning is enabled by -Wall.

---

-Wsequence-point

Warns about code that may have undefined semantics because of sequence point rule violations in the C and C++ standards. it's lowkey a crime scene. 🚨

The C and C++ standards define the order in which expressions in a C/C++ program get evaluated in terms of sequence points, which represent a partial ordering between the execution of parts of the program: those executed before the sequence point, and those executed after it. These occur:

• after the evaluation of a full expression (one which is not part of a larger expression),
• after the evaluation of the first operand of a &&, ||, ? : or , (comma) operator,
• before a function is called (but after the evaluation of its arguments and the expression denoting the called function),
• and in certain other places.

Other than as expressed by the sequence point rules, the order of evaluation of subexpressions of an expression is not specified. All these rules describe only a partial order rather than a total order — because, for example, if two functions are called within one expression with no sequence point between them, the order in which the functions are called is not specified. However, the standards committee have ruled that function calls do not overlap.

It is not specified when between sequence points modifications to the values of objects take effect. Programs whose behavior depends on this have undefined behavior; the C and C++ standards specify that "Between the previous and next sequence point an object shall have its stored value modified at most once by the evaluation of an expression. Furthermore, the prior value shall be read only to determine the value to be stored.". If a program breaks these rules, the results on any particular implementation are entirely unpredictable. 🙈

Examples of code with undefined behavior are a = a++;, a[n] = b[n++] and a[i++] = i;. Some more complicated cases are not diagnosed by this option, and it may give an occasional false positive result, but in general it has been found fairly effective at detecting this sort of problem in programs.

The C++17 standard will define the order of evaluation of operands in more cases: in particular it requires that the right-hand side of an assignment be evaluated before the left-hand side, so the above examples are no longer undefined. But this option will still warn about them, to help people avoid writing code that is undefined in C and earlier revisions of C++. ✨

The standard is worded confusingly, therefore there is some debate over the precise meaning of the sequence point rules in subtle cases. Links to discussions of the problem, including proposed formal definitions, may be found on the GCC readings page, at <https://gcc.gnu.org/readings.html>.

This warning is enabled by -Wall for C and C++.

---

-Wno-return-local-addr

Do not warn about returning a pointer (or in C++, a reference) to a variable that goes out of scope after the function returns. 😬 choosing chaos, honestly.

---

-Wreturn-mismatch

Warns about return statements without an expression in functions which do not return void. Also warns about a return statement with an expression in a function whose return type is void — unless the expression type is also void. As a GNU extension, the latter case is accepted without a warning unless -Wpedantic is used. 🫶

Attempting to use the return value of a non-void function other than main that flows off the end by reaching the closing curly brace that terminates the function is undefined.

This warning is specific to C and enabled by default. In C99 and later language dialects, it is treated as an error. It can be downgraded to a warning using -fpermissive (along with other warnings), or for just this warning, with -Wno-error=return-mismatch.

---

-Wreturn-type

Warns whenever a function is defined with a return type that defaults to int (unless -Wimplicit-int is active, which takes precedence). Also warns if execution may reach the end of the function body, or if the function does not contain any return statement at all. girl, where is your return. 😭

Attempting to use the return value of a non-void function other than main that flows off the end by reaching the closing curly brace that terminates the function is undefined.

Unlike in C, in C++, flowing off the end of a non-void function other than main results in undefined behavior even when the value of the function is not used. 😳

This warning is enabled by default in C++ and by -Wall otherwise.

---

-Wno-shift-count-negative

Controls warnings if a shift count is negative. This warning is enabled by default.

---

-Wno-shift-count-overflow

Controls warnings if a shift count is greater than or equal to the bit width of the type. This warning is enabled by default.

---

-Wshift-negative-value

Warns if left shifting a negative value. This warning is enabled by -Wextra in C99 (and newer) and C++11 to C++17 modes.

---

-Wno-shift-overflow, -Wshift-overflow=n

These options control warnings about left shift overflows. 💫

• -Wshift-overflow=1 — This is the warning level of -Wshift-overflow and is enabled by default in C99 and C++11 modes (and newer). This warning level does not warn about left-shifting 1 into the sign bit. (However, in C, such an overflow is still rejected in contexts where an integer constant expression is required.) No warning is emitted in C++20 mode (and newer), as signed left shifts always wrap.
• -Wshift-overflow=2 — This warning level also warns about left-shifting 1 into the sign bit, unless C++14 mode (or newer) is active.

---

-Wswitch

Warns whenever a switch statement has an index of enumerated type and lacks a case for one or more of the named codes of that enumeration. (The presence of a default label prevents this warning.) case labels that do not correspond to enumerators also provoke warnings when this option is used — unless the enumeration is marked with the flag_enum attribute. This warning is enabled by -Wall.

---

-Wswitch-default

Warns whenever a switch statement does not have a default case. every switch needs its default bestie. 🎀

---

-Wswitch-enum

Warns whenever a switch statement has an index of enumerated type and lacks a case for one or more of the named codes of that enumeration. case labels that do not correspond to enumerators also provoke warnings when this option is used — unless the enumeration is marked with the flag_enum attribute. The only difference between -Wswitch and this option is that this option gives a warning about an omitted enumeration code even if there is a default label. 💅

---

-Wno-switch-bool

Do not warn when a switch statement has an index of boolean type and the case values are outside the range of a boolean type. It is possible to suppress this warning by casting the controlling expression to a type other than bool. For example:

switch ((int) (a == 4))
  {
  ...
  }

This warning is enabled by default for C and C++ programs. ✨

---

-Wno-switch-outside-range

This option controls warnings when a switch case has a value that is outside of its respective type range. This warning is enabled by default for C and C++ programs.

---

-Wno-switch-unreachable

Do not warn when a switch statement contains statements between the controlling expression and the first case label, which will never be executed. For example:

switch (cond)
  {
   i = 15;
  ...
   case 5:
  ...
  }

• -Wswitch-unreachable — heads up bestie, this one does not warn if the statement chillin' between the controlling expression and the first case label is just a declaration:

switch (cond)
  {
   int i;
  ...
   case 5:
   i = 5;
  ...
  }

This warning is enabled by default for C and C++ programs. 💅

• -Wsync-nand (C and C++ only) — warns when the __sync_fetch_and_nand and __sync_nand_and_fetch built-in functions get used. Fair warning: these functions changed semantics in GCC 4.4, so like, they're giving messy-ex energy. 🙈
• -Wtrivial-auto-var-init — warns when -ftrivial-auto-var-init cannot initialize the automatic variable. A super common situation is an automatic variable declared between the controlling expression and the first case label of a switch statement. (Same chaotic spot as above, no shade.)
• -Wunused-but-set-parameter — warns whenever a function parameter gets assigned to, but is otherwise unused (aside from its declaration).

To suppress this warning, use the unused attribute. ✨

This warning is also enabled by -Wunused together with -Wextra.

• -Wunused-but-set-variable — warns whenever a local variable gets assigned to, but is otherwise unused (aside from its declaration). This warning is enabled by -Wall.

To suppress this warning, use the unused attribute.

This warning is also enabled by -Wunused, which is enabled by -Wall.

• -Wunused-function — warns whenever a static function is declared but not defined, or a non-inline static function is unused. This warning is enabled by -Wall.
• -Wunused-label — warns whenever a label is declared but not used. This warning is enabled by -Wall.

To suppress this warning, use the unused attribute.

• -Wunused-local-typedefs (C, Objective-C, C++ and Objective-C++ only) — warns when a typedef locally defined in a function is not used. This warning is enabled by -Wall.
• -Wunused-parameter — warns whenever a function parameter is unused aside from its declaration. This option is not enabled by -Wunused unless -Wextra is also specified. (The way that's just rude of it, honestly. 😭)

To suppress this warning, use the unused attribute.

• -Wno-unused-result — do not warn if a caller of a function marked with attribute warn_unused_result does not use its return value. The default is -Wunused-result.
• -Wunused-variable — warns whenever a local or static variable is unused aside from its declaration. This option implies -Wunused-const-variable=1 for C, but not for C++. This warning is enabled by -Wall.

To suppress this warning, use the unused attribute.

• -Wunused-const-variable
• -Wunused-const-variable=n — warns whenever a constant static variable is unused aside from its declaration.

To suppress this warning, use the unused attribute.

• -Wunused-const-variable=1 — warns about unused static const variables defined in the main compilation unit, but not about static const variables declared in any header included. 🌸

-Wunused-const-variable=1 is enabled by either -Wunused-variable or -Wunused for C, but not for C++. In C this declares variable storage, but in C++ this is not an error since const variables take the place of #defines.

• -Wunused-const-variable=2 — this warning level also warns for unused constant static variables in headers (excluding system headers). It is equivalent to the short form -Wunused-const-variable. This level must be explicitly requested in both C and C++ because it might be hard to clean up all headers included. (The messiest girlies stay messy, it's fine. 💀)

• -Wunused-value — warns whenever a statement computes a result that is explicitly not used. To suppress this warning, cast the unused expression to void. This includes an expression-statement or the left-hand side of a comma expression that contains no side effects. For example, an expression such as x[i,j] causes a warning, while x[(void)i,j] does not. Iconic behaviour from the compiler honestly. 💅

This warning is enabled by -Wall.

• -Wunused — all the above -Wunused options combined, except those documented as needing to be specified explicitly. The full bundle, no cap. 🎀

In order to get a warning about an unused function parameter, you must either specify -Wextra -Wunused (note that -Wall implies -Wunused), or separately specify -Wunused-parameter and/or -Wunused-but-set-parameter.

-Wunused enables only -Wunused-const-variable=1 rather than -Wunused-const-variable, and only for C, not C++.

• -Wuse-after-free (C, Objective-C, C++ and Objective-C++ only)
• -Wuse-after-free=n — warns about uses of pointers to dynamically allocated objects that have been rendered indeterminate by a call to a deallocation function. The warning is enabled at all optimization levels but may yield different results with optimization than without.

-Wuse-after-free is equivalent to -Wuse-after-free=1.

• -Wuse-after-free=1 — at level 1 the warning attempts to diagnose only unconditional uses of pointers made indeterminate by a deallocation call or a successful call to realloc, regardless of whether or not the call resulted in an actual reallocation of memory. This includes double-free calls as well as uses in arithmetic and relational expressions. Although undefined, uses of indeterminate pointers in equality (or inequality) expressions are not diagnosed at this level.
• -Wuse-after-free=2 — at level 2, in addition to unconditional uses, the warning also diagnoses conditional uses of pointers made indeterminate by a deallocation call. As at level 2, uses in equality (or inequality) expressions are not diagnosed. For example, the second call to free in the following function is diagnosed at this level:

struct A { int refcount; void *data; };

void release (struct A *p)
{
  int refcount = --p->refcount;
  free (p);
  if (refcount == 0)
    free (p->data);   // warning: p may be used after free
}

• -Wuse-after-free=3 — at level 3, the warning also diagnoses uses of indeterminate pointers in equality expressions. All uses of indeterminate pointers are undefined but equality tests sometimes appear after calls to realloc as an attempt to determine whether the call resulted in relocating the object to a different address. They are diagnosed at a separate level to aid gradually transitioning legacy code to safe alternatives. For example, the equality test in the function below is diagnosed at this level:

void adjust_pointers (int**, int);

void grow (int **p, int n)
{
  int **q = (int**)realloc (p, n *= 2);
  if (q == p)
    return;
  adjust_pointers ((int**)q, n);
}

To avoid the warning at this level, store offsets into allocated memory instead of pointers. This approach obviates needing to adjust the stored pointers after reallocation. (Slay, offsets stay winning. 👑)

-Wuse-after-free=2 is included in -Wall.

• -Wuseless-cast (C, Objective-C, C++ and Objective-C++ only) — warns when an expression is cast to its own type. This warning does not occur when a class object is converted to a non-reference type, as that is a way to create a temporary:

struct S { };
void g (S&&);
void f (S&& arg)
{
  g (S(arg)); // make arg prvalue so that it can bind to S&&
}

• -Wuninitialized — warns if an object with automatic or allocated storage duration is used without having been initialized. In C++, also warns if a non-static reference or non-static const member appears in a class without constructors. 👀

In addition, passing a pointer (or in C++, a reference) to an uninitialized object to a const-qualified argument of a built-in function known to read the object is also diagnosed by this warning. (-Wmaybe-uninitialized is issued for ordinary functions.)

If you want to warn about code that uses the uninitialized value of the variable in its own initializer, use the -Winit-self option.

These warnings occur for individual uninitialized elements of structure, union or array variables as well as for variables that are uninitialized as a whole. They do not occur for variables or elements declared volatile. Because these warnings depend on optimization, the exact variables or elements for which there are warnings depend on the precise optimization options and version of GCC used. (So like, it's giving ✨situationship with your optimizer✨ — the results may vary.)

Note that there may be no warning about a variable that is used only to compute a value that itself is never used, because such computations may be deleted by data flow analysis before the warnings are printed.

In C++, this warning also warns about using uninitialized objects in member-initializer-lists. For example, GCC warns about b being uninitialized in the following snippet:

struct A {
  int a;
  int b;
  A() : a(b) { }
};

• -Wno-invalid-memory-model — this option controls warnings for invocations of __atomic Builtins, __sync Builtins, and the C11 atomic generic functions with a memory consistency argument that is either invalid for the operation or outside the range of values of the memory_order enumeration. For example, since the __atomic_store and __atomic_store_n built-ins are only defined for the relaxed, release, and sequentially consistent memory orders, the following code is diagnosed:

void store (int *i)
{
  __atomic_store_n (i, 0, memory_order_consume);
}

-Winvalid-memory-model is enabled by default.

• -Wmaybe-uninitialized — for an object with automatic or allocated storage duration, if there exists a path from the function entry to a use of the object that is initialized, but there exist some other paths for which the object is not initialized, the compiler emits a warning if it cannot prove the uninitialized paths are not executed at run time. 🦋

okay so one more lil' thing about -Wuninitialized-adjacent behavior, bestie, because it's lowkey important:

because passing a pointer (or in C++, a reference) to an uninitialized object to a const-qualified function argument is also diagnosed by this warning. And to be clear: -Wuninitialized is issued for built-in functions known to read the object. If you annotate the function with attribute access (none), that tells GCC the argument isn't used to access the object, and boom — the warning is avoided. 🫶

These warnings are only possible in optimizing compilation, because otherwise GCC literally does not keep track of the state of variables. No cap.

And these warnings are made optional because GCC may not be able to determine when the code is correct in spite of appearing to have an error. Here's one example of how that can happen:

{
  int x;
  switch (y)
    {
    case 1: x = 1;
      break;
    case 2: x = 4;
      break;
    case 3: x = 5;
    }
  foo (x);
}

If the value of y is always 1, 2 or 3, then x is always initialized — but GCC doesn't know this. 😭 To suppress the warning, you need to provide a default case with assert(0) or similar code.

This option also warns when a non-volatile automatic variable might be changed by a call to longjmp. The compiler sees only the calls to setjmp. It cannot know where longjmp will be called; in fact, a signal handler could call it at any point in the code. As a result, you may get a warning even when there is in fact no problem, because longjmp cannot in fact be called at the place that would cause a problem. It's giving "false alarm" energy fr fr.

Some spurious warnings can be avoided if you declare all the functions you use that never return as noreturn. ✨

This warning is enabled by -Wall or -Wextra.

---

• -Wunknown-pragmas — Warn when a #pragma directive is encountered that is not understood by GCC. Heads up though: if this command-line option is used, warnings are even issued for unknown pragmas in system header files. That is not the case if the warnings are only enabled by the -Wall command-line option.
• -Wno-pragmas — Do not warn about misuses of pragmas, such as incorrect parameters, invalid syntax, or conflicts between pragmas. See also -Wunknown-pragmas. 🎀
• -Wno-pragma-once-outside-header — Do not warn when #pragma once is used in a file that is not a header file, such as a main file.
• -Wno-prio-ctor-dtor — Do not warn if a priority from 0 to 100 is used for constructor or destructor. The use of constructor and destructor attributes allow you to assign a priority to the constructor/destructor to control its order of execution before main is called or after it returns. The priority values must be greater than 100, as the compiler reserves priority values between 0--100 for the implementation. 👑
• -Wstrict-aliasing — This option is only active when -fstrict-aliasing is active. It warns about code that might break the strict aliasing rules that the compiler is using for optimization. The warning does not catch all cases, but does attempt to catch the more common pitfalls. It is included in -Wall. It is equivalent to -Wstrict-aliasing=3.
• -Wstrict-aliasing=n — This option is only active when -fstrict-aliasing is active. It warns about code that might break the strict aliasing rules that the compiler is using for optimization. Higher levels correspond to higher accuracy (fewer false positives). Higher levels also correspond to more effort, similar to the way -O works. -Wstrict-aliasing is equivalent to -Wstrict-aliasing=3.
  • Level 1: Most aggressive, quick, least accurate. Possibly useful when higher levels do not warn but -fstrict-aliasing still breaks the code, as it has very few false negatives. However, it has many false positives. Warns for all pointer conversions between possibly incompatible types, even if never dereferenced. Runs in the front end only.
  • Level 2: Aggressive, quick, not too precise. May still have many false positives (not as many as level 1 though), and few false negatives (but possibly more than level 1). Unlike level 1, it only warns when an address is taken. Warns about incomplete types. Runs in the front end only.
  • Level 3 (default for -Wstrict-aliasing): Should have very few false positives and few false negatives. Slightly slower than levels 1 or 2 when optimization is enabled. Takes care of the common pun+dereference pattern in the front end: *(int*)&some_float. If optimization is enabled, it also runs in the back end, where it deals with multiple statement cases using flow-sensitive points-to information. Only warns when the converted pointer is dereferenced. Does not warn about incomplete types.
• -Wstrict-overflow
• -Wstrict-overflow=n

This option is only active when signed overflow is undefined. It warns about cases where the compiler optimizes based on the assumption that signed overflow does not occur. Note that it does not warn about all cases where the code might overflow: it only warns about cases where the compiler implements some optimization. Thus this warning depends on the optimization level.

An optimization that assumes that signed overflow does not occur is perfectly safe if the values of the variables involved are such that overflow never does, in fact, occur. Therefore this warning can easily give a false positive: a warning about code that is not actually a problem. To help focus on important issues, several warning levels are defined. No warnings are issued for the use of undefined signed overflow when estimating how many iterations a loop requires, in particular when determining whether a loop will be executed at all. 🧚

• -Wstrict-overflow=1 — Warn about cases that are both questionable and easy to avoid. For example the compiler simplifies x + 1 > x to 1. This level of -Wstrict-overflow is enabled by -Wall; higher levels are not, and must be explicitly requested.
• -Wstrict-overflow=2 — Also warn about other cases where a comparison is simplified to a constant. For example: abs (x) >= 0. This can only be simplified when signed integer overflow is undefined, because abs (INT_MIN) overflows to INT_MIN, which is less than zero. -Wstrict-overflow (with no level) is the same as -Wstrict-overflow=2.
• -Wstrict-overflow=3 — Also warn about other cases where a comparison is simplified. For example: x + 1 > 1 is simplified to x > 0.
• -Wstrict-overflow=4 — Also warn about other simplifications not covered by the above cases. For example: (x * 10) / 5 is simplified to x * 2.
• -Wstrict-overflow=5 — Also warn about cases where the compiler reduces the magnitude of a constant involved in a comparison. For example: x + 2 > y is simplified to x + 1 >= y. This is reported only at the highest warning level because this simplification applies to many comparisons, so this warning level gives a very large number of false positives. 💀

• -Wstring-compare — Warn for calls to strcmp and strncmp whose result is determined to be either zero or non-zero in tests for such equality owing to the length of one argument being greater than the size of the array the other argument is stored in (or the bound in the case of strncmp). Such calls could be mistakes. For example, the call to strcmp below is diagnosed because its result is necessarily non-zero irrespective of the contents of the array a.

extern char a[4];
void f (char *d)
{
  strcpy (d, "string");
  ...
  if (0 == strcmp (a, d))   // cannot be true
    puts ("a and d are the same");
}

-Wstring-compare is enabled by -Wextra. ✨

• -Wno-stringop-overflow
• -Wstringop-overflow
• -Wstringop-overflow=type

Warn for code that can be statically determined to cause buffer overflows or memory overruns, such as calls to memcpy and strcpy that overflow the destination buffer. The optional argument is one greater than the type of Object Size Checking to perform to determine the size of the destination. The argument is meaningful only for string functions that operate on character arrays; raw memory functions like memcpy always use type-zero Object Size Checking.

The option also warns for calls that specify a size in excess of the largest possible object or at most SIZE_MAX / 2 bytes.

The option produces the best results with optimization enabled but can detect a small subset of simple buffer overflows even without optimization in calls to the GCC built-in functions like __builtin_memcpy that correspond to the standard functions. In any case, the option warns about just a subset of buffer overflows detected by the corresponding overflow checking built-ins, such as __builtin___memcpy_chk, which can perform run-time checking if the access cannot be identified as safe at compile time. 👑

For example, the option issues a warning for the strcpy call below because it copies at least 5 characters (the string "blue" including the terminating NUL) into the buffer of size 4.

enum Color { blue, purple, yellow };
const char* f (enum Color clr)
{
  static char buf [4];
  const char *str;
  switch (clr)
    {
      case blue: str = "blue"; break;
      case purple: str = "purple"; break;
      case yellow: str = "yellow"; break;
    }

  return strcpy (buf, str);   // warning here
}

The effect of this option is not limited to string or memory manipulation functions. In this example, a warning is diagnosed because a 1-element array is passed to a function requiring at least a 4-element array argument:

void f (int[static 4]);

void g (void)
{
  int *p = (int *) malloc (1 * sizeof(int));
  f (p);   // warning here
}

Option -Wstringop-overflow=2 is enabled by default. 🦋

• -Wstringop-overflow — the big one about writing past the end of stuff. Got four whole levels, bestie, and yes they all hit different 💅
  • -Wstringop-overflow=1 — this one uses type-zero Object Size Checking to figure out how big the destination objects are. At this setting it will not yell at you for writes past the end of subobjects of bigger objects that you reached through pointers, unless it knows how big the largest surrounding object is. And when the destination could be one of several objects? It just assumes it's the biggest one of them. On Linux, when optimization is on, this setting warns for the exact same code as when _FORTIFY_SOURCE is defined to a non-zero value. cute little twin energy ✨
  • -Wstringop-overflow=2 — type-one Object Size Checking for the destination sizes. At this setting it warns about overflows when you write to members of the largest complete objects whose exact size is known. But it does not warn about excessive writes to those same members on unknown objects reached through pointers, because those pointers might be pointing at arrays with an unknown number of elements. This is the default setting of the option, so it's what you get if you're just vibing 👑
  • -Wstringop-overflow=3 — type-two Object Size Checking. This setting warns when you overflow the smallest object or data member. It's the most restrictive setting of the option, which means yeah… it may throw warnings at totally safe code 😭 dramatic queen behavior
  • -Wstringop-overflow=4 — type-three Object Size Checking. Warns about overflowing any data members, and when the destination is one of several objects it uses the size of the largest one to decide whether to warn. Just like -Wstringop-overflow=3, this setting may warn about benign code too. lowkey overprotective but we love her 🙈
• -Wno-stringop-overread — warns for calls to string manipulation functions like memchr or strcpy that are determined to read past the end of the source sequence. Option -Wstringop-overread is enabled by default, so she's always watching 🧚
• -Wno-stringop-truncation — no warning for calls to bounded string manipulation functions such as strncat, strncpy, and stpncpy that might either truncate the copied string or leave the destination untouched.

Say you got this one, where the call to strncat gives a bound smaller than the length of the source string — so the copy of the source gets truncated and the call gets diagnosed:

void append (char *buf, size_t bufsize)
{
  strncat (buf, ".txt", 3);
}

To dodge the warning, use bufsize - strlen (buf) - 1) as the bound. (yes, exactly like that, we're not fixing it, it's a copy-paste situation and we're all pretending we don't see it 😭)

Another one: this call to strncpy copies into d just the characters before the terminating NUL, and never appends the NUL to the end. Assuming the result of strncpy is necessarily a NUL-terminated string is a super common mistake, so the call gets diagnosed. If you know your result isn't supposed to be NUL-terminated, call memcpy instead:

void copy (char *d, const char *s)
{
  strncpy (d, s, strlen (s));
}

And here, the call to strncpy uses the size of the destination buffer as the bound. If the source string's length is equal to or bigger than that size, the result of the copy won't be NUL-terminated. So yeah, diagnosed. To avoid it, pass sizeof buf - 1 as the bound and set the last element of the buffer to NUL:

void copy (const char *s)
{
  char buf[80];
  strncpy (buf, s, sizeof buf);
  ...
}

Now listen — if a character array is meant to hold a sequence of bytes with no terminating NUL, you can annotate it with attribute nonstring to skip this warning. But heads up bestie, those arrays are not suitable arguments for functions that expect NUL-terminated strings. To help catch accidental misuse of them, GCC issues warnings unless it can prove the use is safe. so like, prove it or get cooked 🫶

• -Wstrict-flex-arrays (C and C++ only) — warns about improper usages of flexible array members, according to the level of the strict_flex_array (level) attribute attached to the trailing array field of a structure if it's available, and otherwise according to the level of the option -fstrict-flex-arrays=level. -Wstrict-flex-arrays only does anything when level is greater than 0, no cap.
  • level=1 → warnings get issued for a trailing array reference of a structure that has 2 or more elements, if the trailing array is referenced as a flexible array member.
  • level=2 → on top of level=1, extra warnings get issued for a trailing one-element array reference of a structure if the array is referenced as a flexible array member.
  • level=3 → on top of level=2, extra warnings get issued for a trailing zero-length array reference of a structure if the array is referenced as a flexible array member.

This option hits way harder when -ftree-vrp is active (the default for -O2 and above), but some warnings can still get diagnosed even with no optimization at all. iconic 💖

• -Wsuggest-attribute=attribute-name — warns for cases where adding an attribute might be beneficial. The attribute-names currently supported are listed below, and yes we're listing all of them.
  • -Wsuggest-attribute=pure, -Wsuggest-attribute=const, -Wsuggest-attribute=noreturn, -Wmissing-noreturn, -Wsuggest-attribute=malloc, -Wsuggest-attribute=returns_nonnull — warns about functions that might be candidates for the attributes pure, const, noreturn, malloc or returns_nonnull. The compiler only warns about functions visible in other compilation units, or (for pure and const) if it cannot prove the function returns normally. A function returns normally if it doesn't contain an infinite loop or return abnormally by throwing, calling abort, or trapping. This analysis needs option -fipa-pure-const, which is enabled by default at -O and higher. Higher optimization levels = more accurate analysis, periodt 🎀
  • -Wsuggest-attribute=format, -Wmissing-format-attribute — warns about function pointers that might be candidates for format attributes. Note: these are only possible candidates, not absolute ones. GCC guesses that function pointers with format attributes used in assignment, initialization, parameter passing, or return statements should have a matching format attribute in the resulting type. I.e. the left-hand side of the assignment or initialization, the type of the parameter variable, or the return type of the containing function respectively should also carry a format attribute to avoid the warning.

GCC also warns about function definitions that might be candidates for format attributes. Again — only possible candidates, not guarantees. GCC guesses that format attributes might be appropriate for any function that calls something like vprintf or vscanf, but that's not always the case, and some functions where format attributes are appropriate may not be detected. it's giving vibes-based ✨

• -Wsuggest-attribute=cold — warns about functions that might be candidates for the cold attribute. This one's based on static detection, and it generally only warns about functions that always lead to a call to another cold function — like wrappers of C++ throw, or fatal error reporting functions that lead to abort. 💀

• -Walloc-size — warns about calls to allocation functions decorated with attribute alloc_size that specify insufficient size for the target type of the pointer the result is assigned to, including calls to the built-in forms of aligned_alloc, alloca, calloc, malloc, and realloc. 👑
• -Walloc-zero — warns about calls to allocation functions decorated with attribute alloc_size that specify zero bytes, including calls to the built-in forms of aligned_alloc, alloca, calloc, malloc, and realloc. Why do we care? Because the behavior of these functions with a zero size differs between implementations (and in realloc's case it's been deprecated), so relying on it can cause subtle portability bugs and should be avoided. ngl this is the kind of bug that ruins your whole week 🙈
• -Wcalloc-transposed-args — warns about calls to allocation functions decorated with attribute alloc_size that take two arguments, where sizeof is used as the earlier size argument and not as the later one. This is a coding style warning. The first argument to calloc is documented as the number of elements in the array, and the second is the size of each element — so calloc (n, sizeof (int)) is preferred over calloc (sizeof (int), n). If putting sizeof in the earlier argument and not the latter is intentional, you can suppress the warning with calloc (sizeof (struct S) + 0, n) or calloc (1 * sizeof (struct S), 4), or by just using sizeof in the later argument as well. we love an explain-your-intent moment 💅
• -Walloc-size-larger-than=byte-size — warns about calls to functions decorated with attribute alloc_size that try to allocate objects larger than the specified number of bytes, or where the result of the size computation in an integer type with infinite precision would exceed the value of PTRDIFF_MAX on the target. -Walloc-size-larger-than=PTRDIFF_MAX is enabled by default. Warnings controlled by this option can be disabled either by specifying a byte-size of SIZE_MAX or more, or with -Wno-alloc-size-larger-than.
• -Wno-alloc-size-larger-than — disables -Walloc-size-larger-than= warnings. This option is equivalent to -Walloc-size-larger-than=SIZE_MAX or larger. bye bestie 🌸
• -Walloca — this option warns on all uses of alloca in the source. all of them. every single one. no mercy 🫶
• -Walloca-larger-than=byte-size — okay bestie, this one is a whole vibe check for alloca 💅 It yells at you when you call alloca with an integer argument whose value is either zero, or that isn't bounded by a controlling predicate keeping it at most byte-size. It also warns for alloca calls where the bound value is just straight-up unknown. And arguments of non-integer types? Treated as unbounded, even if they look like they're constrained to the expected range. It's giving "I don't trust your vibes". 😭

So like, a bounded case of alloca could be:

void func (size_t n)
{
  void *p;
  if (n <= 1000)
    p = alloca (n);
  else
    p = malloc (n);
  f (p);
}

In that example, passing -Walloca-larger-than=1000 would not warn, because the call to alloca is known to be at most 1000 bytes. Iconic. But if you passed -Walloca-larger-than=500? The compiler would emit a warning. Caught. 🔥

Unbounded uses — on the other hand — are uses of alloca with no controlling predicate constraining its integer argument. Like:

void func ()
{
  void *p = alloca (n);
  f (p);
}

If -Walloca-larger-than=500 were passed, that would trigger a warning — but this time because of the lack of bounds checking, not the size itself.

And ngl, even seemingly correct code involving signed integers could warn:

void func (signed int n)
{
  if (n < 500)
    {
      p = alloca (n);
      f (p);
    }
}

Because in that example, n could be negative, causing a larger than expected argument to be implicitly cast into the alloca call. Rude. 🙈

It also warns when alloca is used in a loop. (Loop alloca is not the moment, fr fr.)

-Walloca-larger-than=PTRDIFF_MAX is enabled by default — but it's usually only effective when -ftree-vrp is active (default for -O2 and above).

See also -Wvla-larger-than=byte-size.

• -Wno-alloca-larger-than — kills the -Walloca-larger-than= warnings. This option is equivalent to -Walloca-larger-than=SIZE_MAX or larger. Basically "let me live". ✨
• -Warith-conversion — do warn about implicit conversions from arithmetic operations even when conversion of the operands to the same type can't change their values. This affects warnings from -Wconversion, -Wfloat-conversion, and -Wsign-conversion.

void f (char c, int i)
{
  c = c + i; // warns with B<-Wconversion>
  c = c + 1; // only warns with B<-Warith-conversion>
}

• -Warray-bounds
• -Warray-bounds=n — warn about out-of-bounds subscripts or offsets into arrays. Enabled by -Wall. It's more effective when -ftree-vrp is active (the default for -O2 and above), but a subset of instances gets issued even without optimization. 👑

By default, the trailing array of a structure will be treated as a flexible array member by -Warray-bounds or -Warray-bounds=n if it's declared as either a flexible array member per C99 standard onwards ([]), a GCC zero-length array extension ([0]), or an one-element array ([1]). As a result, out-of-bounds subscripts or offsets into zero-length arrays or one-element arrays are not warned by default. Sneaky little loophole, no cap.

You can add the option -fstrict-flex-arrays or -fstrict-flex-arrays=level to control how this option treats the trailing array of a structure as a flexible array member:

• when level<=1, no change to the default behavior.
• when level=2, additional warnings will be issued for out-of-bounds subscripts or offsets into one-element arrays;
• when level=3, in addition to level=2, additional warnings will be issued for out-of-bounds subscripts or offsets into zero-length arrays.

hmm, levels? yeah, levels 💅

• -Warray-bounds=1 — this is the default warning level of -Warray-bounds and is enabled by -Wall; higher levels are not, and must be explicitly requested.
• -Warray-bounds=2 — this warning level also warns about the intermediate results of pointer arithmetic that may yield out-of-bounds values. Heads up though: this warning level may give a larger number of false positives and is deactivated by default.

---

• -Wunterminated-string-initialization (C and Objective-C only) — warn about character arrays initialized as unterminated character sequences with a string literal, unless the declaration being initialized has the nonstring attribute. For example:

char arr[3] = "foo"; /* Warning.  */
char arr2[3] __attribute__((nonstring)) = "bar"; /* No warning.  */

This warning is enabled by -Wextra. And if -Wc++-compat is enabled, the warning has slightly different wording and warns even if the declaration being initialized has the nonstring warning — because in C++ such initializations are an error. Periodt.

• -Warray-compare — warn about equality and relational comparisons between two operands of array type. This comparison was deprecated in C++20. Like:

int arr1[5];
int arr2[5];
bool same = arr1 == arr2;

-Warray-compare is enabled by -Wall. 🎀

• -Warray-parameter
• -Warray-parameter=n — warn about redeclarations of functions involving parameters of array or pointer types of inconsistent kinds or forms, and enable the detection of out-of-bounds accesses to such parameters by warnings such as -Warray-bounds. Iconic foresight honestly.

So if the first function declaration uses the array form for a parameter declaration, the bound specified in the array is assumed to be the minimum number of elements expected to be provided in calls to the function and the maximum number of elements accessed by it. Failing to provide arguments of sufficient size, or accessing more than the maximum number of elements, may be diagnosed by warnings such as -Warray-bounds or -Wstringop-overflow. At level 1, the warning diagnoses inconsistencies involving array parameters declared using the T[static N] form.

For example, the warning triggers for the second declaration of f because the first one with the keyword static specifies that the array argument must have at least four elements, while the second allows an array of any size to be passed to f.

void f (int[static 4]);
void f (int[]);           // warning (inconsistent array form)

void g (void)
{
  int *p = (int *)malloc (1 * sizeof (int));
  f (p);                  // warning (array too small)
  ...
}

At level 2 the warning also triggers for redeclarations involving any other inconsistency in array or pointer argument forms denoting array sizes. Pointers and arrays of unspecified bound are considered equivalent and do not trigger a warning.

void g (int*);
void g (int[]);     // no warning
void g (int[8]);    // warning (inconsistent array bound)

-Warray-parameter=2 is included in -Wall. And the -Wvla-parameter option triggers warnings for similar inconsistencies involving Variable Length Array arguments. Slay for consistency. 🦋

The short form of the option -Warray-parameter is equivalent to -Warray-parameter=2. The negative form -Wno-array-parameter is equivalent to -Warray-parameter=0.

• -Wattribute-alias=n
• -Wno-attribute-alias — warn about declarations using the alias and similar attributes whose target is incompatible with the type of the alias. It's giving "your alias doesn't match your target, bestie". 😬
  • -Wattribute-alias=1 — the default warning level of the -Wattribute-alias option diagnoses incompatibilities between the type of the alias declaration and that of its target. Such incompatibilities are typically indicative of bugs.
  • -Wattribute-alias=2 — at this level -Wattribute-alias also diagnoses cases where the attributes of the alias declaration are more restrictive than the attributes applied to its target. These mismatches can potentially result in incorrect code generation. In other cases they may be benign and could be resolved simply by adding the missing attribute to the target. For comparison, see the -Wmissing-attributes option, which controls diagnostics when the alias declaration is less restrictive than the target, rather than more restrictive.

Attributes considered include alloc_align, alloc_size, cold, const, hot, leaf, malloc, nonnull, noreturn, nothrow, pure, returns_nonnull, and returns_twice.

-Wattribute-alias is equivalent to -Wattribute-alias=1. This is the default. You can disable these warnings with either -Wno-attribute-alias or -Wattribute-alias=0. Choose your fighter. 💜

• -Wbidi-chars=[none|unpaired|any|ucn] — warn about possibly misleading UTF-8 bidirectional control characters in comments, string literals, character constants, and identifiers. Lowkey terrifying: such characters can change left-to-right writing direction into right-to-left (and vice versa), which can cause confusion between the logical order and visual order. This may be dangerous; for instance, it may seem that a piece of code is not commented out, whereas it in fact is. 💀

There are three levels of warning supported by GCC. The default is -Wbidi-chars=unpaired, which warns about improperly terminated bidi contexts. -Wbidi-chars=none turns the warning off. -Wbidi-chars=any warns about any use of bidirectional control characters.

By default, this warning does not warn about UCNs. It is, however, possible to turn on such checking by using -Wbidi-chars=unpaired,ucn or -Wbidi-chars=any,ucn. Using -Wbidi-chars=ucn is valid, and is equivalent to -Wbidi-chars=unpaired,ucn, if no previous -Wbidi-chars=any was specified.

NOTE: In Ubuntu 25.10 and later versions, -Wbidi-chars is enabled by default for C, C++, ObjC, ObjC++, unless -Wno-bidi-chars or -Wbidi-chars=none is found. 🌈

• -Wbool-compare — warn about boolean expression compared with an integer value different from true/false. For instance, the following comparison is always false:

int n = 5;
...
if ((n > 1) == 2) { ... }

This warning is enabled by -Wall. ⭐

• -Wbool-operation — bestie, this one's like a vibe check for boolean math 💅. It yells at you about suspicious operations on expressions of a boolean type. Example: bitwise negation of a boolean is very likely a bug in the program, ok? And for C, this warning also drags incrementing or decrementing a boolean, because that rarely makes sense. (In C++, decrementing a boolean is always invalid. Incrementing a boolean is invalid in C++17, and deprecated otherwise.) Also, heads up: this warning is enabled by -Wall.
• -Wduplicated-branches — warns when an if-else has identical branches 👯. Like, this warning detects cases like:

        if (p != NULL)
          return 0;
        else
          return 0;

It doesn't warn when both branches contain just a null statement. This warning also warns for conditional operators:

          int i = x ? *p : *p;

• -Wduplicated-cond — warns about duplicated conditions in an if-else-if chain 🔁. For instance, warn for the following code:

        if (p->q != NULL) { ... }
        else if (p->q != NULL) { ... }

• -Wframe-address — warns when the __builtin_frame_address or __builtin_return_address is called with an argument greater than 0 😬. Such calls may return indeterminate values or crash the program, no cap. The warning is included in -Wall.
• -Wno-discarded-qualifiers (C and Objective-C only) — do not warn if type qualifiers on pointers are being discarded 🙈. Typically, the compiler warns if a "const char *" variable is passed to a function that takes a "char *" parameter. This option can be used to suppress such a warning.
• -Wno-discarded-array-qualifiers (C and Objective-C only) — do not warn if type qualifiers on arrays which are pointer targets are being discarded 🌸. Typically, the compiler warns if a "const int (*)[]" variable is passed to a function that takes a "int (*)[]" parameter. This option can be used to suppress such a warning.
• -Wno-incompatible-pointer-types (C and Objective-C only) — do not warn when there is a conversion between pointers that have incompatible types 🫶. This warning is for cases not covered by -Wno-pointer-sign, which warns for pointer argument passing or assignment with different signedness.

By default, in C99 and later dialects of C, GCC treats this issue as an error 💀. The error can be downgraded to a warning using -fpermissive (along with certain other errors), or for this error alone, with -Wno-error=incompatible-pointer-types.

This warning is upgraded to an error by -pedantic-errors.

• -Wno-int-conversion (C and Objective-C only) — do not warn about incompatible integer to pointer and pointer to integer conversions 🔄. This warning is about implicit conversions; for explicit conversions the warnings -Wno-int-to-pointer-cast and -Wno-pointer-to-int-cast may be used.

By default, in C99 and later dialects of C, GCC treats this issue as an error 💀. The error can be downgraded to a warning using -fpermissive (along with certain other errors), or for this error alone, with -Wno-error=int-conversion.

This warning is upgraded to an error by -pedantic-errors.

• -Wzero-as-null-pointer-constant — warns when a literal 0 is used as null pointer constant 🚫.
• -Wzero-length-bounds — warns about accesses to elements of zero-length array members that might overlap other members of the same object 👀. Declaring interior zero-length arrays is discouraged because accesses to them are undefined.

For example, the first two stores in function "bad" are diagnosed because the array elements overlap the subsequent members "b" and "c". The third store is diagnosed by -Warray-bounds because it is beyond the bounds of the enclosing object.

        struct X { int a[0]; int b, c; };
        struct X x;

        void bad (void)
        {
          x.a[0] = 0;   // -Wzero-length-bounds
          x.a[1] = 1;   // -Wzero-length-bounds
          x.a[2] = 2;   // -Warray-bounds
        }

Option -Wzero-length-bounds is enabled by -Warray-bounds.

• -Wno-div-by-zero — do not warn about compile-time integer division by zero ➗🚫. Floating-point division by zero is not warned about, because it can be a legitimate way of obtaining infinities and NaNs.
• -Wsystem-headers — prints warning messages for constructs found in system header files 📚. Warnings from system headers are normally suppressed, on the assumption that they usually do not indicate real problems and would only make the compiler output harder to read. Using this command-line option tells GCC to emit warnings from system headers as if they occurred in user code.

But wait, bestie — using -Wall in conjunction with this option does not warn about unknown pragmas in system headers. For that, -Wunknown-pragmas must also be used 🙃.

• -Wtautological-compare — warns if a self-comparison always evaluates to true or false 🤡. This warning detects various mistakes such as:

        int i = 1;
        ...
        if (i > i) { ... }

This warning also warns about bitwise comparisons that always evaluate to true or false, for instance:

        if ((a & 16) == 10) { ... }

will always be false.

This warning is enabled by -Wall.

• -Wtrailing-whitespace
• *-Wtrailing-whitespace=kind* — warns about trailing whitespace at the end of lines, including inside of comments, but excluding trailing whitespace in raw string literals ✨. -Wtrailing-whitespace is equivalent to -Wtrailing-whitespace=blanks and warns just about trailing space and horizontal tab characters. -Wtrailing-whitespace=any warns about those or trailing form feed or vertical tab characters. -Wno-trailing-whitespace or -Wtrailing-whitespace=none disables the warning, which is the default. This is a coding style warning. 💅
• -Wleading-whitespace=kind — warns about style issues in leading whitespace, but *not about the amount of indentation 🙅‍♀️. Some projects use coding styles where only spaces are used for indentation, others use only tabs, others use zero or more tabs (for multiples of -ftabstop=n) followed by zero or fewer than *n spaces. No warning is emitted on lines which contain solely whitespace (although -Wtrailing-whitespace= warning might be emitted), no warnings are emitted inside of raw string literals. Warnings are also emitted for leading whitespace inside of multi-line comments.

Let's break down the kinds, bestie:

• -Wleading-whitespace=spaces warns about leading whitespace other than spaces, for projects which want to indent just by spaces.
• -Wleading-whitespace=tabs warns about leading whitespace other than horizontal tabs, for projects which want to indent just by horizontal tabs.
• -Wleading-whitespace=blanks warns about leading whitespace other than spaces and horizontal tabs, or about horizontal tab after a space in the leading whitespace, or about n or more consecutive spaces in leading whitespace (where n is argument of -ftabstop=n, 8 by default).
• -Wleading-whitespace=none disables the warning, which is the default.

This is a coding style warning. 🎀

• -Wtrampolines — warns about trampolines generated for pointers to nested functions 🤸. A trampoline is a small piece of data or code that is created at run time on the stack when the address of a nested function is taken, and is used to call the nested function indirectly. For some targets, it is made up of data only and thus requires no special treatment. But, for most targets, it is made up of code and thus requires the stack to be made executable in order for the program to work properly 😭.
• -Wfloat-equal — warns if floating-point values are used in equality comparisons 🌊💔.

The idea behind this is that sometimes it is convenient (for the programmer) to consider floating-point values as approximations to infinitely precise real numbers. If you are doing this, then you need to compute (by analyzing the code, or in some other way) the maximum or likely maximum error that the computation introduces, and allow for it when performing comparisons (and when producing output, but that's a different problem). In particular, instead of testing for equality, you should check to see whether the two values have ranges that overlap; and this is done with the relational operators, so equality comparisons are probably mistaken.

• -Wtraditional (C and Objective-C only) — warns about certain constructs that behave differently in traditional and ISO C 🕰️. Also warn about ISO C constructs that have no traditional C equivalent, and/or problematic constructs that should be avoided.
  • Macro parameters that appear within string literals in the macro body. In traditional C macro replacement takes place within string literals, but in ISO C it does not.
  • In traditional C, some preprocessor directives did not exist. Traditional preprocessors only considered a line to be a directive if the # appeared in column 1 on the line. Therefore -Wtraditional warns about directives that traditional C understands but ignores because the # does not appear as the first character on the line. It also suggests you hide directives like "#pragma" not understood by traditional C by indenting them. Some traditional implementations do not recognize "#elif", so this option suggests avoiding it altogether.
  • A function-like macro that appears without arguments.
  • The unary plus operator.
  • The U integer constant suffix, or the F or L floating-point constant suffixes. (Traditional C does support the L suffix on integer constants.) Note, these suffixes appear in macros defined in the system headers of most modern systems, e.g. the _MIN/_MAX macros in "<limits.h>". Use of these macros in user code might normally lead to spurious warnings, however GCC's integrated preprocessor has enough context to avoid warning in these cases.
  • A function declared external in one block and then used after the end of the block.
  • A "switch" statement has an operand of type "long".
  • A non-"static" function declaration follows a "static" one. This construct is not accepted by some traditional C compilers.
• The ISO type of an integer constant has a different width or signedness from its traditional type. This warning is only issued if the base of the constant is ten. Like, hex or octal values — which are basically just bit patterns wearing a little outfit — are not warned about. Not the vibe, move on.
• Usage of ISO string concatenation is detected. Caught it, bestie. 🙈
• Initialization of automatic aggregates. Yep, that's a thing that flags.
• Identifier conflicts with labels. Because traditional C literally does not have a separate namespace for labels, so they're all up in each other's business. 😭
• Initialization of unions. If the initializer is zero, the warning is omitted — and the reasoning is that a zero initializer in user code shows up conditioned on e.g. "__STDC__" to dodge missing-initializer warnings, and it's relying on default initialization to zero in the traditional C case. She's thorough, okay.
• Conversions by prototypes between fixed/floating-point values and vice versa. The absence of these prototypes when compiling with traditional C causes serious problems — like, no cap. This is only a subset of the possible conversion warnings; for the full set use -Wtraditional-conversion. ✨
• Use of ISO C style function definitions. This warning intentionally is not issued for prototype declarations or variadic functions, because these ISO C features appear in your code when you're using libiberty's traditional C compatibility macros, "PARAMS" and "VPARAMS". It's also bypassed for nested functions, because that feature is already a GCC extension and therefore not relevant to traditional C compatibility. She said what she said. 💅

---

• -Wtraditional-conversion (C and Objective-C only) — Warn if a prototype causes a type conversion that's different from what would happen to the same argument in the absence of a prototype. This covers conversions of fixed point → floating and floating → fixed, plus conversions that change the width or signedness of a fixed-point argument — except when it's the same as the default promotion.
• -Wdeclaration-after-statement (C and Objective-C only) — Warn when a declaration pops up after a statement in a block. This construct, which you know from C++, was introduced with ISO C99 and is by default allowed in GCC. ISO C90 does not support it, periodt.

This warning is upgraded to an error by -pedantic-errors.

• -Wshadow — Warn whenever a local variable or type declaration shadows another variable, parameter, type, class member (in C++), or instance variable (in Objective-C) — or whenever a built-in function gets shadowed. Heads up: in C++, the compiler warns if a local variable shadows an explicit typedef, but not if it shadows a struct/class/enum. If this warning is enabled, it also includes all instances of local shadowing. Which means -Wno-shadow=local and -Wno-shadow=compatible-local are straight-up ignored when -Wshadow is used. Same as -Wshadow=global.
• -Wno-shadow-ivar (Objective-C only) — Do not warn whenever a local variable shadows an instance variable in an Objective-C method. Chill mode activated. 🦋
• -Wshadow=global — Warn for any shadowing. Same as -Wshadow.
• -Wshadow=local — Warn when a local variable shadows another local variable or parameter.
• -Wshadow=compatible-local — Warn when a local variable shadows another local variable or parameter whose type is compatible with that of the shadowing variable. In C++, "type compatibility" here means the type of the shadowing variable can be converted to that of the shadowed variable. The whole reason this flag exists (on top of -Wshadow=local) is the idea that when a local variable shadows another one of incompatible type, it's most likely intentional — not a bug or a typo. Exhibit A:

        for (SomeIterator i = SomeObj.begin(); i != SomeObj.end(); ++i)
        {
          for (int i = 0; i < N; ++i)
          {
            ...
          }
          ...
        }

Since the two variables named i up there have incompatible types, enabling only -Wshadow=compatible-local does not emit a warning. And because their types are incompatible, if a programmer accidentally uses one in place of the other, type checking is expected to catch that and emit an error or warning. Using this flag instead of -Wshadow=local can possibly reduce the number of warnings triggered by intentional shadowing. Note that this also means shadowing "const char *i" by "char *i" does not emit a warning. Sneaky. 👀

This warning is also enabled by -Wshadow=local.

• *-Wlarger-than=byte-size* — Warn whenever an object is defined whose size exceeds byte-size. -Wlarger-than=PTRDIFF_MAX is enabled by default. Warnings controlled by this option can be disabled either by specifying a byte-size of SIZE_MAX or more, or by -Wno-larger-than.

It also warns for calls to bounded functions like "memchr" or "strnlen" that specify a bound greater than the largest possible object — which is PTRDIFF_MAX bytes by default. These warnings can only be disabled by -Wno-larger-than.

• -Wno-larger-than — Disable -Wlarger-than= warnings. This option is equivalent to -Wlarger-than=SIZE_MAX or larger.
• -Wframe-larger-than=byte-size — Warn if the size of a function frame exceeds byte-size. The computation done to figure out the stack frame size is approximate and *not conservative, so the actual requirements may be somewhat greater than *byte-size even if you don't get a warning. On top of that, any space allocated via "alloca", variable-length arrays, or related constructs is not included by the compiler when deciding whether or not to issue a warning. -Wframe-larger-than=PTRDIFF_MAX is enabled by default. Warnings controlled by this option can be disabled either by specifying a byte-size of SIZE_MAX or more, or by -Wno-frame-larger-than.
• -Wno-frame-larger-than — Disable -Wframe-larger-than= warnings. This option is equivalent to -Wframe-larger-than=SIZE_MAX or larger.
• -Wfree-nonheap-object — Warn when you attempt to deallocate an object that was either not allocated on the heap, or by using a pointer that wasn't returned from a prior call to the corresponding allocation function. For example, because the call to "stpcpy" returns a pointer to the terminating nul character and not to the beginning of the object, the call to "free" below is diagnosed:

        void f (char *p)
        {
          p = stpcpy (p, "abc");
          // ...
          free (p);   // warning
        }

-Wfree-nonheap-object is included in -Wall.

• -Wstack-usage=byte-size — Warn if the stack usage of a function might exceed byte-size. The computation done to determine stack usage is conservative (unlike that other one 😌), and any space allocated via "alloca", variable-length arrays, or related constructs **is included by the compiler when determining whether or not to issue a warning.

The message stays in keeping with the output of -fstack-usage.

• If the stack usage is fully static but exceeds the specified amount, it's:

          warning: stack usage is 1120 bytes

• If the stack usage is (partly) dynamic but bounded, it's:

          warning: stack usage might be 1648 bytes

• If the stack usage is (partly) dynamic and not bounded, it's:

          warning: stack usage might be unbounded

-Wstack-usage=PTRDIFF_MAX is enabled by default. Warnings controlled by this option can be disabled either by specifying a byte-size of SIZE_MAX or more, or by -Wno-stack-usage. 💜

• -Wno-stack-usage — Disable -Wstack-usage= warnings. This option is equivalent to -Wstack-usage=SIZE_MAX or larger.
• -Wunsafe-loop-optimizations — Warn if the loop cannot be optimized because the compiler can't assume anything about the bounds of the loop indices. With -funsafe-loop-optimizations, warn if the compiler makes such assumptions. Iconic risk-taking behavior, honestly. 🔥
• -Wno-pedantic-ms-format (MinGW targets only) — When used together with -Wformat and -pedantic without GNU extensions, this option disables the warnings about non-ISO "printf" / "scanf" format width specifiers "I32", "I64", and "I" used on Windows targets — the ones that depend on the MS runtime. Not our problem anymore. 🎀
• -Wpointer-arith — Warn about anything that depends on the "size of" a function type or of "void". GNU C assigns these types a size of 1, purely for convenience when doing calculations with "void *" pointers and pointers to functions. In C++, it also warns when an arithmetic operation involves "NULL". This warning is also enabled by -Wpedantic.

This warning is upgraded to an error by -pedantic-errors.

• -Wno-pointer-compare — Do not warn if a pointer is compared with a zero character constant. That usually means the pointer was meant to be dereferenced. Like:

        const char *p = foo ();
        if (p == '\0')
          return 42;

Note that the code above is invalid in C++11. Oops. 😬

This warning is enabled by default.

• -Wno-tsan — Disable warnings about unsupported features in ThreadSanitizer.

ThreadSanitizer does not support "std::atomic_thread_fence" and can report false positives. She's trying her best, okay.

• -Wtype-limits — Warn if a comparison is always true or always false because of the limited range of the data type — but do not warn for constant expressions. For example, warn if an unsigned variable is compared against zero with "<" or ">=". This warning is also enabled by -Wextra.
• -Wabsolute-value (C and Objective-C only) — Warn for calls to standard functions that compute the absolute value of an argument when a more appropriate standard function is available. For example, calling abs(3.14) triggers the warning, because the appropriate function to compute the absolute value of a double argument is "fabs". The option also triggers warnings when the argument in a call to such a function has an unsigned type. Nothing personal, just math. 💅 This warning can be suppressed with an explicit type cast, and it's also enabled by -Wextra.
• -Wcomment
• -Wcomments — Warn whenever a comment-start sequence /* appears in a /* comment, or whenever a backslash-newline appears in a // comment. This warning is enabled by -Wall.
• -Wtrigraphs — Warn if any trigraphs are encountered that might change the meaning of the program. Trigraphs inside comments are not warned about — except the ones that would form escaped newlines. 🌀

This option is implied by -Wall. And if -Wall isn't even on the table? It's still on, bestie — unless trigraphs are enabled. So if you want the trigraph conversion without the nagging, but you still want all the other -Wall warnings, you do:

-trigraphs -Wall -Wno-trigraphs

and you're golden. ✨

---

• -Wundef — yells at you when an undefined identifier gets evaluated inside an #if directive. Those mystery identifiers just get swapped for zero, which is exactly the kind of quiet chaos you want a warning about. 💀
• -Wexpansion-to-defined — warns whenever defined shows up in the expansion of a macro (yes, including when the macro gets expanded by an #if directive). This usage is not portable, so it's giving red flag 🚩. Also switched on by -Wpedantic and -Wextra.
• -Wunused-macros — warns about macros defined in the main file that never get used. A macro counts as "used" if it's expanded or tested for existence at least once. The preprocessor also warns if the macro hasn't been used by the time it's redefined or #undef'd.

Built-in macros, macros defined on the command line, and macros defined in include files are exempt — no warnings for those, they're protected 🫶.

Note: if a macro is actually used but only inside skipped conditional blocks, the preprocessor still reports it as unused. To dodge that warning, you can tighten up the scope of the macro's definition — like moving it into the first skipped block. Or you can just give it a dummy use, something like:

#if defined the_macro_causing_the_warning
#endif

• -Wno-endif-labels — do not warn when an #else or an #endif is followed by text. This is a whole vibe in older programs, code that looks like:

#if FOO
...
#else FOO
...
#endif FOO

Bestie, that second and third FOO should be chilling inside comments. This warning is on by default.

• -Wbad-function-cast (C and Objective-C only) — warns when a function call gets cast to a type that doesn't match. Example: a call to a function returning an integer type being cast to a pointer type. Ma'am, that's not it. 😭
• -Wc90-c99-compat (C and Objective-C only) — warns about features that aren't in ISO C90 but are in ISO C99. Like variable length arrays, the long long type, the bool type, compound literals, designated initializers, and so on. This option is independent of the standards mode. Warnings are disabled inside the expression that follows __extension__.
• -Wc99-c11-compat (C and Objective-C only) — warns about features that aren't in ISO C99 but are in ISO C11. Think anonymous structures and unions, the _Atomic type qualifier, the _Thread_local storage-class specifier, the _Alignas specifier, the Alignof operator, the _Generic keyword, and so forth. Also independent of the standards mode, and warnings are disabled in the expression that follows __extension__.
• -Wc11-c23-compat (C and Objective-C only)
• -Wc11-c2x-compat (C and Objective-C only)

Warns about features that aren't in ISO C11 but are in ISO C23. Like omitting the string in _Static_assert, using [[]] syntax for attributes, decimal floating-point types, and so on. Independent of the standards mode, warnings disabled in the expression after __extension__. Heads up: the name -Wc11-c2x-compat is deprecated 💀.

When you're not compiling in C23 mode, these warnings get upgraded to errors by -pedantic-errors.

• -Wc23-c2y-compat (C and Objective-C only)
• -Wc23-c2y-compat (C and Objective-C only)

Warns about features that aren't in ISO C23 but are in ISO C2Y. Like _Generic selecting with a type name instead of an expression. Independent of the standards mode, and again warnings are disabled in the expression that follows __extension__.

When you're not compiling in C2Y mode, -pedantic-errors upgrades these warnings to errors.

• -Wc++-compat (C and Objective-C only) — warns about ISO C constructs that live outside the common subset of ISO C and ISO C++, e.g. asking for an implicit conversion from void * to a pointer to a non-void type.
• -Wc++11-compat (C++ and Objective-C++ only) — warns about C++ constructs whose meaning changes between ISO C++ 1998 and ISO C++ 2011, e.g. identifiers that were fine in ISO C++ 1998 but became keywords in ISO C++ 2011. This warning turns on -Wnarrowing and is enabled by -Wall.
• -Wc++14-compat (C++ and Objective-C++ only) — warns about C++ constructs whose meaning differs between ISO C++ 2011 and ISO C++ 2014. Enabled by -Wall.
• -Wc++17-compat (C++ and Objective-C++ only) — warns about C++ constructs whose meaning differs between ISO C++ 2014 and ISO C++ 2017. Enabled by -Wall.
• -Wc++20-compat (C++ and Objective-C++ only) — warns about C++ constructs whose meaning differs between ISO C++ 2017 and ISO C++ 2020. Enabled by -Wall.
• -Wno-c++11-extensions (C++ and Objective-C++ only) — stop warning about C++11 constructs in code that's being compiled with an older C++ standard. Even without this option, some C++11 constructs only get diagnosed if -Wpedantic is used.
• -Wno-c++14-extensions (C++ and Objective-C++ only) — same deal, but for C++14 constructs compiled under an older C++ standard. Even without it, some C++14 constructs only get diagnosed with -Wpedantic.
• -Wno-c++17-extensions (C++ and Objective-C++ only) — same energy, C++17 constructs, older C++ standard. Some C++17 constructs still only get diagnosed with -Wpedantic.
• -Wno-c++20-extensions (C++ and Objective-C++ only) — you already know the pattern: C++20 constructs, older standard. Some C++20 constructs only get diagnosed with -Wpedantic.
• -Wno-c++23-extensions (C++ and Objective-C++ only) — C++23 constructs, older standard, no warning. And yep, some C++23 constructs only get diagnosed with -Wpedantic.
• -Wno-c++26-extensions (C++ and Objective-C++ only) — C++26 constructs compiled under an older C++ standard, silenced. Some C++26 constructs only get diagnosed if you're running -Wpedantic.

---

• -Wcast-qual — warns whenever a pointer is cast in a way that strips a type qualifier off the target type. Like when a const char * gets cast to a plain old char *. Excuse me?? 👀

It also warns when a cast introduces a type qualifier in an unsafe way. For example, casting char ** to const char ** is unsafe, like this:

/* p is char ** value.  */
const char **q = (const char **) p;
/* Assignment of readonly string to const char * is OK.  */
*q = "string";
/* Now char** pointer points to read-only memory.  */
**p = 'b';

• -Wcast-align — warns whenever a pointer is cast such that the target's required alignment gets bumped up. Example: a char * cast to an int * on machines where integers can only be accessed at two- or four-byte boundaries.
• -Wcast-align=strict — warns whenever a pointer is cast such that the target's required alignment increases. Example: a char * cast to an int * — but this time regardless of the target machine. No machine left behind 💅
• -Wcast-function-type — warns when a function pointer gets cast to an incompatible function pointer. If the cast involves function types with a variable argument list, only the types of the initial arguments that are provided get considered. Any pointer-type parameter matches any other pointer-type parameter. Benign differences in integral types are ignored (like int vs. long on ILP32 targets). Type qualifiers get ignored too. The function type void (*) (void) is special — it matches everything, which you can use to suppress this warning. 👑 In a cast involving pointer-to-member types, this warning fires whenever the cast changes the pointer-to-member type. This one's enabled by -Wextra.
• -Wcast-user-defined — warns when a cast to a reference type doesn't involve a user-defined conversion that the programmer probably expected to be called.

struct A { operator const int&(); } a;
auto r = (int&)a; // warning

This warning is on by default.

• -Wwrite-strings — when compiling C, this gives string constants the type const char[length], so copying the address of one into a non-const char * pointer produces a warning. These warnings help you catch compile-time code that might try to write into a string constant — but only if you've been extremely careful about using const in your declarations and prototypes. Otherwise it's just a nuisance, which is exactly why -Wall does not request these warnings.

When compiling C++, it warns about the deprecated conversion from string literals to char *. In C++ programs this warning is enabled by default.

And in C++11 mode or later, -pedantic-errors upgrades this warning to an error. 😤

• -Wclobbered — warns for variables that might get changed by longjmp or vfork. Also enabled by -Wextra.
• -Wno-complain-wrong-lang — by default, language front ends complain when a command-line option is valid but just doesn't apply to that front end. You can turn that off with -Wno-complain-wrong-lang, which is mostly useful when you're invoking a single compiler driver on multiple source files written in different languages, like:

$ g++ -fno-rtti a.cc b.f90

Here the driver g++ invokes the C++ front end to compile a.cc and the Fortran front end to compile b.f90. That latter front end diagnoses:

f951: Warning: command-line option '-fno-rtti' is valid for C++/D/ObjC++ but not for Fortran

and that lil warning can be disabled with -Wno-complain-wrong-lang. 🎀

• -Wcompare-distinct-pointer-types (C and Objective-C only) — warns if pointers of distinct types get compared without a cast. This warning is enabled by default.
• -Wconversion — bestie, this one yells at you for implicit conversions that might change a value 💅. That means stuff like real↔integer conversions, like abs (x) when x is double; signed↔unsigned, like unsigned ui = -1; and conversions to smaller types, like sqrtf (M_PI). But it stays quiet for explicit casts like abs ((int) x) and ui = (unsigned) -1, and it also stays quiet if the value isn't even changed by the conversion, like in abs (2.0). Wanna mute the signed↔unsigned gossip specifically? Use -Wno-sign-conversion.

For C++, it also warns for confusing overload resolution for user-defined conversions, and for conversions that never use a type conversion operator: conversions to void, the same type, a base class, or a reference to them. Warning about signed↔unsigned conversions is off by default in C++ unless -Wsign-conversion is explicitly enabled.

One more: warnings about conversion from arithmetic on a small type back to that same type only show up with -Warith-conversion. periodt.

• -Wdangling-else — this is the ✨which if does the else belong to✨ drama, and honestly it's giving chaos. Here's the cursed example:

{
  if (a)
    if (b)
      foo ();
  else
    bar ();
}

In C/C++, every else branch belongs to the innermost possible if statement, which here is if (b). That's often not what the programmer expected — the indentation in that example says it all 🫠. When there's potential for this confusion, GCC throws a warning when this flag is specified. Wanna kill the warning? Add explicit braces around the innermost if statement so there's no way the else can belong to the enclosing if. The fixed-up version looks like:

{
  if (a)
    {
      if (b)
        foo ();
      else
        bar ();
    }
}

This warning is enabled by -Wparentheses.

• -Wdangling-pointer, -Wdangling-pointer=n — no cap, this one's watching pointers (or C++ references) to objects with automatic storage duration after their lifetime has ended 💀. That includes local variables declared in nested blocks, compound literals, and other unnamed temporary objects. It also warns about storing the address of such objects in escaped pointers. The warning is enabled at all optimization levels, but may yield different results with optimization than without.
  • -Wdangling-pointer=1 — at level 1, it only diagnoses unconditional uses of dangling pointers.
  • -Wdangling-pointer=2 — at level 2, in addition to unconditional uses, it also diagnoses conditional uses of dangling pointers.

The short form -Wdangling-pointer is equivalent to -Wdangling-pointer=2, while -Wno-dangling-pointer and -Wdangling-pointer=0 have the same effect of disabling the warnings. And -Wdangling-pointer=2 is included in -Wall 👑

This example triggers the warning at level 1 — the address of the unnamed temporary is unconditionally referenced outside of its scope:

char f (char c1, char c2, char c3)
{
  char *p;
  {
    p = (char[]) { c1, c2, c3 };
  }
  // warning: using dangling pointer 'p' to an unnamed temporary
  return *p;
}

In the following function, the store of the address of the local variable x in the escaped pointer *p triggers the warning at level 1:

void g (int **p)
{
  int x = 7;
  // warning: storing the address of local variable 'x' in '*p'
  *p = &x;
}

And in this example, the array a is out of scope when the pointer s is used. Since the code that sets s is conditional, the warning triggers at level 2:

extern void frob (const char *);
void h (char *s)
{
  if (!s)
    {
      char a[12] = "tmpname";
      s = a;
    }
  // warning: dangling pointer 's' to 'a' may be used
  frob (s);
}

• -Wdate-time — warns when the macros __TIME__, __DATE__ or __TIMESTAMP__ show up, because they might prevent bit-wise-identical reproducible compilations 😭. it's giving "why is my build not reproducible".
• -Wempty-body — warns if an empty body occurs in an if, else or do while statement. Also enabled by -Wextra ✨.
• -Wno-endif-labels — do not warn about stray tokens after #else and #endif. she's chill like that.
• -Wenum-compare — warns about a comparison between values of different enumerated types. In C++, enumerated type mismatches in conditional expressions are also diagnosed, and the warning is enabled by default. In C, this warning is enabled by -Wall.
• -Wenum-conversion — warns when a value of enumerated type is implicitly converted to a different enumerated type. This warning is enabled by -Wextra in C.
• -Wenum-int-mismatch (C and Objective-C only) — warns about mismatches between an enumerated type and an integer type in declarations. Like this:

enum E { l = -1, z = 0, g = 1 };
int foo(void);
enum E foo(void);

In C, an enumerated type is compatible with char, a signed integer type, or an unsigned integer type. But since the choice of the underlying type of an enumerated type is implementation-defined, such mismatches may cause portability issues 🚩. In C++, such mismatches are an error. In C, this warning is enabled by -Wall and -Wc++-compat.

• -Wjump-misses-init (C, Objective-C only) — warns if a goto statement or a switch statement jumps forward across the initialization of a variable, or jumps backward to a label after the variable has been initialized. It only warns about variables that are initialized when they are declared. This warning is only supported for C and Objective-C; in C++ this sort of branch is an error in any case. It's included in -Wc++-compat, and can be disabled with the -Wno-jump-misses-init option.
• -Wsign-compare — warns when a comparison between signed and unsigned values could produce an incorrect result when the signed value is converted to unsigned 😬. In C++, this warning is also enabled by -Wall. In C, it is also enabled by -Wextra.
• -Wsign-conversion — warns for implicit conversions that may change the sign of an integer value, like assigning a signed integer expression to an unsigned integer variable. An explicit cast silences the warning. In C, this option is enabled also by -Wconversion.
• -Wflex-array-member-not-at-end (C and C++ only) — warns when a structure containing a C99 flexible array member as the last field is not at the end of another structure. Like, it warns about stuff like this:

struct flex  { int length; char data[]; };
struct mid_flex { int m; struct flex flex_data; int n; };

• -Wfloat-conversion — warns for implicit conversions that reduce the precision of a real value 🌸. This includes conversions from real to integer, and from higher precision real to lower precision real values. This option is also enabled by -Wconversion.
• -Wno-scalar-storage-order — do not warn on suspicious constructs involving reverse scalar storage order. 🙈
• -Wsizeof-array-div — warns about divisions of two sizeof operators when the first one is applied to an array and the divisor does not equal the size of the array element. In that case, the computation will not yield the number of elements in the array — which is probably what the user intended 🥲. So it warns about stuff like:

int fn ()
{
  int arr[10];
  return sizeof (arr) / sizeof (short);
}

This warning is enabled by -Wall.

• -Wsizeof-pointer-div — warns for suspicious divisions of two sizeof expressions that divide the pointer size by the element size, which is the usual way to compute the array size but won't work out correctly with pointers 💀. It warns e.g. about sizeof (ptr) / sizeof (ptr[0]) if ptr is not an array, but a pointer. This warning is enabled by -Wall.
• -Wsizeof-pointer-memaccess — warns for suspicious length parameters to certain string and memory built-in functions if the argument uses sizeof. It triggers, for example, for memset (ptr, 0, sizeof (ptr)); if ptr is not an array, but a pointer, and suggests a possible fix — or about memcpy (&foo, ptr, sizeof (&foo)); 😭.

It also warns about calls to bounded string copy functions like strncat or strncpy that specify as the bound a sizeof expression of the source array. For example, in this function the call to strncat specifies the size of the source string as the bound. That is almost certainly a mistake, so the call gets diagnosed:

void make_file (const char *name)
{
  char path[PATH_MAX];
  strncpy (path, name, sizeof path - 1);
  strncat (path, ".text", sizeof ".text");
  ...
}

The -Wsizeof-pointer-memaccess option is enabled by -Wall.

• -Wno-sizeof-array-argument — do not warn when the sizeof operator is applied to a parameter that is declared as an array in a function definition. This warning is enabled by default for C and C++ programs.
• -Wmemset-elt-size — warns for suspicious calls to the memset built-in function, if the first argument references an array, and the third argument is a number equal to the number of elements, but not equal to the size of the array in memory 🫠. That's a big hint that the user forgot to multiply by the element size. This warning is enabled by -Wall.

---

• -Wmemset-transposed-args — bestie, this one's a vibe check for sus memset calls 👀 specifically when arg two isn't zero but arg three is zero. Like memset (buf, sizeof buf, 0) — that gets flagged because you literally meant memset (buf, 0, sizeof buf). But here's the tea: the diagnostic only fires if the third argument is a literal zero. If it's an expression that gets folded to zero, or a cast of zero to some type, then it's way less likely you transposed stuff, so no warning. This one's enabled by -Wall. ✨
• -Waddress — warns about sus uses of address expressions 💅. That means comparing the address of a function or a declared object to the null pointer constant, like:

        void f (void);
        void g (void)
        {
          if (!f)   // warning: expression evaluates to false
            abort ();
        }

Also comparisons of a pointer to a string literal, like:

        void f (const char *x)
        {
          if (x == "abc")   // warning: expression evaluates to false
            puts ("equal");
        }

And tests of the results of pointer addition or subtraction for equality to null, like:

        void f (const int *p, int i)
        {
          return p + i == NULL;
        }

Such uses typically mean a programmer oopsie: the address of most functions and objects necessarily evaluates to true (only exception? weak symbols 🫠), so using one in a conditional might mean you forgot parentheses in a function call or forgot to dereference in an array expression. The object-pointer subset of this warning can be silenced by casting the pointer operand to an integer type like intptr_t or uintptr_t. Comparing against string literals gives unspecified behavior and isn't portable, and ngl it kinda screams "you meant to call strcmp". The warning is suppressed if the sus expression came from macro expansion. -Waddress is enabled by -Wall. 🌸

• -Wno-address-of-packed-member — don't warn when you take the address of a packed member of a struct or union, which usually gives you an unaligned pointer value. Enabled by default, no cap. 🎀
• -Wlogical-op — warns about sus uses of logical operators in expressions ✨. Includes using logical operators where a bit-wise operator was probably expected. Also warns when the operands of a logical operator are literally the same:

        extern int a;
        if (a < 0 && a < 0) { ... }

• -Wlogical-not-parentheses — warns about logical not used on the left hand side operand of a comparison 👑. It does not warn if the right operand is considered a boolean expression. Its whole purpose is catching sus code like:

        int a;
        ...
        if (!a > 1) { ... }

You can suppress it by wrapping the LHS in parentheses:

        if ((!a) > 1) { ... }

This warning is enabled by -Wall. 💖

• -Waggregate-return — warns if any functions that return structures or unions are defined or called 🌈. (And in languages where you can return an array, that also gets you a warning.)
• -Wno-aggressive-loop-optimizations — don't warn if the compiler detects undefined behavior in a loop with a constant number of iterations 💀. Meanwhile -Waggressive-loop-optimizations is enabled by default.
• -Wno-attributes — don't warn if an unexpected __attribute__ shows up, like unrecognized attributes, function attributes applied to variables, etc 🙈. Note: this does not stop errors for incorrect use of supported attributes.

Warnings about ill-formed uses of standard attributes get upgraded to errors by -pedantic-errors.

Also, with -Wno-attributes=, you can suppress warnings about unknown scoped attributes (in C++11 and C23) 🦋. Example: -Wno-attributes=vendor::attr disables the warning about this declaration:

        [[vendor::attr]] void f();

And you can kill the warning for all attributes in a namespace using -Wno-attributes=vendor::, which prevents warnings about both of these declarations:

        [[vendor::safe]] void f();
        [[vendor::unsafe]] void f2();

Heads up though: -Wno-attributes= does not imply -Wno-attributes. ⭐

• -Wno-builtin-declaration-mismatch — warns if a built-in function is declared with an incompatible signature or as a non-function 😭, or when a built-in function declared with a type that doesn't include a prototype gets called with arguments whose promoted types don't match what the function expects. When -Wextra is specified, it also warns when a built-in function that takes arguments is declared without a prototype. The -Wbuiltin-declaration-mismatch warning is enabled by default. To dodge the warning, include the appropriate header so the built-in function's prototypes come into scope.

For example, the call to memset below gets diagnosed because the function expects a value of type size_t as its argument but the type of 32 is int. With -Wextra, the declaration of the function is diagnosed too:

        extern void* memset ();
        void f (void *d)
        {
          memset (d, '\0', 32);
        }

• -Wno-builtin-macro-redefined — don't warn if certain built-in macros get redefined 🙅♀️. This suppresses warnings for redefinition of __TIMESTAMP__, __TIME__, __DATE__, __FILE__, and __BASE_FILE__.
• -Wfree-labels (C and Objective-C only) — warns if a label is applied to a non-statement, or shows up at the end of a compound statement 🌸. Such labels are allowed by C23 and later dialects of C, and are available as a GCC extension in all the other dialects.

This warning is also enabled by -Wc11-c23-compat. And it gets turned into an error if you're building for a C version before C23 by -pedantic-errors. 💅

• -Wheader-guard — warns if a valid preprocessor header multiple inclusion guard has a #define directive right after the #ifndef or #if !defined directive for the multiple inclusion guard, which defines a different macro from the guard macro with a similar name 😵💫, the actual multiple inclusion guard macro isn't defined at the corresponding #ifndef directive at the end of the header, and the #define directive defines an object-like macro with an empty definition. In that case, it's usually just a misspelled guard name — either in the #ifndef or #if !defined directive, or in the subsequent #define directive. This warning is enabled by -Wall. ✨
• -Wstrict-prototypes (C and Objective-C only) — warns if a function is declared or defined without specifying the argument types 🫶. (An old-style function definition is permitted without a warning if it's preceded by a declaration that specifies the argument types.)
• -Wold-style-declaration (C and Objective-C only) — warns for obsolescent usages, per the C Standard, in a declaration 👑. For example, it warns if storage-class specifiers like static aren't the first things in a declaration. This warning is also enabled by -Wextra.
• -Wold-style-definition (C and Objective-C only) — warns if an old-style function definition is used 💀. A warning is given even if there's a previous prototype. A definition using () is not considered an old-style definition in C23 mode, because it's equivalent to (void) in that case — but it is considered an old-style definition for older standards.
• -Wdeprecated-non-prototype (C and Objective-C only) — warns if a function declared with an empty parameter list () is called with one or more arguments, or if a function definition with one or more parameters shows up after such a declaration 😭. Both cases are errors in C23 and later dialects of C.

This warning is also enabled by -Wc11-c23-compat.

• -Wmissing-parameter-name (C and Objective-C only) — warns if a function definition omits a parameter name, specifying only its type 🎀. This can be used to document that a parameter is unused in the definition. It's part of C23 and later dialects of C, and available as a GCC extension in all other dialects.

This warning is also enabled by -Wc11-c23-compat. It gets turned into an error if you're building for a C version before C23 by -pedantic-errors. ⭐

• -Wmissing-parameter-type (C and Objective-C only) — fires when a function parameter is declared without a type specifier in K&R-style functions:

        void foo(bar) { }

This warning is also enabled by -Wextra. 🌈

• -Wno-declaration-missing-parameter-type (C and Objective-C only) — don't warn if a function declaration contains a parameter name without a type 🙈. Such function declarations don't provide a function prototype and they block most type checking in function calls.

This warning is enabled by default. In C99 and later dialects of C, it's treated as an error. The error can be downgraded to a warning using -fpermissive (along with certain other errors), or for this error alone, with -Wno-error=declaration-missing-parameter-type.

This warning is upgraded to an error by -pedantic-errors. 💅

• -Wmissing-prototypes (C and Objective-C only) — warns if a global function is defined without a previous prototype declaration 🦋. This warning shows up even if the definition itself provides a prototype. Use this option to catch global functions that don't have a matching prototype declaration in a header file. Not valid for C++ though, because all function declarations provide prototypes and a non-matching declaration declares an overload rather than conflicting with an earlier declaration. In C++, use -Wmissing-declarations instead to detect missing declarations. ✨
• -Wmissing-variable-declarations (C and Objective-C only) — warns if a global variable is defined without a previous declaration 🌸. Use this option to catch global variables that don't have a matching extern declaration in a header file. 💖
• -Wmissing-declarations — bestie, this one yells at you if a global function gets defined without a previous declaration, even if the definition itself hands you a prototype. Use it to catch global functions that were never declared in header files. In C, no warnings for functions with previous non-prototype declarations; use -Wmissing-prototypes for missing prototypes. In C++, no warnings for function templates, inline functions, or functions in anonymous namespaces. slay responsibly 💅
• -Wmissing-field-initializers — warns when a structure's initializer leaves some fields missing. Like, the code below totally triggers this because x.h is implicitly zero:

struct s { int f, g, h; };
struct s x = { 3, 4 };

In C, this option does not warn about designated initializers, so this little glow-up does not trigger a warning:

struct s { int f, g, h; };
struct s x = { .f = 3, .g = 4 };

Also in C, it does not warn about the universal zero initializer { 0 }:

struct s { int f, g, h; };
struct s x = { 0 };

And in C++, it does not warn about the empty { } initializer, for example:

struct s { int f, g, h; };
s x = { };

This warning is included in -Wextra. To get other -Wextra warnings without this one, use -Wextra -Wno-missing-field-initializers. the more you know ✨

• -Wno-missing-requires — by default, the compiler warns about a concept-id appearing as a C++20 simple-requirement:

bool satisfied = requires { C<T> };

Here satisfied will be true if C<T> is a valid expression, which it is for all T. Presumably the user meant to write this instead:

bool satisfied = requires { requires C<T> };

so satisfied is only true if concept C is satisfied for type T. This warning can be disabled with -Wno-missing-requires. the way C++ do be C++ 😭

• -Wno-missing-template-keyword — the member access tokens ., -> and :: must be followed by the "template" keyword if the parent object is dependent and the member being named is a template.

template <class X>
void DoStuff (X x)
{
  x.template DoSomeOtherStuff<X>(); // Good.
  x.DoMoreStuff<X>(); // Warning, x is dependent.
}

In rare cases it is possible to get false positives. To silence this, wrap the expression in parentheses. For example, the following is treated as a template, even where m and N are integers:

void NotATemplate (my_class t)
{
  int N = 5;

  bool test = t.m < N > (0); // Treated as a template.
  test = (t.m < N) > (0); // Same meaning, but not treated as a template.
}

This warning can be disabled with -Wno-missing-template-keyword. false positives are the villain origin story fr fr

• -Wno-multichar — do not warn if a multicharacter constant ('FOOF') is used. Usually they indicate a typo in the user's code, as they have implementation-defined values, and should not be used in portable code. just don't do it bestie
• -Wnormalized=[none|id|nfc|nfkc] — in ISO C and ISO C++, two identifiers are different if they are different sequences of characters. However, sometimes when characters outside the basic ASCII character set are used, you can have two different character sequences that look the same. To avoid confusion, the ISO 10646 standard sets out some normalization rules which when applied ensure that two sequences that look the same are turned into the same sequence. GCC can warn you if you are using identifiers that have not been normalized; this option controls that warning.

There are four levels of warning supported by GCC. The default is -Wnormalized=nfc, which warns about any identifier that is not in the ISO 10646 "C" normalized form, NFC. NFC is the recommended form for most uses. It is equivalent to -Wnormalized.

Unfortunately, there are some characters allowed in identifiers by ISO C and ISO C++ that, when turned into NFC, are not allowed in identifiers. That is, there's no way to use these symbols in portable ISO C or C++ and have all your identifiers in NFC. -Wnormalized=id suppresses the warning for these characters. It is hoped that future versions of the standards involved will correct this, which is why this option is not the default.

You can switch the warning off for all characters by writing -Wnormalized=none or -Wno-normalized. You should only do this if you are using some other normalization scheme (like "D"), because otherwise you can easily create bugs that are literally impossible to see. invisible bugs? not the vibe 😭

Some characters in ISO 10646 have distinct meanings but look identical in some fonts or display methodologies, especially once formatting has been applied. For instance "\u207F", "SUPERSCRIPT LATIN SMALL LETTER N", displays just like a regular "n" that has been placed in a superscript. ISO 10646 defines the NFKC normalization scheme to convert all these into a standard form as well, and GCC warns if your code is not in NFKC if you use -Wnormalized=nfkc. This warning is comparable to warning about every identifier that contains the letter O because it might be confused with the digit 0, and so is not the default, but may be useful as a local coding convention if the programming environment cannot be fixed to display these characters distinctly. unicode girlies stay winning 🦋

• -Wno-attribute-warning — do not warn about usage of functions declared with "warning" attribute. By default, this warning is enabled. -Wno-attribute-warning can be used to disable the warning or -Wno-error=attribute-warning can be used to disable the error when compiled with -Werror flag.
• -Wno-deprecated — do not warn about usage of deprecated features.

In C++, explicitly specifying -Wdeprecated also enables warnings about some features that are deprecated in later language standards, specifically -Wcomma-subscript, -Wvolatile, -Wdeprecated-enum-float-conversion, -Wdeprecated-enum-enum-conversion, -Wdeprecated-literal-operator, and -Wdeprecated-variadic-comma-omission. the deprecated lore is deep 🍓

• -Wno-deprecated-declarations — do not warn about uses of functions, variables, and types marked as deprecated by using the "deprecated" attribute.
• -Wno-overflow — do not warn about compile-time overflow in constant expressions.
• -Wno-odr — warn about One Definition Rule violations during link-time optimization. Enabled by default. (yes, the flag says -Wno- but the text literally says warn — I don't make the rules 💀)
• -Wopenacc-parallelism — warn about potentially suboptimal choices related to OpenACC parallelism.
• -Wno-openmp — warn about suspicious OpenMP code. again with the naming chaos ✨
• -Wopenmp-simd — warn if the vectorizer cost model overrides the OpenMP simd directive set by user. The -fsimd-cost-model=unlimited option can be used to relax the cost model.
• -Woverride-init (C and Objective-C only) — warn if an initialized field without side effects is overridden when using designated initializers.

This warning is included in -Wextra. To get other -Wextra warnings without this one, use -Wextra -Wno-override-init.

• -Wno-override-init-side-effects (C and Objective-C only) — do not warn if an initialized field with side effects is overridden when using designated initializers. This warning is enabled by default.
• -Wpacked — warn if a structure is given the packed attribute, but the packed attribute has no effect on the layout or size of the structure. Such structures may be mis-aligned for little benefit. For instance, in this code, the variable f.x in struct bar is misaligned even though struct bar does not itself have the packed attribute:

struct foo {
  int x;
  char a, b, c, d;
} __attribute__((packed));
struct bar {
  char z;
  struct foo f;
};

misaligned for no reason? it's giving tragic 🫠

• -Wnopacked-bitfield-compat — the 4.1, 4.2 and 4.3 series of GCC ignore the "packed" attribute on bit-fields of type "char". This was fixed in GCC 4.4 but the change can lead to differences in the structure layout. GCC informs you when the offset of such a field has changed in GCC 4.4. For example there is no longer a 4-bit padding between field a and b in this structure:

struct foo
{
  char a:4;
  char b:8;
} __attribute__ ((packed));

This warning is enabled by default. Use -Wno-packed-bitfield-compat to disable this warning.

• -Wpacked-not-aligned (C, C++, Objective-C and Objective-C++ only) — warn if a structure field with explicitly specified alignment in a packed struct or union is misaligned. For example, a warning will be issued on struct S, like, "warning: alignment 1 of 'struct S' is less than 8", in this code:

struct __attribute__ ((aligned (8))) S8 { char a[8]; };
struct __attribute__ ((packed)) S {
  struct S8 s8;
};

This warning is enabled by -Wall. alignment drama, but make it compiler-core 💅

• -Wpadded — warn if padding is included in a structure, either to align an element of the structure or to align the whole structure. Sometimes when this happens it is possible to rearrange the fields of the structure to reduce the padding and so make the structure smaller.
• -Wredundant-decls — warn if anything is declared more than once in the same scope, even in cases where multiple declaration is valid and changes nothing.
• -Wrestrict — warn when an object referenced by a restrict-qualified parameter (or, in C++, a __restrict-qualified parameter) is aliased by another argument, or when copies between such objects overlap. For example, the call to the strcpy function below attempts to truncate the string by replacing its initial characters with the last four. However, because the call writes the terminating NUL into a[4], the copies overlap and the call is diagnosed.

void foo (void)
{
  char a[] = "abcd1234";
  strcpy (a, a + 4);
  ...
}

The -Wrestrict option detects some instances of simple overlap even without optimization but works best at -O2 and above. It is included in -Wall. overlap? caught in 4K 🔥

• -Wnested-externs (C and Objective-C only) — warn if an extern declaration is encountered within a function.
• -Winline — bestie, this one yells at you if a function you declared as inline straight up cannot be inlined. Rude, right? BUT 💅 even with this flag on, the compiler stays quiet about failures to inline functions declared in system headers. It's giving selective.

So like, the compiler has a whole bunch of ✨heuristics✨ to decide whether or not it's gonna inline something. For example, it looks at how big the function being inlined actually is, and how much inlining has already happened in the current function. Which means — and this is the unhinged part — seemingly insignificant changes in your source program can make -Winline warnings appear or ✨vanish✨ out of nowhere. Iconic chaos fr fr.

• -Winterference-size — warns you about use of the C++17 std::hardware_destructive_interference_size when you didn't specify its value with --param destructive-interference-size. It also side-eyes questionable values for that option. 🫶

So this variable exists to control class layout, specifically to avoid false sharing in concurrent code:

struct independent_fields {
  alignas(std::hardware_destructive_interference_size)
    std::atomic<int> one;
  alignas(std::hardware_destructive_interference_size)
    std::atomic<int> two;
};

Here one and two are meant to be far enough apart that storing to one won't force a reload of the cache line just to touch two. That's the whole point, no cap.

By default, --param destructive-interference-size and --param constructive-interference-size are set based on the current -mtune option — typically the L1 cache line size for your particular target CPU, and sometimes a range if you're tuning for a generic target. So every single translation unit that depends on ABI compatibility for these variables must be compiled with the same -mtune (or -mcpu). Do not play with this. 😭

If ABI stability matters to you — like if this is being used in a header for a library — you should probably just not use the hardware interference size variables at all. Alternatively, you can force a particular value with --param.

And if you're confident your use of the variable doesn't affect ABI outside a single build of your project, you can just turn off the warning with -Wno-interference-size. Easy 💅

• -Wint-in-bool-context — warns on suspicious use of integer values where booleans were expected. Like conditional expressions (?:) using non-boolean integer constants in a boolean context, e.g. if (a <= b ? 2 : 3). Or left shifting signed integers in a boolean context, like for (a = 0; 1 << a; a++);. Same energy for all kinds of multiplications regardless of data type. This one is enabled by -Wall. 👑
• -Wno-int-to-pointer-cast — suppresses warnings from casts to pointer type of an integer of a different size. And heads up: in C++, casting to a pointer type of smaller size is an error, not a warning. Wint-to-pointer-cast is enabled by default. ⭐
• -Wno-pointer-to-int-cast (C and Objective-C only) — suppresses warnings from casts from a pointer to an integer type of a different size.
• -Winvalid-pch — warns you if a precompiled header is found in the search path but can't actually be used. It's giving "found it but it's useless" 💀
• -Winvalid-utf8 — warns if an invalid UTF-8 character is found. This warning is on by default for C++23 if -finput-charset=UTF-8 is used, and gets turned into an error with -pedantic-errors.
• -Wno-unicode — don't diagnose invalid forms of delimited or named escape sequences which are treated as separate tokens. Wunicode is enabled by default. 🌈
• -Wlong-long — warns if the long long type is used. This is enabled by either -Wpedantic or -Wtraditional in ISO C90 and C++98 modes. To inhibit the warning messages, use -Wno-long-long.

This warning is upgraded to an error by -pedantic-errors. 😳

• -Wvariadic-macros — warns if variadic macros are used in ISO C90 mode, or if the GNU alternate syntax is used in ISO C99 mode. Enabled by either -Wpedantic or -Wtraditional. To inhibit the warning messages, use -Wno-variadic-macros.
• -Wno-varargs — do not warn upon questionable usage of the macros used to handle variable arguments like va_start. These warnings are enabled by default. 🙈
• -Wvector-operation-performance — warns if a vector operation is not implemented via the SIMD capabilities of the architecture. Mainly useful for performance tuning, obvi. So a vector operation can be implemented:
  • "piecewise" — the scalar operation is performed on every vector element
  • "in parallel" — the vector operation is implemented using scalars of wider type, which normally is more performance efficient
  • "as a single scalar" — the vector fits into a scalar type
• -Wvla — warns if a variable-length array is used in the code. -Wno-vla prevents the -Wpedantic warning of the variable-length array.

This warning is upgraded to an error by -pedantic-errors. 🚨

• -Wvla-larger-than=byte-size — if this option is used, the compiler warns for declarations of variable-length arrays whose size is either unbounded, or bounded by an argument that allows the array size to exceed byte-size bytes. Similar to how -Walloca-larger-than=byte-size works, but with variable-length arrays. 🦋

Ngl, note that GCC may optimize small variable-length arrays of a known value into plain arrays, so this warning may not get triggered for such arrays. Lowkey sneaky.

-Wvla-larger-than=PTRDIFF_MAX is enabled by default but is typically only effective when -ftree-vrp is active (default for -O2 and above).

See also -Walloca-larger-than=byte-size. 💜

• -Wno-vla-larger-than — disables -Wvla-larger-than= warnings. The option is equivalent to -Wvla-larger-than=SIZE_MAX or larger.
• -Wvla-parameter — warns about redeclarations of functions involving arguments of Variable Length Array types of inconsistent kinds or forms, and enables the detection of out-of-bounds accesses to such parameters by warnings such as -Warray-bounds. 👑

So if the first function declaration uses the VLA form, the bound specified in the array is assumed to be the minimum number of elements expected to be provided in calls to the function and the maximum number of elements accessed by it. Failing to provide arguments of sufficient size, or accessing more than the maximum number of elements, may be diagnosed. Periodt.

For example, the warning triggers for the following redeclarations because the first one allows an array of any size to be passed to f while the second one specifies that the array argument must have at least n elements. In addition, calling f with the associated VLA bound parameter in excess of the actual VLA bound triggers a warning as well.

void f (int n, int[n]);
// warning: argument 2 previously declared as a VLA
void f (int, int[]);

void g (int n)
{
    if (n > 4)
      return;
    int a[n];
    // warning: access to a by f may be out of bounds
    f (sizeof a, a);
  ...
}

-Wvla-parameter is included in -Wall. And the -Warray-parameter option triggers warnings for similar problems involving ordinary array arguments. 💖

• -Wvolatile-register-var — warns if a register variable is declared volatile. The volatile modifier does not inhibit all optimizations that may eliminate reads and/or writes to register variables. This warning is enabled by -Wall. 🌸
• -Wno-xor-used-as-pow (C, C++, Objective-C and Objective-C++ only) — disables warnings about uses of ^, the exclusive or operator, where it appears the code meant exponentiation. Specifically, the warning occurs when the left-hand side is the decimal constant 2 or 10 and the right-hand side is also a decimal constant. 💀

In C and C++, ^ means exclusive or — whereas in some other languages (e.g. TeX and some versions of BASIC) it means exponentiation. So the math girlies get confused, it's understandable 🥲

This warning can be silenced by converting one of the operands to hexadecimal, as well as by compiling with -Wno-xor-used-as-pow. Two ways, slay.

• -Wdisabled-optimization — warns if a requested optimization pass is disabled. This warning does not generally mean there's anything wrong with your code, bestie — it just means GCC's optimizers can't handle the code effectively. Often the problem is that your code is too big or too complex, and GCC refuses to optimize programs when the optimization itself is likely to take inordinate amounts of time. Valid honestly. ⏳
• -Wpointer-sign (C and Objective-C only) — warns for pointer argument passing or assignment with different signedness. This option is only supported for C and Objective-C. It is implied by -Wall and by -Wpedantic, which can be disabled with -Wno-pointer-sign.

This warning is upgraded to an error by -pedantic-errors. 🔥

• -Wstack-protector — this option is only active when -fstack-protector is active. It warns about functions that are not protected against stack smashing. 🛡️
• -Woverlength-strings — warns about string constants that are longer than the "minimum maximum" length specified in the C standard. Modern compilers generally allow string constants that are much longer than the standard's minimum limit, but very portable programs should avoid using longer strings. ✨

The limit applies after string constant concatenation, and does not count the trailing NUL. In C90, the limit was 509 characters; in C99, it was raised to 4095. C++98 does not specify a normative minimum maximum, so we do not diagnose overlength strings in C++. Sorry C++ girlies 💅

This option is implied by -Wpedantic, and can be disabled with -Wno-overlength-strings.

• -Wunsuffixed-float-constants (C and Objective-C only) — issues a warning for any floating constant that does not have a suffix. When used together with -Wsystem-headers, it warns about such constants in system header files too. This can be useful when preparing code to use with the FLOAT_CONST_DECIMAL64 pragma from the decimal floating-point extension to C99. 🎀
• -Wno-lto-type-mismatch — during link-time optimization, do not warn about type mismatches in global declarations from different compilation units. Requires -flto to be enabled. Enabled by default. 💜
• -Wno-designated-init (C and Objective-C only) — suppresses warnings when a positional initializer is used to initialize a structure that has been marked with the designated_init attribute. 🦋

---

✨ Options That Control Static Analysis ✨

• -fanalyzer — bestie. this one turns on a whole static analysis of program flow. it goes hunting for "interesting" interprocedural paths through your code, and when it finds problems lurking along them, it yells (aka issues warnings). 💅

ok but real talk: this analysis is much more expensive than the other GCC warnings. like, it's giving "takes forever" energy. 😭

technically? it performs coverage-guided symbolic execution of the code being compiled. and ngl, it is neither sound nor complete — it can have false positives and false negatives. it's a bug-finding tool, not a tool for proving your program is correct. it's the friend who says "hey that looks sketchy" not the one who signs off on your math homework. 🦋

the analyzer is only suitable for use on C code in this release. no cap.

turning this option on effectively enables all of these warnings:

-Wanalyzer-allocation-size -Wanalyzer-deref-before-check
-Wanalyzer-double-fclose -Wanalyzer-double-free
-Wanalyzer-exposure-through-output-file
-Wanalyzer-exposure-through-uninit-copy
-Wanalyzer-fd-access-mode-mismatch -Wanalyzer-fd-double-close
-Wanalyzer-fd-leak -Wanalyzer-fd-phase-mismatch
-Wanalyzer-fd-type-mismatch -Wanalyzer-fd-use-after-close
-Wanalyzer-fd-use-without-check -Wanalyzer-file-leak
-Wanalyzer-free-of-non-heap -Wanalyzer-imprecise-fp-arithmetic
-Wanalyzer-infinite-loop -Wanalyzer-infinite-recursion
-Wanalyzer-jump-through-null -Wanalyzer-malloc-leak
-Wanalyzer-mismatching-deallocation -Wanalyzer-null-argument
-Wanalyzer-null-dereference -Wanalyzer-out-of-bounds
-Wanalyzer-overlapping-buffers -Wanalyzer-possible-null-argument
-Wanalyzer-possible-null-dereference -Wanalyzer-putenv-of-auto-var
-Wanalyzer-shift-count-negative -Wanalyzer-shift-count-overflow
-Wanalyzer-stale-setjmp-buffer -Wanalyzer-tainted-allocation-size
-Wanalyzer-tainted-array-index -Wanalyzer-tainted-assertion
-Wanalyzer-tainted-divisor -Wanalyzer-tainted-offset
-Wanalyzer-tainted-size -Wanalyzer-undefined-behavior-ptrdiff
-Wanalyzer-undefined-behavior-strtok
-Wanalyzer-unsafe-call-within-signal-handler
-Wanalyzer-use-after-free
-Wanalyzer-use-of-pointer-in-stale-stack-frame
-Wanalyzer-use-of-uninitialized-value
-Wanalyzer-va-arg-type-mismatch -Wanalyzer-va-list-exhausted
-Wanalyzer-va-list-leak -Wanalyzer-va-list-use-after-va-end
-Wanalyzer-write-to-const -Wanalyzer-write-to-string-literal

heads up: this option is only available if GCC was configured with analyzer support enabled. if your build didn't opt in, it's just not there, sorry bestie. 🎀

• -Wanalyzer-symbol-too-complex — so if -fanalyzer is on, the analyzer uses a bunch of heuristics to try to track the state of memory, but sufficiently complicated code can defeat them. 🙈

by default the analysis just silently stops tracking the values of expressions once they exceed the threshold defined by --param analyzer-max-svalue-depth=value, and falls back to an imprecise representation for those expressions. so like, it gives up quietly. 💀

the -Wanalyzer-symbol-too-complex option warns you if this occurs. it's the "hey, i stopped tracking this" notification. 🔥

• -Wanalyzer-too-complex — same deal but for control flow + data flow. if -fanalyzer is enabled, the analyzer uses various heuristics to attempt to explore the control flow and data flow in the program, but complicated code can defeat those too.

by default, the analysis silently stops if the code is too complicated for the analyzer to fully explore and it reaches an internal limit. -Wanalyzer-too-complex warns if this occurs. internally screaming, externally silent, until you turn this on. 😭

• -Wno-analyzer-allocation-size — this warning requires -fanalyzer, which enables it; to disable it, use -Wno-analyzer-allocation-size.

this diagnostic warns for paths through the code in which a pointer to a buffer is assigned to point at a buffer with a size that is not a multiple of sizeof (*pointer).

See CWE-131: Incorrect Calculation of Buffer Size (https://cwe.mitre.org/data/definitions/131.html).

• -Wno-analyzer-deref-before-check — this warning requires -fanalyzer, which enables it; use -Wno-analyzer-deref-before-check to disable it.

this diagnostic warns for paths through the code in which a pointer is checked for "NULL" after it has already been dereferenced — which is giving "the pointer could have been NULL the whole time" energy. such cases suggest that the check for NULL is either redundant, or that it needs to be moved to before the pointer is dereferenced. 🙈

this diagnostic also considers values passed to a function argument marked with "__attribute__((nonnull))" as requiring a non-NULL value, and thus will complain if such values are checked for "NULL" after returning from such a function call.

lowkey important: this diagnostic is unlikely to be reported when any level of optimization is enabled, because GCC's optimization logic will typically consider such checks for NULL as being redundant and optimize them away before the analyzer "sees" them. hence optimization should be disabled when attempting to trigger this diagnostic. periodt.

• -Wno-analyzer-double-fclose — this warning requires -fanalyzer, which enables it; use -Wno-analyzer-double-fclose to disable it.

this diagnostic warns for paths through the code in which a "FILE *" can have "fclose" called on it more than once. 💀

See CWE-1341: Multiple Releases of Same Resource or Handle (https://cwe.mitre.org/data/definitions/1341.html).

• -Wno-analyzer-double-free — this warning requires -fanalyzer, which enables it; use -Wno-analyzer-double-free to disable it.

this diagnostic warns for paths through the code in which a pointer can have a deallocator called on it more than once — either "free", or a deallocator referenced by attribute "malloc".

See CWE-415: Double Free (https://cwe.mitre.org/data/definitions/415.html).

• -Wno-analyzer-exposure-through-output-file — this warning requires -fanalyzer, which enables it; use -Wno-analyzer-exposure-through-output-file to disable it.

this diagnostic warns for paths through the code in which a security-sensitive value is written to an output file — like, say, writing a password to a log file. bestie no. 😭

See CWE-532: Information Exposure Through Log Files (https://cwe.mitre.org/data/definitions/532.html).

• -Wanalyzer-exposure-through-uninit-copy — this warning requires both -fanalyzer and the use of a plugin to specify a function that copies across a "trust boundary". use -Wno-analyzer-exposure-through-uninit-copy to disable it.

this diagnostic warns for "infoleaks" — paths through the code in which uninitialized values are copied across a security boundary. the classic example: code within an OS kernel that copies a partially-initialized struct on the stack to user space. 🙈

See CWE-200: Exposure of Sensitive Information to an Unauthorized Actor (https://cwe.mitre.org/data/definitions/200.html).

• -Wno-analyzer-fd-access-mode-mismatch — this warning requires -fanalyzer, which enables it; use -Wno-analyzer-fd-access-mode-mismatch to disable it.

this diagnostic warns for paths through code in which a "read" on a write-only file descriptor is attempted, or vice versa.

this diagnostic also warns for code paths in a which a function with attribute "fd_arg_read (N)" is called with a file descriptor opened with "O_WRONLY" at referenced argument "N", or a function with attribute "fd_arg_write (N)" is called with a file descriptor opened with "O_RDONLY" at referenced argument N. 💅

• -Wno-analyzer-fd-double-close — this warning requires -fanalyzer, which enables it; use -Wno-analyzer-fd-double-close to disable it.

this diagnostic warns for paths through code in which a file descriptor can be closed more than once.

See CWE-1341: Multiple Releases of Same Resource or Handle (https://cwe.mitre.org/data/definitions/1341.html).

• -Wno-analyzer-fd-leak — this warning requires -fanalyzer, which enables it; use -Wno-analyzer-fd-leak to disable it.

this diagnostic warns for paths through code in which an open file descriptor is leaked. rip to those fds. 🕯️

See CWE-775: Missing Release of File Descriptor or Handle after Effective Lifetime (https://cwe.mitre.org/data/definitions/775.html).

• -Wno-analyzer-fd-phase-mismatch — this warning requires -fanalyzer, which enables it; use -Wno-analyzer-fd-phase-mismatch to disable it.

this diagnostic warns for paths through code in which an operation is attempted in the wrong phase of a file descriptor's lifetime. for example, it will warn on attempts to call "accept" on a stream socket that has not yet had "listen" successfully called on it. it's giving "you skipped a step, bestie". ✨

See CWE-666: Operation on Resource in Wrong Phase of Lifetime (https://cwe.mitre.org/data/definitions/666.html).

• -Wno-analyzer-fd-type-mismatch — this warning requires -fanalyzer, which enables it; use -Wno-analyzer-fd-type-mismatch to disable it.

this diagnostic warns for paths through code in which an operation is attempted on the wrong type of file descriptor. for example, it will warn on attempts to use socket operations on a file descriptor obtained via "open", or when attempting to use a stream socket operation on a datagram socket. 🦋

• -Wno-analyzer-fd-use-after-close — this warning requires -fanalyzer, which enables it; use -Wno-analyzer-fd-use-after-close to disable it.

this diagnostic warns for paths through code in which a read or write is called on a closed file descriptor. 💀

this diagnostic also warns for paths through code in which a function with attribute "fd_arg (N)" or "fd_arg_read (N)" or "fd_arg_write (N)" is called with a closed file descriptor at referenced argument "N".

• -Wno-analyzer-fd-use-without-check — this warning requires -fanalyzer, which enables it; use -Wno-analyzer-fd-use-without-check to disable it.

this diagnostic warns for paths through code in which a file descriptor is used without being checked for validity.

this diagnostic also warns for paths through code in which a function with attribute "fd_arg (N)" or "fd_arg_read (N)" or "fd_arg_write (N)" is called with a file descriptor, at referenced argument "N", without being checked for validity. like… vEt it first. 😭

• -Wno-analyzer-file-leak — this warning requires -fanalyzer, which enables it; use -Wno-analyzer-file-leak to disable it.

this diagnostic warns for paths through the code in which a "<stdio.h>" "FILE *" stream object is leaked.

See CWE-775: Missing Release of File Descriptor or Handle after Effective Lifetime (https://cwe.mitre.org/data/definitions/775.html).

• -Wno-analyzer-free-of-non-heap — ngl bestie, this one rides along with -fanalyzer (that's what turns it on), and you silence it with -Wno-analyzer-free-of-non-heap. It clocks paths through your code where free gets called on a pointer that did NOT come from the heap — think an on-stack buffer, or a global. Yeah. Chaos. 💀

CWE said it best, so: CWE-590: Free of Memory not on the Heap ("https://cwe.mitre.org/data/definitions/590.html").

• -Wno-analyzer-imprecise-fp-arithmetic — requires -fanalyzer (which is the whole reason it's alive), and -Wno-analyzer-imprecise-fp-arithmetic is your off switch. It flags paths where floating-point arithmetic shows up in spots that genuinely need precise computation. As of right now it only warns when floating-point operands are used inside the calculation of an allocation size. That's it, that's the whole scope. 🎀
• -Wno-analyzer-infinite-loop — needs -fanalyzer to be enabled; kill it with -Wno-analyzer-infinite-loop. This diagnostic clocks paths that look like they're headed straight into an infinite loop, and the analyzer is picky about when it speaks up. Specifically, it drops this warning when it "sees" a loop where:
  • no externally-visible work could be being done inside the loop
  • there's literally no way to escape the loop
  • the analyzer is confident enough about the program state across the whole loop to know the two things above are actually true

One classic way this fires: there's an execution path through a loop where taking that path on one iteration means the exact same path gets taken on every iteration after. Like, forever. 🌀

For example, consider:

while (1)
  {
    char opcode = *cpu_state.pc;
    switch (opcode)
     {
     case OPCODE_FOO:
       handle_opcode_foo (&cpu_state);
       break;
     case OPCODE_BAR:
       handle_opcode_bar (&cpu_state);
       break;
     }
  }

Why is the analyzer mad at this? Because if opcode ever matches none of the cases, the switch falls through to the implicit default case, which makes the loop body a total "no-op" with cpu_state.pc never changing — so the same value of opcode comes back around on all subsequent iterations, and boom, infinite loop. 😭

See CWE-835: Loop with Unreachable Exit Condition ('Infinite Loop') ("https://cwe.mitre.org/data/definitions/835.html").

• -Wno-analyzer-infinite-recursion — yep, -fanalyzer enables it, and -Wno-analyzer-infinite-recursion disables it. This one spots paths that appear to lead to infinite recursion. Here's the tea: when the analyzer "sees" a recursive call, it compares the state of memory at the entry to the new frame against the state at the entry to the previous frame of that function on the stack. If nothing in memory appears to be changing, the warning fires. Any changes it observes to parameters or globals are assumed to lead to termination of the recursion, so those suppress the warning. 🤔

Heads up though — this diagnostic is likely to miss cases of infinite recursion that get converted to iteration by the optimizer before the analyzer ever "sees" them. So if you're trying to trigger it, disable optimization. Periodt.

Compare with -Winfinite-recursion, which gives you a similar diagnostic but is implemented in a completely different way.

See CWE-674: Uncontrolled Recursion ("https://cwe.mitre.org/data/definitions/674.html").

• -Wno-analyzer-jump-through-null — needs -fanalyzer, off with -Wno-analyzer-jump-through-null. It flags paths where a NULL function pointer gets called. That's the vibe, and it's a bad vibe. 🙈
• -Wno-analyzer-malloc-leak — -fanalyzer enables it; -Wno-analyzer-malloc-leak disables it. It catches paths where a pointer allocated via an allocator gets leaked — either malloc, or any function marked with attribute malloc.

See CWE-401: Missing Release of Memory after Effective Lifetime ("https://cwe.mitre.org/data/definitions/401.html").

• -Wno-analyzer-mismatching-deallocation — -fanalyzer enables it, -Wno-analyzer-mismatching-deallocation turns it off. This one flags paths where the wrong deallocation function is called on a pointer value, judging by which function allocated that pointer in the first place. It'll call out mismatches between free, scalar delete and vector delete[], plus anything marked as allocator/deallocator pairs using attribute malloc.

See CWE-762: Mismatched Memory Management Routines ("https://cwe.mitre.org/data/definitions/762.html").

• -Wno-analyzer-out-of-bounds — -fanalyzer enables it; -Wno-analyzer-out-of-bounds disables it. It warns on paths where a buffer is definitely read or written out-of-bounds. It applies when the analyzer can work out a constant offset and for accesses past the end of a buffer with a constant capacity too. And lowkey? It also does limited checking for accesses past the end when both the offset and the capacity are symbolic. 👑

See CWE-119: Improper Restriction of Operations within the Bounds of a Memory Buffer ("https://cwe.mitre.org/data/definitions/119.html").

Oh and when the analyzer can manage it, it'll emit a literal text art diagram showing the spatial relationship between the memory region it predicts would be accessed and the range of memory that's valid to access: whether they overlap, are touching, are close or far apart; which one comes before or after in memory; the relative sizes involved; the direction of the access (read vs write); and in some cases the actual values of data involved. Iconic behaviour honestly. You can suppress the diagram with -fdiagnostics-text-art-charset=none.

• -Wno-analyzer-overlapping-buffers — requires -fanalyzer, and -Wno-analyzer-overlapping-buffers disables it. It flags paths where overlapping buffers get passed to an API whose behaviour on overlapping buffers is undefined. 🫠

Specifically, the diagnostic fires on calls to these functions:

• "memcpy"
• "strcat"
• "strcpy"

in cases where the buffers are known to overlap.

• -Wno-analyzer-possible-null-argument — needs -fanalyzer; -Wno-analyzer-possible-null-argument is the off switch. It warns on paths where a possibly-NULL value is passed to a function argument marked with __attribute__((nonnull)) as requiring a non-NULL value.

See CWE-690: Unchecked Return Value to NULL Pointer Dereference ("https://cwe.mitre.org/data/definitions/690.html").

• -Wno-analyzer-possible-null-dereference — requires -fanalyzer; disable with -Wno-analyzer-possible-null-dereference. It flags paths where a possibly-NULL value gets dereferenced. 😬

See CWE-690: Unchecked Return Value to NULL Pointer Dereference ("https://cwe.mitre.org/data/definitions/690.html").

• -Wno-analyzer-null-argument — -fanalyzer enables it, -Wno-analyzer-null-argument disables it. It warns on paths where a value known to be NULL is passed to a function argument marked with __attribute__((nonnull)) as requiring a non-NULL value.

See CWE-476: NULL Pointer Dereference ("https://cwe.mitre.org/data/definitions/476.html").

• -Wno-analyzer-null-dereference — needs -fanalyzer; -Wno-analyzer-null-dereference shuts it up. It flags paths where a value known to be NULL gets dereferenced. 💀

See CWE-476: NULL Pointer Dereference ("https://cwe.mitre.org/data/definitions/476.html").

• -Wno-analyzer-putenv-of-auto-var — -fanalyzer enables it, and -Wno-analyzer-putenv-of-auto-var disables it. This one warns on paths where a call to putenv is handed a pointer to an automatic variable or an on-stack buffer.

See POS34-C. Do not call putenv() with a pointer to an automatic variable as the argument ("https://wiki.sei.cmu.edu/confluence/x/6NYxBQ").

• -Wno-analyzer-shift-count-negative — requires -fanalyzer; turn it off with -Wno-analyzer-shift-count-negative. It warns on paths where a shift is attempted with a negative count. It's analogous to the -Wshift-count-negative diagnostic implemented in the C/C++ front ends, except this one is implemented by analyzing interprocedural paths instead of just parsing the syntax tree. One caveat though: the analyzer doesn't prioritize detection of such paths, so false negatives are more likely relative to other warnings. 🌸
• -Wno-analyzer-shift-count-overflow — -fanalyzer enables it, -Wno-analyzer-shift-count-overflow disables it. It warns on paths where a shift is attempted with a count greater than or equal to the precision of the operand's type. Same deal as above: analogous to the -Wshift-count-overflow diagnostic in the C/C++ front ends, but implemented via interprocedural path analysis rather than just parsing the syntax tree. And again, the analyzer doesn't prioritize detecting these paths, so false negatives are more likely compared to other warnings. 🦋
• -Wno-analyzer-stale-setjmp-buffer — needs -fanalyzer; -Wno-analyzer-stale-setjmp-buffer disables it. It flags paths where longjmp is called to rewind to a jmp_buf relating to a setjmp call in a function that has already returned. Stale is not the moment bestie. 💅

When setjmp is called on a jmp_buf to jot down a rewind spot, it records the stack frame. And bestie, that stack frame becomes invalid the second the function containing the setjmp call returns. So if you try to rewind back to it with longjmp? You're pointing at a stack frame that literally does not exist anymore, and that's how you get a crash (or worse). 💀

---

• -Wno-analyzer-tainted-allocation-size — needs -fanalyzer to turn it on in the first place; use -Wno-analyzer-tainted-allocation-size to switch it off.

This one flags paths through your code where a value that could be under an attacker's control gets used as the size of an allocation without being sanitized. Which means: attacker injects a comically huge allocation → potential denial of service. Not the vibe. 😭

See CWE-789: Memory Allocation with Excessive Size Value (https://cwe.mitre.org/data/definitions/789.html).

• -Wno-analyzer-tainted-assertion — needs -fanalyzer to enable it; use -Wno-analyzer-tainted-assertion to disable it.

This diagnostic calls out paths where an attacker-controllable value is used in a condition without being sanitized first, and that condition guards a call to a function marked with the noreturn attribute (like __builtin_unreachable). Those functions usually mean the program is terminating abnormally, like an assertion-failure handler. Example:

assert (some_tainted_value < SOME_LIMIT);

In cases like that:

• when assertion-checking is enabled: an attacker could trigger a denial of service by injecting an assertion failure
• when assertion-checking is disabled, like by defining NDEBUG: an attacker could inject data that subverts the process, since it presumably violates a precondition the code is assuming is true.

Heads up: when assertion-checking is off, the assertions usually get removed by the preprocessor before the analyzer ever has a chance to "see" them, so this diagnostic can only warn on builds where assertion-checking is enabled.

For this warning's purposes, any function marked with the noreturn attribute counts as a possible assertion-failure handler, including __builtin_unreachable. And note that these functions are sometimes removed by the optimizer before the analyzer "sees" them — so optimization should be disabled if you're trying to trigger this diagnostic.

See CWE-617: Reachable Assertion (https://cwe.mitre.org/data/definitions/617.html).

It can also report spicy constructions like:

switch (some_tainted_value) {
case 0:
  /* [...etc; various valid cases omitted...] */
  break;

default:
  __builtin_unreachable (); /* BUG: attacker can trigger this  */
}

even though, strictly speaking, that's not an assertion failure.

• -Wno-analyzer-tainted-array-index — needs -fanalyzer to enable it; use -Wno-analyzer-tainted-array-index to disable it.

Flags paths where an attacker-controllable value is used as the index of an array access without being sanitized, so an attacker could inject an out-of-bounds access. 🫣

See CWE-129: Improper Validation of Array Index (https://cwe.mitre.org/data/definitions/129.html).

• -Wno-analyzer-tainted-divisor — needs -fanalyzer to enable it; use -Wno-analyzer-tainted-divisor to disable it.

Flags paths where an attacker-controllable value gets used as the divisor in a division or modulus operation without being sanitized — so an attacker can inject a division-by-zero.

See CWE-369: Divide By Zero (https://cwe.mitre.org/data/definitions/369.html).

• -Wno-analyzer-tainted-offset — needs -fanalyzer to enable it; use -Wno-analyzer-tainted-offset to disable it.

Flags paths where an attacker-controllable value is used as a pointer offset without being sanitized, so an attacker could inject an out-of-bounds access.

See CWE-823: Use of Out-of-range Pointer Offset (https://cwe.mitre.org/data/definitions/823.html).

• -Wno-analyzer-tainted-size — needs -fanalyzer to enable it; use -Wno-analyzer-tainted-size to disable it.

Flags paths where an attacker-controllable value is used as the size of an operation like memset without being sanitized, so an attacker could inject an out-of-bounds access. 🍓

See CWE-129: Improper Validation of Array Index (https://cwe.mitre.org/data/definitions/129.html).

• -Wno-analyzer-undefined-behavior-ptrdiff — needs -fanalyzer, which enables it; use -Wno-analyzer-undefined-behavior-ptrdiff to disable it.

Flags paths where a pointer subtraction happens but the pointers refer to different chunks of memory. That code is leaning on undefined behavior, because pointer subtraction is only defined when both pointers point within (or just after) the same array.

See CWE-469: Use of Pointer Subtraction to Determine Size (https://cwe.mitre.org/data/definitions/469.html).

• -Wno-analyzer-undefined-behavior-strtok — needs -fanalyzer, which enables it; use -Wno-analyzer-undefined-behavior-strtok to disable it.

Flags paths where a call to strtok has undefined behavior.

Specifically: passing NULL as the first parameter for the initial call to strtok within a process has undefined behavior.

• -Wno-analyzer-unsafe-call-within-signal-handler — needs -fanalyzer, which enables it; use -Wno-analyzer-unsafe-call-within-signal-handler to disable it.

Flags paths where a function known to be async-signal-unsafe (like fprintf) is called from a signal handler. 💅

See CWE-479: Signal Handler Use of a Non-reentrant Function (https://cwe.mitre.org/data/definitions/479.html).

• -Wno-analyzer-use-after-free — needs -fanalyzer, which enables it; use -Wno-analyzer-use-after-free to disable it.

Flags paths where a pointer is used after a deallocator was called on it — either free, or a deallocator referenced by the malloc attribute.

See CWE-416: Use After Free (https://cwe.mitre.org/data/definitions/416.html).

• -Wno-analyzer-use-of-pointer-in-stale-stack-frame — needs -fanalyzer, which enables it; use -Wno-analyzer-use-of-pointer-in-stale-stack-frame to disable it.

Flags paths where a pointer is dereferenced that points to a variable in a stale stack frame. 😭

• -Wno-analyzer-va-arg-type-mismatch — needs -fanalyzer, which enables it; use -Wno-analyzer-va-arg-type-mismatch to disable it.

Flags interprocedural paths where the analyzer catches an attempt to use va_arg to extract a value passed to a variadic call, but the type used does not match the type of the expression passed to the call.

See CWE-686: Function Call With Incorrect Argument Type (https://cwe.mitre.org/data/definitions/686.html).

• -Wno-analyzer-va-list-exhausted — needs -fanalyzer, which enables it; use -Wno-analyzer-va-list-exhausted to disable it.

Flags interprocedural paths where the analyzer catches an attempt to use va_arg to grab the next value passed to a variadic call, but every value in the va_list has already been consumed. 🙈

See CWE-685: Function Call With Incorrect Number of Arguments (https://cwe.mitre.org/data/definitions/685.html).

• -Wno-analyzer-va-list-leak — needs -fanalyzer, which enables it; use -Wno-analyzer-va-list-leak to disable it.

Flags interprocedural paths where the analyzer catches that va_start or va_copy was called on a va_list with no corresponding call to va_end.

• -Wno-analyzer-va-list-use-after-va-end — needs -fanalyzer, which enables it; use -Wno-analyzer-va-list-use-after-va-end to disable it.

Flags interprocedural paths where the analyzer catches an attempt to use a va_list after va_end has been called on it. va_list.

• -Wno-analyzer-write-to-const — needs -fanalyzer, which enables it; use -Wno-analyzer-write-to-const to disable it.

Flags paths where the analyzer catches an attempt to write through a pointer to a const object. That said, the analyzer doesn't prioritize detecting these paths, so false negatives are more likely here compared to other warnings. 🤷

• -Wno-analyzer-write-to-string-literal — needs -fanalyzer, which enables it; use -Wno-analyzer-write-to-string-literal to disable it.

Flags paths where the analyzer catches an attempt to write through a pointer to a string literal. Same deal as above though — the analyzer doesn't prioritize detecting these paths, so false negatives are more likely relative to other warnings.

• -Wno-analyzer-use-of-uninitialized-value — needs -fanalyzer, which enables it; use -Wno-analyzer-use-of-uninitialized-value to disable it.

Flags paths where an uninitialized value is used. 🎀

See CWE-457: Use of Uninitialized Variable (https://cwe.mitre.org/data/definitions/457.html).

---

The analyzer has hardcoded knowledge about the behavior of the following memory-management functions:

• "alloca" 🎀
• the built-in functions "__builtin_alloc", "__builtin_alloc_with_align", "__builtin_calloc", "__builtin_free", "__builtin_malloc", "__builtin_memcpy", "__builtin_memcpy_chk", "__builtin_memset", "__builtin_memset_chk", "__builtin_realloc", "__builtin_stack_restore", and "__builtin_stack_save" ✨
• "calloc"
• "free"
• "malloc"
• "memset"
• "operator delete"
• "operator delete []"
• "operator new"
• "operator new []"
• "realloc"
• "strdup"
• "strndup"

of the following functions for working with file descriptors: 🦋

• "open"
• "close"
• "creat"
• "dup", "dup2" and "dup3"
• "isatty"
• "pipe", and "pipe2"
• "read"
• "write"
• "socket", "bind", "listen", "accept", and "connect"

of the following functions for working with "<stdio.h>" streams: 🍓

• the built-in functions "__builtin_fprintf", "__builtin_fprintf_unlocked", "__builtin_fputc", "__builtin_fputc_unlocked", "__builtin_fputs", "__builtin_fputs_unlocked", "__builtin_fwrite", "__builtin_fwrite_unlocked", "__builtin_printf", "__builtin_printf_unlocked", "__builtin_putc", "__builtin_putchar", "__builtin_putchar_unlocked", "__builtin_putc_unlocked", "__builtin_puts", "__builtin_puts_unlocked", "__builtin_vfprintf", and "__builtin_vprintf" 💖
• "fopen"
• "fclose"
• "ferror"
• "fgets"
• "fgets_unlocked"
• "fileno"
• "fread"
• "getc"
• "getchar"
• "fprintf"
• "printf"
• "fwrite"

and of the following functions: 🌸

• the built-in functions "__builtin_expect", "__builtin_expect_with_probability", "__builtin_strchr", "__builtin_strcpy", "__builtin_strcpy_chk", "__builtin_strlen", "__builtin_va_copy", and "__builtin_va_start" ⭐
• the GNU extensions "error" and "error_at_line"
• "getpass"
• "longjmp"
• "putenv"
• "setjmp"
• "siglongjmp"
• "signal"
• "sigsetjmp"
• "strcat"
• "strchr"
• "strlen"

on top of all that, a whole bunch of functions rocking an "__analyzer_" prefix have special meaning to the analyzer, and bestie, that's all documented over in the GCC Internals manual. 💅

The parameters that actually matter for steering how much the analyzer explores are: 🔍

• --param analyzer-bb-explosion-factor=value
• --param analyzer-max-enodes-per-program-point=value
• --param analyzer-max-recursion-depth=value
• --param analyzer-min-snodes-for-call-summary=value

The following options control the analyzer.

• -fanalyzer-call-summaries — simplifies the interprocedural analysis by computing the effect of certain calls, instead of exploring every single path through the function from the callsite out to each possible return. Lowkey a glow-up for your compile times.

If enabled, call summaries are only used for functions with more than one call site, and that are sufficiently complicated (as per --param analyzer-min-snodes-for-call-summary=value).

• -fanalyzer-checker=name — restricts the analyzer to run just the named checker, and enables it. One checker, one vibe, no distractions. 💅
• -fanalyzer-debug-text-art-headings — this option is intended for analyzer developers. If enabled, the analyzer adds extra annotations to any diagrams it generates. ✨ (yes the diagrams get little headings, it's giving arts-and-crafts)
• -fno-analyzer-feasibility — this option is intended for analyzer developers.

By default the analyzer verifies that there's a feasible control flow path for each diagnostic it emits: that the conditions that hold are not mutually exclusive. Diagnostics where no feasible path can be found get rejected. You can suppress this filtering with -fno-analyzer-feasibility, for debugging issues in this code.

• -fanalyzer-fine-grained — this option is intended for analyzer developers.

Internally the analyzer builds an "exploded graph" that combines control flow graphs with data flow information.

By default, an edge in this graph can contain the effects of a run of multiple statements within a basic block. With -fanalyzer-fine-grained, each statement gets its own edge. ⚡

• -fanalyzer-show-duplicate-count — this option is intended for analyzer developers: if multiple diagnostics have been detected as being duplicates of each other, it emits a note when reporting the best diagnostic, giving the number of additional diagnostics that were suppressed by the deduplication logic. Like "btw there were 12 more of these, you're welcome" 😭
• -fanalyzer-show-events-in-system-headers — by default the analyzer emits simplified diagnostics paths by hiding events fully located within a system header. With -fanalyzer-show-events-in-system-headers such events are no longer suppressed. 🙈
• -fno-analyzer-state-merge — this option is intended for analyzer developers.

By default the analyzer attempts to simplify analysis by merging sufficiently similar states at each program point as it builds its "exploded graph". With -fno-analyzer-state-merge this merging can be suppressed, for debugging state-handling issues.

• -fno-analyzer-state-purge — this option is intended for analyzer developers.

By default the analyzer attempts to simplify analysis by purging aspects of state at a program point that appear to no longer be relevant — e.g. the values of locals that aren't accessed later in the function and which aren't relevant to leak analysis.

With -fno-analyzer-state-purge this purging of state can be suppressed, for debugging state-handling issues. 🧹🚫

• -fno-analyzer-suppress-followups — this option is intended for analyzer developers.

By default the analyzer will stop exploring an execution path after encountering certain diagnostics, in order to avoid potentially issuing a cascade of follow-up diagnostics.

The diagnostics that terminate analysis along a path are:

• -Wanalyzer-null-argument
• -Wanalyzer-null-dereference
• -Wanalyzer-use-after-free
• -Wanalyzer-use-of-pointer-in-stale-stack-frame
• -Wanalyzer-use-of-uninitialized-value

With -fno-analyzer-suppress-followups the analyzer will continue to explore such paths even after such diagnostics, which may be helpful for debugging issues in the analyzer, or for microbenchmarks for detecting undefined behavior. 💀

• -fanalyzer-transitivity — this option enables transitivity of constraints within the analyzer. 🔗
• -fno-analyzer-undo-inlining — this option is intended for analyzer developers.

-fanalyzer runs relatively late compared to other code analysis tools, and some optimizations have already been applied to the code. In particular function inlining may have occurred, leading to the interprocedural execution paths emitted by the analyzer containing function frames that don't correspond to those in the original source code. 💀

By default the analyzer attempts to reconstruct the original function frames, and to emit events showing the inlined calls.

With -fno-analyzer-undo-inlining this attempt to reconstruct the original frame information can be disabled, which may be of help when debugging issues in the analyzer.

• -fanalyzer-verbose-edges — this option is intended for analyzer developers. It enables more verbose, lower-level detail in the descriptions of control flow within diagnostic paths. 🌈
• -fanalyzer-verbose-state-changes — this option is intended for analyzer developers. It enables more verbose, lower-level detail in the descriptions of events relating to state machines within diagnostic paths. 🌈
• -fanalyzer-verbosity=level — this option controls the complexity of the control flow paths that are emitted for analyzer diagnostics.

The level can be one of:

• 0 — at this level, interprocedural call and return events are displayed, along with the most pertinent state-change events relating to a diagnostic. For example, for a double-"free" diagnostic, both calls to "free" will be shown.
• 1 — as per the previous level, but also show events for the entry to each function.
• 2 — as per the previous level, but also show events relating to control flow that are significant to triggering the issue (e.g. "true path taken" at a conditional).

This level is the default. 👑

• 3 — as per the previous level, but show all control flow events, not just significant ones.
• 4 — this level is intended for analyzer developers; it adds various other events intended for debugging the analyzer.

• -fdump-analyzer — dumps internal details about what the analyzer is doing to file.analyzer.txt. -fdump-analyzer-stderr overrides this option.
• -fdump-analyzer-stderr — dumps internal details about what the analyzer is doing to stderr. This option overrides -fdump-analyzer.
• -fdump-analyzer-callgraph — dumps a representation of the call graph suitable for viewing with GraphViz to file.callgraph.dot. 📊
• -fdump-analyzer-exploded-graph — dumps a representation of the "exploded graph" suitable for viewing with GraphViz to file.eg.dot. Nodes are color-coded based on state-machine states to emphasize state changes. 🎨
• -fdump-analyzer-exploded-nodes — emits diagnostics showing where nodes in the "exploded graph" are in relation to the program source.
• -fdump-analyzer-exploded-nodes-2 — dumps a textual representation of the "exploded graph" to file.eg.txt.
• -fdump-analyzer-exploded-nodes-3 — dumps a textual representation of the "exploded graph" to one dump file per node, to file.eg-id.txt. This is typically a large number of dump files. 😭
• -fdump-analyzer-exploded-paths — dumps a textual representation of the "exploded path" for each diagnostic to file.idx.kind.epath.txt.
• -fdump-analyzer-feasibility — dumps internal details about the analyzer's search for feasible paths. The details are written in a form suitable for viewing with GraphViz to filenames of the form file.*.fg.dot, file.*.tg.dot, and file.*.fpath.txt.
• -fdump-analyzer-infinite-loop — dumps internal details about the analyzer's search for infinite loops. The details are written in a form suitable for viewing with GraphViz to filenames of the form file.*.infinite-loop.dot. 🔁
• -fdump-analyzer-json — dumps a compressed JSON representation of analyzer internals to file.analyzer.json.gz. The precise format is subject to change. ⚠️
• -fdump-analyzer-state-purge — okay so this one is basically -fdump-analyzer-supergraph but make it ✨fancier✨: it dumps that whole "supergraph" so you can stare at it in GraphViz, except the graph gets annotated with the tea on what state is gonna get purged at each node. It writes to file.state-purge.dot. 💅
• -fdump-analyzer-supergraph — dumps representations of the "supergraph" that you can view with GraphViz, into file.supergraph.dot AND file.supergraph-eg.dot. These show every control flow graph in the program, with interprocedural edges for calls and returns, bestie. The second dump has annotations showing nodes in the "exploded graph" plus the diagnostics that go with them. 🦋
• -fdump-analyzer-untracked — emits custom warnings with internal details, and this one is literally for analyzer developers only (iykyk). 🔍

✨ Options for Debugging Your Program ✨

So if you wanna tell GCC to spit out extra info for a debugger to snack on, like, almost every single time you just add -g to your other options and you're done. Some debug formats can co-exist (like DWARF with CTF) as long as each of them gets enabled explicitly by adding its own respective command line option to your other options. 🫶

GCC lets you use -g together with -O. But heads up — the shortcuts that optimized code takes can occasionally be surprising: some variables you declared might not exist at all 😭, flow of control may briefly move somewhere you did not expect, some statements may not be executed because they compute constant results or their values are already at hand, and some statements may execute in different places because they got moved out of loops. Still, you can debug optimized output. So it makes total sense to use the optimizer on programs that might have bugs.

If you're not using some other optimization option, consider using -Og with -g. Because with no -O option at all, some compiler passes that collect info useful for debugging don't run at all — so -Og might get you a better debugging experience. 👑

• -g — produces debugging information in the operating system's native format (stabs, COFF, XCOFF, or DWARF). GDB can work with this debugging info. 💖

On most systems that use stabs format, -g enables use of extra debugging information that only GDB can use. This extra info makes debugging work better in GDB but probably makes other debuggers crash or refuse to read the program. If you want to control for certain whether to generate the extra information, use -gvms (peep below 👇).

• -ggdb — produces debugging information for use by GDB. That means it uses the most expressive format available (DWARF, stabs, or the native format if neither of those are supported), including GDB extensions if at all possible. It's giving ✨maximum expressiveness✨
• -gdwarf
• -gdwarf-version — produces debugging information in DWARF format (if that's supported). The value of version may be either 2, 3, 4 or 5; the default version for most targets is 5 (with the exception of VxWorks, TPF and Darwin / macOS, which default to version 2, and AIX, which defaults to version 4). 📊

Note that with DWARF Version 2, some ports require and always use some non-conflicting DWARF 3 extensions in the unwind tables.

Version 4 may require GDB 7.0 and -fvar-tracking-assignments for maximum benefit. Version 5 requires GDB 8.0 or higher.

GCC no longer supports DWARF Version 1, which is substantially different than Version 2 and later. For historical reasons, some other DWARF-related options such as -fno-dwarf2-cfi-asm retain a reference to DWARF Version 2 in their names, but apply to all currently-supported versions of DWARF. 💀 (the naming is a lie but it's a historical lie, we forgive her)

• -gbtf — requests BTF debug information. BTF is the default debugging format for the eBPF target. On other targets, like x86, BTF debug information can be generated along with DWARF debug information when both of the debug formats are enabled explicitly via their respective command line options. 🐝
• -gprune-btf
• -gno-prune-btf — prunes BTF information before emission. When pruning, only type information for types used by global variables and file-scope functions will be emitted. If compiling for the BPF target with BPF CO-RE enabled, type information will also be emitted for types used in BPF CO-RE relocations. On top of that, struct and union types which are only referred to via pointers from members of other struct or union types shall be pruned and replaced with BTF_KIND_FWD, as though those types were only present in the input as forward declarations.

This option substantially reduces the size of produced BTF information, but at significant loss in the amount of detailed type information. It is primarily useful when compiling for the BPF target, to minimize the size of the resulting object, and to eliminate BTF information which is not immediately relevant to the BPF program loading process. ✂️

This option is enabled by default for the BPF target when generating BTF information.

• -gctf
• -gctflevel — requests CTF debug information and uses level to specify how much CTF debug information should be produced. If -gctf is specified without a value for level, the default level of CTF debug information is 2. 🌸

CTF debug information can be generated along with DWARF debug information when both of the debug formats are enabled explicitly via their respective command line options.

Level 0 produces no CTF debug information at all. So -gctf0 negates -gctf. 🚫

Level 1 produces CTF information for tracebacks only. This includes callsite information, but does not include type information.

Level 2 produces type information for entities (functions, data objects etc.) at file-scope or global-scope only.

• -gvms — produces debugging information in Alpha/VMS debug format (if that's supported). This is the format used by DEBUG on Alpha/VMS systems. 🎀
• -gcodeview — produces debugging information in CodeView debug format (if that's supported). This is the format used by Microsoft Visual C++ on Windows. 🪟
• -glevel
• -ggdblevel
• -gvmslevel — requests debugging information and also uses level to specify how much information. The default level is 2.

Level 0 produces no debug information at all. So -g0 negates -g.

Level 1 produces minimal information, enough for making backtraces in parts of the program that you don't plan to debug. This includes descriptions of functions and external variables, and line number tables, but no information about local variables.

Level 3 includes extra information, such as all the macro definitions present in the program. Some debuggers support macro expansion when you use -g3. 🔥

If you use multiple -g options, with or without level numbers, the last such option is the one that is effective. (last one wins, periodt 💅)

-gdwarf does not accept a concatenated debug level, to avoid confusion with -gdwarf-level. Instead use an additional -glevel option to change the debug level for DWARF. 🙈

• -fno-eliminate-unused-debug-symbols — by default, no debug information is produced for symbols that are not actually used. Use this option if you want debug information for all symbols. ⭐
• -femit-class-debug-always — instead of emitting debugging information for a C++ class in only one object file, emit it in all object files using the class. This option should be used only with debuggers that are unable to handle the way GCC normally emits debugging information for classes, because using this option increases the size of debugging information by as much as a factor of two. 😭 (double the bloat, bestie, be careful)
• -fno-merge-debug-strings — directs the linker to not merge together strings in the debugging information that are identical in different object files. Merging is not supported by all assemblers or linkers. Merging decreases the size of the debug information in the output file at the cost of increasing link processing time. Merging is enabled by default.
• -fdebug-prefix-map=old=new — when compiling files residing in directory old, record debugging information describing them as if the files resided in directory new instead. This can be used to replace a build-time path with an install-time path in the debug info. It can also be used to change an absolute path to a relative path by using . for new. This can give more reproducible builds, which are location independent, but may require an extra command to tell GDB where to find the source files. See also -ffile-prefix-map and -fcanon-prefix-map. 🗺️
• -fvar-tracking — runs the variable tracking pass. It computes where variables are stored at each position in code. Better debugging information is then generated (if the debugging information format supports this information). 🕵️

It's enabled by default when compiling with optimization (-Os, -O, -O2, ...), debugging information (-g) and the debug info format supports it.

• -fvar-tracking-assignments — annotates assignments to user variables early in the compilation and attempts to carry the annotations over throughout the compilation all the way to the end, in an attempt to improve debug information while optimizing. Use of -gdwarf-4 is recommended along with it. 📝

It can be enabled even if var-tracking is disabled, in which case annotations are created and maintained, but discarded at the end. By default, this flag is enabled together with -fvar-tracking, except when selective scheduling is enabled.

• -gsplit-dwarf — if DWARF debugging information is enabled, separate as much debugging information as possible into a separate output file with the extension .dwo. This option lets the build system avoid linking files with debug information. To be useful, this option requires a debugger capable of reading .dwo files. 🦋
• -gdwarf32
• -gdwarf64 — if DWARF debugging information is enabled, the -gdwarf32 selects the 32-bit DWARF format and the -gdwarf64 selects the 64-bit DWARF format. The default is target specific, on most targets it's -gdwarf32 though. The 32-bit DWARF format is smaller, but can't support more than 2GiB of debug information in any of the DWARF debug information sections. The 64-bit DWARF format allows larger debug information and might not be well supported by all consumers yet. 👑
• -gdescribe-dies — adds description attributes to some DWARF DIEs that have no name attribute, such as artificial variables, external references and call site parameter DIEs. ✨
• -gpubnames — bestie this one generates DWARF .debug_pubnames and .debug_pubtypes sections ✨
• -ggnu-pubnames — same .debug_pubnames and .debug_pubtypes sections but in a format that's ready to be converted into a GDB index. Heads up tho: this is only useful if your linker can produce GDB index version 7 💅
• -fdebug-types-section — when you're on DWARF Version 4 or higher, type DIEs can live in their own .debug_types section instead of just being part of .debug_info. Putting them in a separate comdat section is more efficient bc then the linker can yeet the duplicates 🔥 BUT not every DWARF consumer out there supports .debug_types sections yet, and on some objects it actually produces bigger debug info instead of smaller. So it's giving "sometimes slay, sometimes flop" 😭
• -grecord-gcc-switches, -gno-record-gcc-switches — this one appends the command-line options you used to invoke the compiler (the ones that might affect code generation) onto the DW_AT_producer attribute in your DWARF debugging info. The options get concatenated with spaces separating them from each other and from the compiler version. It's enabled by default ✨ Also peep -frecord-gcc-switches if you want a different way of cramming compiler options into the object file 👑
• -gstrict-dwarf — disallows using extensions from a later DWARF standard version than whatever you selected with -gdwarf-version. On most targets, using non-conflicting DWARF extensions from later standard versions is allowed, but not with this bad boy active 🚫
• -gno-strict-dwarf — the opposite vibe: allows using extensions of a later DWARF standard version than selected with -gdwarf-version 💖
• -gas-loc-support — tells the compiler that the assembler supports .loc directives. It can then use them so the assembler generates the DWARF2+ line number tables itself. This is generally what you want, bc assembler-generated line-number tables are sooo much more compact than the ones the compiler can make on its own 🎀 This option gets enabled by default if, at GCC configure time, the assembler was found to support those directives.
• -gno-as-loc-support — forces GCC to generate DWARF2+ line number tables internally, if DWARF2+ line number tables are to be generated at all 🙈
• -gas-locview-support — informs the compiler that the assembler supports "view" assignment and reset assertion checking in .loc directives. Enabled by default if, at GCC configure time, the assembler was found to support them ✨
• -gno-as-locview-support — forces GCC to assign view numbers internally, if -gvariable-location-views are explicitly requested 💅
• -gcolumn-info, -gno-column-info — emit location column information into DWARF debugging info, not just file and line. Enabled by default, periodt ✨
• -gstatement-frontiers, -gno-statement-frontiers — this makes GCC drop markers into the internal representation at the beginning of statements and keep them roughly in place throughout compilation, using them to guide the output of is_stmt markers in the line number table. It's enabled by default when compiling with optimization (-Os, -O1, -O2, ...) and outputting DWARF 2 debug info at the normal level 🔥
• -gvariable-location-views, -gvariable-location-views=incompat5, -gno-variable-location-views — augments variable location lists with progressive view numbers implied from the line number table. This lets debug info consumers inspect state at certain points of the program, even when no instructions associated with the corresponding source locations are present at that point 👑 If the assembler lacks support for view numbers in line number tables, this causes the compiler to emit the line number table, which generally makes them somewhat less compact. The augmented line number tables and location lists are fully backward-compatible, so they can be consumed by debug info consumers that aren't aware of these augmentations — but they won't get any benefit from them either 😭

This is enabled by default when outputting DWARF 2 debug info at the normal level, as long as there's assembler support, -fvar-tracking-assignments is enabled, and -gstrict-dwarf is not. When assembler support isn't available, this may still be enabled, but it'll force GCC to output internal line number tables, and if -ginternal-reset-location-views isn't enabled, that will most certainly lead to silently mismatching location views 💀

There's a proposed representation for view numbers that is not backward compatible with the location list format introduced in DWARF 5, and you can enable it with -gvariable-location-views=incompat5. Heads up: this option may be removed in the future, and it's only provided as a reference implementation of the proposed representation. Debug info consumers are not expected to support this extended format, and they'd be rendered unable to decode location lists using it 🚫

• -ginternal-reset-location-views, -gno-internal-reset-location-views — attempts to figure out location views that can be omitted from location view lists. This requires the compiler to have very accurate insn length estimates, which isn't always the case, and it may cause incorrect view lists to be generated silently when using an assembler that doesn't support location view lists 😬 The GNU assembler will flag any such error as a "view number mismatch". This is only enabled on ports that define a reliable estimation function.
• -ginline-points, -gno-inline-points — generates extended debug info for inlined functions 🎀 Location view tracking markers get inserted at inlined entry points so that address and view numbers can be computed and output in debug info. This can be enabled independently of location views, in which case the view numbers won't be output — but it can only be enabled along with statement frontiers, and it's only enabled by default if location views are enabled ✨
• -gz[=type] — produce compressed debug sections in DWARF format, if that's supported. If type isn't given, the default type depends on what your assembler and linker can do. type can be one of none (don't compress debug sections) or zlib (use zlib compression in ELF gABI format). If the linker doesn't support writing compressed debug sections, the option gets rejected 🚫 Otherwise, if the assembler doesn't support them, -gz is silently ignored when producing object files 🙈
• -femit-struct-debug-baseonly — emits debug info for struct-like types only when the base name of the compilation source file matches the base name of the file where the struct is defined. This substantially reduces the size of debugging information, but at significant potential loss in type info to the debugger 💀 Peep -femit-struct-debug-reduced for a less aggressive option, and -femit-struct-debug-detailed for more detailed control. This option works only with DWARF debug output.
• -femit-struct-debug-reduced — emits debug info for struct-like types only when the base name of the compilation source file matches the base name of the file where the type is defined — unless the struct is a template or defined in a system header 👑 This significantly reduces the size of debugging information, with some potential loss in type info to the debugger. See -femit-struct-debug-baseonly for a more aggressive option, and -femit-struct-debug-detailed for more detailed control. Also works only with DWARF debug output ✨
• -femit-struct-debug-detailed[=spec-list] — specify the struct-like types for which the compiler generates debug info. The whole intent here is to reduce duplicate struct debug info between different object files within the same program 🎀 This is the detailed version of -femit-struct-debug-reduced and -femit-struct-debug-baseonly, which serve most needs.

A specification has the syntax:

[dir:|ind:][ord:|gen:](any|sys|base|none)

The optional first word limits the spec to structs that are used directly (dir:) or used indirectly (ind:). A struct type is used directly when it's the type of a variable or member. Indirect uses arise through pointers to structs. That is, when use of an incomplete struct is valid, the use is indirect. An example is struct one direct; struct two * indirect;.

The optional second word limits the spec to ordinary structs (ord:) or generic structs (gen:). Generic structs are a bit complicated to explain 💅 For C++, these are non-explicit specializations of template classes, or non-template classes within the above. Other programming languages have generics, but -femit-struct-debug-detailed doesn't implement them yet 🙈

The third word specifies the source files for those structs for which the compiler should emit debug info. The values none and any have the normal meaning. The value base means the base of the name of the file in which the type declaration appears must match the base of the name of the main compilation file. In practice, this means that when compiling foo.c, debug info is generated for types declared in that file and foo.h, but not other header files. The value sys means those types satisfying base or declared in system or compiler headers.

You may need to experiment to determine the best settings for your application fr fr.

The default is -femit-struct-debug-detailed=all.

This option works only with DWARF debug output ⭐

• -fno-dwarf2-cfi-asm — emits DWARF unwind info as a compiler-generated .eh_frame section instead of using GAS .cfi_* directives 💅
• -fno-eliminate-unused-debug-types — normally, when producing DWARF output, GCC avoids producing debug symbol output for types that are nowhere used in the source file being compiled. Sometimes it's useful to have GCC emit debug info for all types declared in a compilation unit, regardless of whether or not they're actually used in that compilation unit — like, say, if in the debugger you want to cast a value to a type that isn't actually used in your program (but is declared) 👑 More often, though, this results in a significant amount of wasted space 😭

---

🌈 Options That Control Optimization 🌈

These options control various sorts of optimizations, bestie. Buckle up 💖

okay so without ANY optimization flag, your compiler is in chill mode: its whole goal is to compile fast and make debugging actually make sense. statements are independent, bestie—if you break between statements you can just change any variable or move the program counter to another statement in the function and get exactly what you expect from the source. no cap. 💅

but the second you turn on optimization flags, the compiler starts trying to make your code faster and/or smaller—at the cost of compile time and possibly your ability to debug. it's giving trade-offs. 🦋

the compiler optimizes based on what it knows about your whole program. so if you compile multiple files at once into one output file, it can use info from ALL the files when compiling each one. lowkey smart. ✨

not every optimization has its own flag. only the ones with flags are listed here, so don't come for me if you see something missing. 🙈

most optimizations are totally off at -O0 or if you didn't set any -O level on the command line—even if you explicitly turn on individual flags. same vibe with -Og: it suppresses a bunch of passes. 💀

depending on your target and how GCC was built, the actual set of optimizations at each -O level might differ slightly from what's listed here. wanna know exactly? run GCC with -Q --help=optimizers and it'll spill the tea on what's enabled at each level. 👑

• -O, -O1 — Optimize. Optimizing compilation takes a bit more time and a lot more memory for big functions. 😭

With -O, the compiler tries to reduce code size and execution time without doing any optimizations that take a ton of compile time. ⏳

-O is the recommended level for large machine-generated code—it's the sensible balance between compile time and memory use. higher levels do more complex optimizations than -O. 🎀

-O turns on these flags:

-fauto-inc-dec -fbranch-count-reg -fcombine-stack-adjustments
-fcompare-elim -fcprop-registers -fdce -fdefer-pop -fdelayed-branch
-fdse -fforward-propagate -fguess-branch-probability -fif-conversion
-fif-conversion2 -finline-functions-called-once -fipa-modref
-fipa-profile -fipa-pure-const -fipa-reference
-fipa-reference-addressable -fivopts -fmerge-constants
-fmove-loop-invariants -fmove-loop-stores -fomit-frame-pointer
-freorder-blocks -fshrink-wrap -fshrink-wrap-separate
-fsplit-wide-types -fssa-backprop -fssa-phiopt -ftree-bit-ccp
-ftree-ccp -ftree-ch -ftree-coalesce-vars -ftree-copy-prop
-ftree-dce -ftree-dominator-opts -ftree-dse -ftree-forwprop
-ftree-fre -ftree-phiprop -ftree-pta -ftree-scev-cprop -ftree-sink
-ftree-slsr -ftree-sra -ftree-ter -funit-at-a-time

• -O2 — Optimize even more. GCC performs nearly all supported optimizations that don't involve a space-speed tradeoff. compared to -O, this increases both compilation time and the performance of the generated code. 🔥

-O2 turns on everything from -O1. it also turns on:

-falign-functions  -falign-jumps -falign-labels  -falign-loops
-fcaller-saves -fcode-hoisting -fcrossjumping -fcse-follow-jumps
-fcse-skip-blocks -fdelete-null-pointer-checks -fdevirtualize
-fdevirtualize-speculatively -fexpensive-optimizations
-ffinite-loops -fgcse  -fgcse-lm -fhoist-adjacent-loads
-finline-functions -finline-small-functions -findirect-inlining
-fipa-bit-cp  -fipa-cp  -fipa-icf -fipa-ra  -fipa-sra  -fipa-vrp
-fisolate-erroneous-paths-dereference -flra-remat -foptimize-crc
-foptimize-sibling-calls -foptimize-strlen -fpartial-inlining
-fpeephole2 -freorder-blocks-algorithm=stc
-freorder-blocks-and-partition  -freorder-functions
-frerun-cse-after-loop -fschedule-insns  -fschedule-insns2
-fsched-interblock  -fsched-spec -fstore-merging -fstrict-aliasing
-fthread-jumps -ftree-builtin-call-dce -ftree-loop-vectorize
-ftree-pre -ftree-slp-vectorize -ftree-switch-conversion
-ftree-tail-merge -ftree-vrp -fvect-cost-model=very-cheap

quick warning: check the -fgcse note below about using -O2 on programs with computed gotos. ⚠️

NOTE for Ubuntu besties: in Ubuntu 8.10 and later, -D_FORTIFY_SOURCE=2; in Ubuntu 24.04 and later, -D_FORTIFY_SOURCE=3, is set by default, and activates when -O is 2 or higher. this adds extra compile-time and run-time checks for several libc functions. to disable, use -U_FORTIFY_SOURCE or -D_FORTIFY_SOURCE=0. 💖

NOTE for Debian 13 and Ubuntu 24.04+: on 32-bit architectures armel, armhf, hppa, m68k, mips, mipsel, powerpc and sh4, -D_TIME_BITS=64 and -D_FILE_OFFSET_BITS=64 are set by default. 🌈

• -O3 — Optimize yet more. -O3 turns on everything from -O2 plus:

-fgcse-after-reload -fipa-cp-clone -floop-interchange
-floop-unroll-and-jam -fpeel-loops -fpredictive-commoning
-fsplit-loops -fsplit-paths -ftree-loop-distribution
-ftree-partial-pre -funswitch-loops -fvect-cost-model=dynamic
-fversion-loops-for-strides

• -O0 — Reduce compile time and make debugging produce expected results. this is the default. 🐢

At -O0, GCC completely disables most optimization passes; they don't run even if you explicitly enable them on the command line, or if -Q --help=optimizers says they're enabled by default. many optimizations depend on code analysis or canonicalization passes enabled by -O, and it wouldn't be useful to run individual passes in isolation. 💅

• -Os — Optimize for size. -Os enables all -O2 optimizations except those that often increase code size:

-falign-functions  -falign-jumps -falign-labels  -falign-loops
-fprefetch-loop-arrays  -freorder-blocks-algorithm=stc

It also enables -finline-functions, tunes for code size rather than execution speed, and performs further optimizations designed to reduce code size. 🦋

• -Ofast — Disregard strict standards compliance. -Ofast enables all -O3 optimizations. it also enables optimizations that aren't valid for all standard-compliant programs. it turns on -ffast-math, -fallow-store-data-races and the Fortran-specific -fstack-arrays (unless -fmax-stack-var-size is specified), and -fno-protect-parens. it turns off -fsemantic-interposition. 🚀
• -Og — Optimize debugging experience. -Og should be your go-to for the standard edit-compile-debug cycle—it gives a reasonable level of optimization while keeping compilation fast and debugging good. it's a better choice than -O0 for debuggable code because some compiler passes that collect debug info are disabled at -O0. 🐛

Like -O0, -Og completely disables a number of optimization passes so individual options controlling them have no effect. otherwise, -Og enables all -O1 optimization flags except those that might interfere with debugging:

-fbranch-count-reg  -fdelayed-branch -fdse  -fif-conversion
-fif-conversion2 -finline-functions-called-once
-fmove-loop-invariants  -fmove-loop-stores  -fssa-phiopt
-ftree-bit-ccp  -ftree-dse  -ftree-pta  -ftree-sra

• -Oz — Optimize aggressively for size rather than speed. this might increase the number of instructions executed if those instructions require fewer bytes to encode. -Oz behaves similarly to -Os including enabling most -O2 optimizations. 🍓

If you use multiple -O options, with or without level numbers, the last one wins. periodt. 💅

Options of the form -fflag are machine-independent flags. most have positive and negative forms; the negative of -ffoo is -fno-foo. in the table below, only one form is listed—the one you'd typically use. you can figure out the other by either removing no- or adding it. 🧚

The following options control specific optimizations. they're either activated by -O options or related to ones that are. use these bad boys when you want to "fine-tune" optimizations. it's giving custom vibes. ✨

• -fno-defer-pop — For machines that must pop arguments after a function call, always pop them as soon as each function returns. at -O1 and higher, -fdefer-pop is the default; this lets the compiler let arguments accumulate on the stack for several function calls and pop them all at once. 🎀
• -fforward-propagate — Perform a forward propagation pass on RTL. the pass tries to combine two instructions and checks if the result can be simplified. if loop unrolling is active, two passes are performed and the second is scheduled after loop unrolling. 🔄

This option is enabled by default at optimization levels -O1, -O2, -O3, -Os. 💖

• -favoid-store-forwarding, -fno-avoid-store-forwarding — Many CPUs stall for many cycles when a load partially depends on previous smaller stores. this pass tries to detect such cases and avoid the penalty by changing the order of the load and store and then fixing up the loaded value. 🛑

Disabled by default. 😴

• -ffp-contract=style — -ffp-contract=off disables floating-point expression contraction. -ffp-contract=fast enables it—like forming fused multiply-add operations if the target has native support. -ffp-contract=on enables contraction if allowed by the language standard. this is implemented for C and C++, where it enables contraction within one expression, but not across different statements. 🧮

The default is -ffp-contract=off for C in a standards compliant mode (-std=c11 or similar), -ffp-contract=fast otherwise. 🎯

• -fomit-frame-pointer — Omit the frame pointer in functions that don't need one. this avoids the instructions to save, set up and restore the frame pointer; on many targets it also makes an extra register available. 💅

On some targets this flag has no effect because the standard calling sequence always uses a frame pointer, so it can't be omitted. 🙅

Note that -fno-omit-frame-pointer doesn't guarantee the frame pointer is used in all functions. several targets always omit the frame pointer in leaf functions. 🍃

Enabled by default at -O1 and higher. 🔥

• -foptimize-crc — okay so this one sniffs out loops that are out here computing CRC (like, doing polynomial long division by hand, one bit at a time 😭) and swaps them for something way faster. It catches 8, 16, 32, and 64 bit CRC, with a constant polynomial that's missing the leading 1 bit, for both the bit-forward and bit-reversed cases. If your target has a CRC instruction and the polynomial in your source code matches the polynomial in the CRC instruction — boom, it generates that instruction. Otherwise? If the target supports a carry-less-multiplication instruction, it builds the CRC with that. Otherwise it falls back to generating a table-based CRC.

Enabled by default at -O2 and higher. ✨

• -foptimize-sibling-calls — optimizes sibling calls and tail recursive calls. cute, tidy, no leftover stack frames. 🎀

Enabled at levels -O2, -O3, -Os.

• -foptimize-strlen — takes a bunch of standard C string functions (like strlen, strchr or strcpy) plus their _FORTIFY_SOURCE counterparts and turns them into faster alternatives. glow up fr.

Enabled at levels -O2, -O3.

• -finline-stringops[=fn] — expands memory and string operations inline (for now, only memset), even when the length is variable or big enough that it would literally need a loop. This is most useful when you're pairing it with -ffreestanding and -fno-builtin.

In some circumstances it lets the compiler generate code that takes advantage of known alignment and length multipliers — but even then it may be less efficient than optimized runtime implementations, and it can grow code size so much that even a less performant but shared implementation runs faster, just because it uses the code caches better. This option is disabled by default. 🫶

• -fno-inline — do NOT expand any functions inline, except the ones marked with the always_inline attribute. This is the default when you're not optimizing.

You can also exempt individual functions from inlining by marking them with the noinline attribute. 💅

• -finline-small-functions — integrates functions into their callers when the function body is smaller than the expected function-call code (so the overall program gets smaller 🐜). The compiler heuristically decides which functions are simple enough to be worth integrating this way. This inlining applies to all functions, even the ones not declared inline.

Enabled at levels -O2, -O3, -Os.

• -findirect-inlining — inlines indirect calls too, the ones that turn out to be known at compile time thanks to previous inlining. This option only does anything at all when inlining itself is turned on by -finline-functions or -finline-small-functions.

Enabled at levels -O2, -O3, -Os.

• -finline-functions — considers all functions for inlining, even if they're not declared inline. The compiler heuristically decides which functions are worth integrating this way.

If all calls to a given function get integrated, and the function is declared static, then normally it's not output as assembler code in its own right. iconic behavior tbh.

Enabled at levels -O2, -O3, -Os. Also enabled by -fprofile-use and -fauto-profile.

• -finline-functions-called-once — considers all static functions called once for inlining into their caller, even if they're not marked inline. If a call to a given function gets integrated, the function is not output as assembler code in its own right.

Enabled at levels -O1, -O2, -O3 and -Os, but not -Og.

• -fearly-inlining — inlines functions marked by always_inline plus functions whose body seems smaller than the function call overhead, early — before doing -fprofile-generate instrumentation and the real inlining pass. Doing this makes profiling significantly cheaper and usually makes inlining faster on programs with big chains of nested wrapper functions. no cap.

Enabled by default.

• -fipa-sra — performs interprocedural scalar replacement of aggregates, removal of unused parameters, and replacement of parameters passed by reference with parameters passed by value.

Enabled at levels -O2, -O3 and -Os.

• *-finline-limit=n* — bestie, by default GCC limits the size of functions that can be inlined, and this flag lets you coarse-control that limit. n is the size of functions that can be inlined, measured in number of pseudo instructions.

Inlining is actually controlled by a bunch of parameters, which you can specify individually using --param name=value. The -finline-limit=n option sets some of these parameters like so:

• max-inline-insns-single — is set to n/2.
• max-inline-insns-auto — is set to n/2.

See below for documentation of the individual parameters controlling inlining and the defaults of those parameters.

Note: there may be no value of -finline-limit that actually results in default behavior. 🙈

Note: "pseudo instruction" here means an abstract measurement of a function's size. It is in no way a count of assembly instructions, so its exact meaning might change from one release to another. 💀

• -fno-keep-inline-dllexport — this is a more fine-grained version of -fkeep-inline-functions, and it applies only to functions declared using the dllexport attribute or declspec.
• -fkeep-inline-functions — in C, emit static functions that are declared inline into the object file, even if the function got inlined into all of its callers. Heads up: this switch does not affect functions using the extern inline extension in GNU C90. In C++, it emits any and all inline functions into the object file.
• -fkeep-static-functions — emits static functions into the object file, even if the function is never used. hoarding but make it compilation.
• -fkeep-static-consts — emits variables declared static const when optimization isn't turned on, even if the variables aren't referenced.

GCC enables this option by default. If you want to force the compiler to check whether a variable is referenced, regardless of whether or not optimization is turned on, use the -fno-keep-static-consts option.

• -fmerge-constants — tries to merge identical constants (string constants and floating-point constants) across compilation units. dedupe queen 👑

This option is the default for optimized compilation if the assembler and linker support it. Use -fno-merge-constants to inhibit this behavior.

Enabled at levels -O1, -O2, -O3, -Os.

• -fmerge-all-constants — tries to merge identical constants and identical variables.

This option implies -fmerge-constants. On top of -fmerge-constants, it also considers stuff like constant initialized arrays, or initialized constant variables with integral or floating-point types. But here's the thing: languages like C or C++ require each variable — including multiple instances of the same variable in recursive calls — to have distinct locations, so using this option results in non-conforming behavior. ⚠️

• -fmodulo-sched — performs swing modulo scheduling immediately before the first scheduling pass. This pass looks at innermost loops and reorders their instructions by overlapping different iterations. overlapping is the whole vibe.
• -fmodulo-sched-allow-regmoves — performs more aggressive SMS-based modulo scheduling, with register moves allowed. Setting this flag deletes certain anti-dependences edges, which triggers the generation of reg-moves based on life-range analysis. This option is effective only when -fmodulo-sched is enabled.
• -fno-branch-count-reg — disables the optimization pass that scans for opportunities to use decrement and branch instructions on a count register, instead of instruction sequences that decrement a register, compare it against zero, and then branch based on the result. This option is only meaningful on architectures that support such instructions — which include x86, PowerPC, IA-64 and S/390. Note that the -fno-branch-count-reg option doesn't remove the decrement and branch instructions from the generated instruction stream that were introduced by other optimization passes.

The default is -fbranch-count-reg at -O1 and higher, except for -Og.

• -fno-function-cse — do not put function addresses in registers; make each instruction that calls a constant function contain the function's address explicitly.

This option results in less efficient code, but some strange hacks that alter the assembler output may get confused by the optimizations performed when this option is not used. 😵‍💫

The default is -ffunction-cse

• -fno-zero-initialized-in-bss — if the target supports a BSS section, GCC by default puts variables that are initialized to zero into BSS, which can save space in the resulting code.

This option turns that behavior off, because some programs explicitly rely on variables going to the data section — e.g., so the resulting executable can find the beginning of that section and/or make assumptions based on that.

The default is -fzero-initialized-in-bss except in Ada.

• -fthread-jumps — performs optimizations that check whether a jump branches to a location where another comparison subsumed by the first is found. If so, the first branch is redirected to either the destination of the second branch or a point immediately following it, depending on whether the condition is known to be true or false.

Enabled at levels -O1, -O2, -O3, -Os.

• -fsplit-wide-types — when you're using a type that occupies multiple registers, like long long on a 32-bit system, this splits the registers apart and allocates them independently. This normally generates better code for those types, but may make debugging more difficult. 🙈

Enabled at levels -O1, -O2, -O3, -Os.

• -fsplit-wide-types-early — fully splits wide types early, instead of very late. This option has no effect unless -fsplit-wide-types is turned on.

This is the default on some targets.

• -fcse-follow-jumps — in common subexpression elimination (CSE), scan through jump instructions when the target of the jump isn't reached by any other path. For example, when CSE encounters an if statement with an else clause, CSE follows the jump when the condition tested is false.

Enabled at levels -O2, -O3, -Os.

• -fcse-skip-blocks — similar to -fcse-follow-jumps, but it makes CSE follow jumps that conditionally skip over blocks. When CSE encounters a simple if statement with no else clause, -fcse-skip-blocks makes CSE follow the jump around the body of the if.
• -frerun-cse-after-loop — bestie, we're gonna run common subexpression elimination again, right after all the loop optimizations are done doing their lil thing. ✨ Enabled at levels -O2, -O3, -Os.
• -fgcse — this one runs a global common subexpression elimination pass, and like, it also does global constant and copy propagation while it's in there. Multitasking icon. 💅

Note: when you're compiling a program that uses computed gotos (a GCC extension, iykyk), you may get better run-time performance if you disable the global common subexpression elimination pass by adding -fno-gcse to the command line.

Enabled at levels -O2, -O3, -Os.

• -fgcse-lm — when this is enabled, global common subexpression elimination tries to move loads that are only killed by stores into themselves. Which like, sounds unhinged but it lets a loop containing a load/store sequence be changed to a load outside the loop, and a copy/store within the loop. Obsessed. 🦋 Enabled by default when -fgcse is enabled.
• -fgcse-sm — when this is enabled, a store motion pass runs after global common subexpression elimination, and it attempts to move stores out of loops. Team it up with -fgcse-lm and loops containing a load/store sequence can be changed to a load before the loop and a store after the loop. Cute little duo fr fr. Not enabled at any optimization level. 💀
• -fgcse-las — when this is enabled, the global common subexpression elimination pass eliminates redundant loads that come after stores to the same memory location (both partial and full redundancies, no crumbs left). Not enabled at any optimization level.
• -fgcse-after-reload — when this is enabled, a redundant load elimination pass is performed after reload. The whole purpose of this pass is to clean up redundant spilling. Little janitor energy. 🧹 Enabled by -O3, -fprofile-use and -fauto-profile.
• -faggressive-loop-optimizations — this tells the loop optimizer to use language constraints to derive bounds for the number of iterations of a loop. It assumes that loop code does not invoke undefined behavior, for example by causing signed integer overflows or out-of-bound array accesses. The bounds for the number of iterations of a loop are used to guide loop unrolling and peeling and loop exit test optimizations. This option is enabled by default.
• -funconstrained-commons — tells the compiler that variables declared in common blocks (e.g. Fortran) may later be overridden with longer trailing arrays. This prevents certain optimizations that depend on knowing the array bounds. So like, it's the "don't assume bestie, we don't actually know how big this is" flag. 🙈
• -fcrossjumping — performs the cross-jumping transformation, which unifies equivalent code and saves code size. The resulting code may or may not perform better than without cross-jumping. No promises, we're just saying. 💁 Enabled at levels -O2, -O3, -Os.
• -fauto-inc-dec — combines increments or decrements of addresses with memory accesses. This pass is always skipped on architectures that don't have instructions to support it. Enabled by default at -O1 and higher on architectures that support this.
• -fdce — performs dead code elimination (DCE) on RTL. Enabled by default at -O1 and higher. Bye to the dead code, we don't know her. ✌️
• -fdse — performs dead store elimination (DSE) on RTL. Enabled by default at -O1 and higher.
• -fif-conversion — attempts to transform conditional jumps into branch-less equivalents. This includes use of conditional moves, min, max, set flags and abs instructions, plus some tricks that are doable by standard arithmetics. The use of conditional execution on chips where it's available is controlled by -fif-conversion2. Enabled at levels -O1, -O2, -O3, -Os, but not with -Og. Noted. 📝
• -fif-conversion2 — uses conditional execution (where available) to transform conditional jumps into branch-less equivalents. Enabled at levels -O1, -O2, -O3, -Os, but not with -Og.
• -fdeclone-ctor-dtor — so the C++ ABI requires multiple entry points for constructors and destructors: one for a base subobject, one for a complete object, and one for a virtual destructor that calls operator delete afterwards. For a hierarchy with virtual bases, the base and complete variants are clones, which means two copies of the function. With this option, the base and complete variants are changed to be thunks that call a common implementation. Deduplicating the queen behavior. 👑

Enabled by -Os.

• -fdelete-null-pointer-checks — assume that programs cannot safely dereference null pointers, and that no code or data element resides at address zero. This option enables simple constant folding optimizations at all optimization levels. In addition, other optimization passes in GCC use this flag to control global dataflow analyses that eliminate useless checks for null pointers; these assume that a memory access to address zero always results in a trap, so that if a pointer is checked after it has already been dereferenced, it cannot be null.

Note however that in some environments this assumption is not true. Use -fno-delete-null-pointer-checks to disable this optimization for programs that depend on that behavior.

This option is enabled by default on most targets. On AVR and MSP430, this option is completely disabled. 🚫

Passes that use the dataflow information are enabled independently at different optimization levels.

• -fdevirtualize — attempts to convert calls to virtual functions to direct calls. This is done both within a procedure and interprocedurally as part of indirect inlining (-findirect-inlining) and interprocedural constant propagation (-fipa-cp). Enabled at levels -O2, -O3, -Os.
• -fdevirtualize-speculatively — attempts to convert calls to virtual functions to speculative direct calls. Based on the analysis of the type inheritance graph, it determines for a given call the set of likely targets. If the set is small, preferably of size 1, it changes the call into a conditional deciding between direct and indirect calls. The speculative calls enable more optimizations, such as inlining. When they seem useless after further optimization, they are converted back into original form. Manifesting, then un-manifesting. ✨
• -fdevirtualize-at-ltrans — streams extra information needed for aggressive devirtualization when running the link-time optimizer in local transformation mode. This option enables more devirtualization but significantly increases the size of streamed data. For this reason it is disabled by default.
• -fexpensive-optimizations — performs a number of minor optimizations that are relatively expensive. Yes, treating yourself. Enabled at levels -O2, -O3, -Os.
• -free — attempts to remove redundant extension instructions. This is especially helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit registers after writing to their lower 32-bit half. Slay. 💅 Enabled for Alpha, AArch64, LoongArch, PowerPC, RISC-V, SPARC, h83000 and x86 at levels -O2, -O3, -Os.
• -fno-lifetime-dse — in C++, the value of an object is only affected by changes within its lifetime: when the constructor begins, the object has an indeterminate value, and any changes during the lifetime of the object are dead when the object is destroyed. Normally dead store elimination will take advantage of this; if your code relies on the value of the object storage persisting beyond the lifetime of the object, you can use this flag to disable this optimization. To preserve stores before the constructor starts (e.g. because your operator new clears the object storage) but still treat the object as dead after the destructor, you can use -flifetime-dse=1. The default behavior can be explicitly selected with -flifetime-dse=2. -flifetime-dse=0 is equivalent to -fno-lifetime-dse.
• -flive-range-shrinkage — attempts to decrease register pressure through register live range shrinkage. This is helpful for fast processors with small or moderate size register sets. 🍓
• *-fira-algorithm=algorithm* — use the specified coloring algorithm for the integrated register allocator. The algorithm argument can be priority, which specifies Chow's priority coloring, or CB, which specifies Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented for all architectures, but for those targets that do support it, it is the default because it generates better code. Main character algorithm, honestly. 👑
• *-fira-region=region* — use specified regions for the integrated register allocator. The region argument should be one of the following:
  • all — use all loops as register allocation regions. This can give the best results for machines with a small and/or irregular register set.
  • mixed — use all loops except for loops with small register pressure as the regions. This value usually gives the best results in most cases and for most architectures, and is enabled by default when compiling with optimization for speed (-O, -O2, ...).
  • one — use all functions as a single region. This typically results in the smallest code size, and is enabled by default for -Os or -O0.
• -fira-hoist-pressure — use IRA to evaluate register pressure in the code hoisting pass for decisions to hoist expressions. This option usually results in smaller code, but it can slow the compiler down. Trade-offs, always trade-offs. 😭

This option is enabled at level -Os for all targets.

• -fira-loop-pressure — use IRA to evaluate register pressure in loops for decisions to move loop invariants. This option usually results in generation of faster and smaller code on machines with large register files (>= 32 registers), but it can slow the compiler down. 🌈

This option is enabled at level -O3 for some targets.

• -fno-ira-share-save-slots — disables sharing of stack slots used for saving call-used hard registers living through a call. Each hard register gets a separate stack slot, and as a result function stack frames are larger. Everyone gets their own room now. 🏠
• -fno-ira-share-spill-slots — disables sharing of stack slots allocated for pseudo-registers. Each pseudo-register that does not get a hard register gets a separate stack slot, and as a result function stack frames are larger.
• -flra-remat — enables CFG-sensitive rematerialization in LRA. Instead of loading values of spilled pseudos, LRA tries to rematerialize (recalculate) values if it is profitable. Enabled at levels -O2, -O3, -Os.
• -fdelayed-branch — if the target machine supports it, this tries to reorder instructions so they can slot into the empty spots that exist after delayed branch instructions 👑 like, filling the gaps, no wasted space, iconic behavior fr.

It's on by default at -O1, -O2, -O3, -Os — but NOT at -Og. Noted 💅

• -fschedule-insns — if the target supports it, attempt to reorder instructions so you don't get execution stalls from data not being ready yet. The way this helps: machines with slow floating point or memory load instructions can keep issuing other instructions while waiting for that load or FP result to actually be needed. It's giving "don't just stand there, do something else" energy 🦋

Conventionally it's enabled at -O2 and -O3. BUT (and this is the plot twist) a lot of targets override that. For example, on x86 it's disabled at all levels, while on AArch64 it's enabled only at -O3. The lore is inconsistent bestie, that's just how it is 😭

• -fschedule-insns2 — basically -fschedule-insns's sequel 🎬 same vibe but it requests an extra pass of instruction scheduling after register allocation has already happened. Super useful on machines with a relatively small number of registers, and where memory load instructions take more than one cycle.

Enabled at -O2, -O3, -Os.

• -fno-sched-interblock — turns off instruction scheduling across basic blocks. Normally that's enabled when scheduling happens before register allocation, aka with -fschedule-insns or at -O2 or higher. So this is the off switch 🚫
• -fno-sched-spec — turns off speculative motion of non-load instructions. Normally enabled when scheduling before register allocation, i.e. with -fschedule-insns or at -O2 or higher.
• -fsched-pressure — enables register-pressure-sensitive instruction scheduling before register allocation. Heads up: this only makes sense if scheduling before register allocation is actually on, meaning with -fschedule-insns or at -O2 or higher. Using it can improve the code GCC generates and shrink its size, by stopping register pressure from climbing above the number of available hard registers — which stops the spills that would come later in register allocation. Preventing the mess before it happens, we love a proactive queen ✨
• -fsched-spec-load — allows speculative motion of some load instructions. Again: only makes sense when scheduling before register allocation, i.e. with -fschedule-insns or at -O2 or higher.
• -fsched-spec-load-dangerous — allows speculative motion of more load instructions. Yes bestie the flag is literally named dangerous 💀 and it also only makes sense when scheduling before register allocation, i.e. with -fschedule-insns or at -O2 or higher.
• -fsched-stalled-insns
• -fsched-stalled-insns=n — this one defines how many insns (if any) can be moved prematurely from the queue of stalled insns into the ready list, during the second scheduling pass.
  • -fno-sched-stalled-insns means no insns get moved prematurely.
  • -fsched-stalled-insns=0 means there's no limit on how many queued insns can be moved prematurely.
  • -fsched-stalled-insns with no value is the same as -fsched-stalled-insns=1.
• -fsched-stalled-insns-dep
• -fsched-stalled-insns-dep=n — defines how many insn groups (cycles) get examined for a dependency on a stalled insn that's a candidate for premature removal from the queue of stalled insns. Two conditions, both required: this only has an effect during the second scheduling pass, and only if -fsched-stalled-insns is used.
  • -fno-sched-stalled-insns-dep is equivalent to -fsched-stalled-insns-dep=0.
  • -fsched-stalled-insns-dep with no value is equivalent to -fsched-stalled-insns-dep=1.
• -fsched2-use-superblocks — when scheduling after register allocation, use superblock scheduling. This allows motion across basic block boundaries, which gives you faster schedules 🌈

⚠️ Heads up, this option is experimental. Not all machine descriptions used by GCC model the CPU closely enough to avoid unreliable results from the algorithm. So like… proceed with caution, babe.

Also it only makes sense when scheduling after register allocation, i.e. with -fschedule-insns2 or at -O2 or higher.

---

okay now the scheduler heuristics, aka GCC's little set of priorities for choosing which instruction goes first 💜

• -fsched-group-heuristic — enables the group heuristic in the scheduler. This one favors the instruction that belongs to a schedule group. Enabled by default when scheduling is enabled, i.e. with -fschedule-insns or -fschedule-insns2 or at -O2 or higher.
• -fsched-critical-path-heuristic — enables the critical-path heuristic. Favors instructions on the critical path. Enabled by default when scheduling is enabled, i.e. with -fschedule-insns or -fschedule-insns2 or at -O2 or higher.
• -fsched-spec-insn-heuristic — enables the speculative instruction heuristic. Favors speculative instructions with greater dependency weakness. Enabled by default when scheduling is enabled, i.e. with -fschedule-insns or -fschedule-insns2 or at -O2 or higher.
• -fsched-rank-heuristic — enables the rank heuristic. Favors the instruction belonging to a basic block with greater size or frequency. Enabled by default when scheduling is enabled, i.e. with -fschedule-insns or -fschedule-insns2 or at -O2 or higher.
• -fsched-last-insn-heuristic — enables the last-instruction heuristic. Favors the instruction that's less dependent on the last instruction scheduled. Enabled by default when scheduling is enabled, i.e. with -fschedule-insns or -fschedule-insns2 or at -O2 or higher.
• -fsched-dep-count-heuristic — enables the dependent-count heuristic. Favors the instruction that has more instructions depending on it. Enabled by default when scheduling is enabled, i.e. with -fschedule-insns or -fschedule-insns2 or at -O2 or higher.
• -freschedule-modulo-scheduled-loops — so here's the tea 🫖 modulo scheduling happens before traditional scheduling. If a loop is modulo scheduled, later scheduling passes may change its schedule. Use this option to control that behavior.
• -fselective-scheduling — schedule instructions using the selective scheduling algorithm. Selective scheduling runs instead of the first scheduler pass.
• -fselective-scheduling2 — schedule instructions using the selective scheduling algorithm. Selective scheduling runs instead of the second scheduler pass.
• -fsel-sched-pipelining — enable software pipelining of innermost loops during selective scheduling. No effect unless one of -fselective-scheduling or -fselective-scheduling2 is turned on. (Dependency chain bestie, keep up 💅)
• -fsel-sched-pipelining-outer-loops — when pipelining loops during selective scheduling, also pipeline outer loops. No effect unless -fsel-sched-pipelining is turned on.

---

• -fsemantic-interposition — buckle up, this one's a whole essay 😭

Some object formats, like ELF, allow interposing of symbols by the dynamic linker. What that means: for symbols exported from the DSO, the compiler can't do interprocedural propagation, inlining and other optimizations, because it has to assume the function or variable in question might change. Is that feature useful? Yes — for example, to rewrite memory allocation functions with a debugging implementation. Is it expensive in terms of code quality? Also yes 💀

With -fno-semantic-interposition, the compiler assumes that if interposition happens for functions, the overwriting function will have precisely the same semantics (and side effects). Same deal for variables: if interposition happens for variables, the constructor of the variable will be the same.

Important exceptions ✋ the flag has no effect for functions explicitly declared inline (where it's never allowed for interposition to change semantics) and for symbols explicitly declared weak.

• -fshrink-wrap — emit function prologues only before the parts of the function that actually need it, rather than at the very top of the function. Enabled by default at -O and higher. Minimal effort, maximum slay 🌸
• -fshrink-wrap-separate — shrink-wrap the separate parts of the prologue and epilogue separately, so those parts only get executed when needed. On by default, but no effect unless -fshrink-wrap is also turned on and the target supports it.
• -fcaller-saves — enable allocation of values to registers that get clobbered by function calls, by emitting extra instructions to save and restore those registers around such calls. Such allocation only happens when it seems to result in better code.

This option is always enabled by default on certain machines — usually the ones that have no call-preserved registers to use instead.

Enabled at levels -O2, -O3, -Os.

• -fcombine-stack-adjustments — tracks stack adjustments (pushes and pops) and stack memory references, then tries to find ways to combine them. Enabled by default at -O1 and higher 🎀
• -fipa-ra — use caller-save registers for allocation if those registers aren't used by any called function. In that case you don't need to save and restore them around calls. This is only possible if the called functions are part of the same compilation unit as the current function and they're compiled before it.

Enabled at levels -O2, -O3, -Os. However, the option is disabled if the generated code is going to be instrumented for profiling (-p, or -pg), or if the callee's register usage can't be known exactly — which happens on targets that don't expose prologues and epilogues in RTL.

• -fconserve-stack — attempt to minimize stack usage. The compiler tries to use less stack space, even if that makes the program slower. This option implies setting the large-stack-frame parameter to 100 and the large-stack-frame-growth parameter to 400. Trade speed for space, we respect the sacrifice 🙈
• -ftree-reassoc — perform reassociation on trees. Enabled by default at -O1 and higher.
• -fcode-hoisting — perform code hoisting. Code hoisting tries to move the evaluation of expressions executed on all paths to the function exit as early as possible. Especially useful as a code size optimization, but it often helps code speed too. Enabled by default at -O2 and higher. Early bird gets the optimized code 🦋
• -ftree-pre — perform partial redundancy elimination (PRE) on trees. Enabled by default at -O2 and -O3.
• -ftree-partial-pre — make partial redundancy elimination (PRE) more aggressive. Enabled by default at -O3. She's not mad, she's just aggressive 💅
• -ftree-forwprop — perform forward propagation on trees. Enabled by default at -O1 and higher.
• -ftree-fre — bestie this one does full redundancy elimination (FRE) on trees 💖. Here's the tea: the whole difference between FRE and PRE is that FRE is pickier — it only looks at expressions that get computed on all paths leading to the redundant computation. So yeah, this analysis is faster than PRE, but the tradeoff is it exposes fewer redundancies. Enabled by default at -O1 and higher ✨.
• -ftree-phiprop — hoists loads from conditional pointers on trees, slay 🎀. This pass is on by default at -O1 and higher.
• -fhoist-adjacent-loads — okay so this speculatively hoists loads from both branches of an if-then-else, but only if the loads are from adjacent locations in the same structure and the target architecture has a conditional move instruction. Enabled by default at -O2 and higher 🦋.
• -ftree-copy-prop — copy propagation on trees, periodt. This pass eliminates unnecessary copy operations. Enabled by default at -O1 and higher ✨.
• -fipa-pure-const — discovers which functions are pure or constant 👑. Enabled by default at -O1 and higher.
• -fipa-reference — discovers which static variables do not escape the compilation unit 🌸. Enabled by default at -O1 and higher.
• -fipa-reference-addressable — discovers read-only, write-only and non-addressable static variables. Enabled by default at -O1 and higher 💜.
• -fipa-reorder-for-locality — this one groups call chains close together in the binary layout so code locality gets better and jump distances between frequently called functions get minimized 🔥. And like, it's different from -freorder-functions because this pass actually looks at the call chains between functions and groups them together, instead of shoving all hot/normal/cold/never-executed functions into separate sections. And it's different from -fprofile-reorder-functions because it wants to improve code locality throughout the entire runtime of the program, not just focus on program startup, iykyk. Heads up though: this option is incompatible with an explicit -flto-partition= option, because it enforces a custom partitioning scheme 😭. And if you're using this, it's recommended to also use profile feedback — but otherwise this option is not enabled by default.
• -fipa-stack-alignment — reduces stack alignment on call sites if possible. Enabled by default 💅.
• -fipa-pta — performs interprocedural pointer analysis plus interprocedural modification and reference analysis 🧚. Real talk: this option can cause excessive memory and compile-time usage on large compilation units. It is not enabled by default at any optimization level, nope, not even one.
• -fipa-profile — performs interprocedural profile propagation 💖. Functions that are called only from cold functions get marked as cold. Also, functions executed once (like "cold", "noreturn", static constructors or destructors) get identified. Then cold functions and the loop-less parts of functions executed once get optimized for size. Enabled by default at -O1 and higher ✨.
• -fipa-modref — performs interprocedural mod/ref analysis. This optimization analyzes the side effects of functions (the memory locations that get modified or referenced) and it enables better optimization across the function call boundary 🦋. Enabled by default at -O1 and higher.
• -fipa-cp — interprocedural constant propagation, the moment 💅. This optimization analyzes the program to figure out when values passed to functions are constants, and then optimizes accordingly. And it can substantially increase performance if your application has constants passed to functions. Enabled by default at -O2, -Os and -O3. It's also enabled by -fprofile-use and -fauto-profile 🔥.
• -fipa-cp-clone — performs function cloning to make interprocedural constant propagation stronger 💪. When enabled, interprocedural constant propagation performs function cloning when an externally visible function can be called with constant arguments. But bestie, because this optimization can create multiple copies of functions, it may significantly increase code size (see --param ipa-cp-unit-growth=value). Enabled by default at -O3. It's also enabled by -fprofile-use and -fauto-profile ✨.
• -fipa-bit-cp — when enabled, performs interprocedural bitwise constant propagation. Enabled by default at -O2 and by -fprofile-use and -fauto-profile. It requires that -fipa-cp is enabled 👑.
• -fipa-vrp — when enabled, performs interprocedural propagation of value ranges 🌈. Enabled by default at -O2. It requires that -fipa-cp is enabled too.
• -fipa-icf — performs Identical Code Folding for functions and read-only variables 🎀. This optimization reduces code size, but ngl it may disturb unwind stacks by replacing a function with an equivalent one that has a different name 😭. And the optimization works more effectively with link-time optimization enabled.

So even though the behavior is similar to the Gold Linker's ICF optimization, GCC ICF works on different levels, and therefore the optimizations are not the same — there are equivalences that are found only by GCC and equivalences found only by Gold 💀.

This flag is enabled by default at -O2 and -Os.

• -flate-combine-instructions — enables two instruction combination passes that run relatively late in the compilation process ⭐. One of the passes runs before register allocation and the other one runs after register allocation. The main aim of these passes is to substitute definitions into all uses. Most targets enable this flag by default at -O2 and -Os 💖.
• -flive-patching=level — controls GCC's optimizations so it produces output that's suitable for live-patching 🩹.

So here's the concept: if the compiler's optimization uses a function's body, or information extracted from its body, to optimize/change another function, then that latter function is called an impacted function of the former. And if a function gets patched, its impacted functions should be patched too, no cap.

The impacted functions are determined by the compiler's interprocedural optimizations. For example, a caller is impacted when inlining a function into its caller, cloning a function and changing its caller to call this new clone, or extracting a function's pureness/constness information to optimize its direct or indirect callers, etc. 😌

Usually the more IPA optimizations you enable, the larger the number of impacted functions for each function. So in order to control the number of impacted functions and more easily compute the list of impacted functions, IPA optimizations can be partially enabled at two different levels.

The level argument should be one of the following:

• inline-clone — only enables inlining and cloning optimizations, which includes inlining, cloning, interprocedural scalar replacement of aggregates and partial inlining. The consequence? When you patch a function, all its callers and its clones' callers are impacted, so they need to be patched as well 💅.

-flive-patching=inline-clone disables the following optimization flags: -fwhole-program -fipa-pta -fipa-reference -fipa-ra -fipa-icf -fipa-icf-functions -fipa-icf-variables -fipa-bit-cp -fipa-vrp -fipa-pure-const -fipa-reference-addressable -fipa-stack-alignment -fipa-modref

• inline-only-static — only enables inlining of static functions. So when you patch a static function, all its callers are impacted and therefore need to be patched as well 🦋.

On top of all the flags that -flive-patching=inline-clone disables, -flive-patching=inline-only-static disables these additional optimization flags: -fipa-cp-clone -fipa-sra -fpartial-inlining -fipa-cp

When -flive-patching is specified without any value, the default value is inline-clone ✨.

This flag is disabled by default.

Warning bestie: note that -flive-patching is not supported with link-time optimization (-flto) 😭.

• -fisolate-erroneous-paths-dereference — detects paths that trigger erroneous or undefined behavior because of dereferencing a null pointer 💀. It isolates those paths from the main control flow and turns the statement with erroneous or undefined behavior into a trap. Enabled by default at -O2 and higher, and it depends on -fdelete-null-pointer-checks also being enabled 🌸.
• -fisolate-erroneous-paths-attribute — detects paths that trigger erroneous or undefined behavior because a null value is being used in a way that's forbidden by a "returns_nonnull" or "nonnull" attribute 👀. Just like the other one, it isolates those paths from the main control flow and turns the statement with erroneous or undefined behavior into a trap. This is not currently enabled, but it may be enabled by -O2 in the future. Foreshadowing ✨.
• -ftree-sink — performs forward store motion on trees 💖. Enabled by default at -O1 and higher.
• -ftree-bit-ccp — performs sparse conditional bit constant propagation on trees and propagates pointer alignment information 🌈. Heads up: this pass only operates on local scalar variables, and it's enabled by default at -O1 and higher — except for -Og. It requires that -ftree-ccp is enabled 💅.
• -ftree-ccp — performs sparse conditional constant propagation (CCP) on trees 🎀. This pass only operates on local scalar variables and is enabled by default at -O1 and higher.
• -fssa-backprop — propagates information about uses of a value up the definition chain in order to simplify the definitions 🦋. Like, for example, this pass strips sign operations if the sign of a value never matters. Enabled by default at -O1 and higher ✨.
• -fssa-phiopt — performs pattern matching on SSA PHI nodes to optimize conditional code 💜. This pass is enabled by default at -O1 and higher, except for -Og.
• -ftree-switch-conversion — performs conversion of simple initializations in a switch into initializations from a scalar array 🔥. Enabled by default at -O2 and higher.
• -ftree-tail-merge — looks for identical code sequences 👀. When it finds one, it replaces one with a jump to the other. This optimization is known as tail merging or cross jumping. Enabled by default at -O2 and higher. And the compilation time in this pass can be limited using the max-tail-merge-comparisons parameter and the max-tail-merge-iterations parameter 👑
• -ftree-cselim — performs conditional store elimination on trees 💖. Enabled by default at -O1 and higher on targets that have conditional move instructions ✨.
• -ftree-dce — does dead code elimination (DCE) on trees, bestie. And it's on by default at -O1 and up, so it's just out here working for free 💅
• -ftree-builtin-call-dce — conditional dead code elimination (DCE) specifically for calls to built-in functions that <em>might</em> set errno but are otherwise free of side effects. This one's enabled by default at -O2 and higher — unless you also passed -Os. Then it sits this one out 🙈
• -ffinite-loops — tells the compiler to just trust that a loop with an exit will eventually take that exit, and not loop forever and ever into the void. Which means it's allowed to delete loops that have no side-effects, since it's no longer worried about them being an infinite vibe. 💀

This one's on by default at -O2 for C++ when you're on -std=c++11 or higher.

• -ftree-dominator-opts — the little cleanup crew. Does a whole bunch of simple scalar cleanups based on a dominator tree traversal: constant/copy propagation, redundancy elimination, range propagation and expression simplification. It also does jump threading (which is just reducing jumps-to-jumps, so like, less bouncing around). Enabled by default at -O1 and up ✨
• -ftree-dse — dead store elimination (DSE) on trees. A dead store is when you store into a memory location and then later overwrite it with another store without any loads happening in between — in which case the first store is pointless and can just be deleted. Rude but correct. Enabled by default at -O1 and up 🎀
• -ftree-ch — loop header copying on trees. It's a slay because it makes code motion optimizations work better, plus it saves one jump. Enabled by default at -O1 and up — but NOT for -Os, because it usually makes your code bigger 😭
• -ftree-loop-optimize — loop optimizations on trees. On by default at -O1 and up, periodt.
• -ftree-loop-linear, -floop-strip-mine, -floop-block — loop nest optimizations. Same thing as -floop-nest-optimize. Heads up though: to actually use this code transformation, GCC has to be configured with --with-isl to enable the Graphite loop transformation infrastructure. No isl, no transformation, iykyk 🦋
• -fgraphite-identity — enables the identity transformation for graphite. For every SCoP it generates the polyhedral representation and transforms it back to gimple. With -fgraphite-identity you can check the costs or benefits of the GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations also get performed by the code generator isl, like index splitting and dead code elimination in loops. It's giving very "round trip just to look at it" energy 💅
• -floop-nest-optimize — enables the isl based loop nest optimizer. It's a generic loop nest optimizer built on the Pluto optimization algorithms, and it calculates a loop structure optimized for data-locality and parallelism. This option is experimental, so like… know that going in 🫶
• -floop-parallelize-all — uses the Graphite data dependence analysis to find loops that can be parallelized, then parallelizes all the loops it can analyze as not containing loop carried dependences — and it does not check whether parallelizing them is actually profitable. It said go, no thoughts, no questions 🔥
• -ftree-coalesce-vars — while transforming your program out of the SSA representation, this tries to reduce copying by coalescing versions of different user-defined variables, instead of only compiler temporaries. Which honestly sounds great until you hear this: it may severely limit your ability to debug an optimized program compiled with -fno-var-tracking-assignments 😭 In its negated form, this flag prevents SSA coalescing of user variables. Enabled by default if optimization is enabled — and it does very little otherwise.
• -ftree-loop-if-convert — tries to transform conditional jumps in the innermost loops into branch-less equivalents. The whole intent is to get control-flow out of the innermost loops so the vectorization pass has an easier time handling them. Enabled by default if vectorization is enabled ✨
• -ftree-loop-distribution — does loop distribution. This can improve cache performance on big loop bodies AND let further loop optimizations (parallelization, vectorization) actually happen. Like, watch this loop:

        DO I = 1, N
          A(I) = B(I) + C
          D(I) = E(I) * F
        ENDDO

become this:

        DO I = 1, N
           A(I) = B(I) + C
        ENDDO
        DO I = 1, N
           D(I) = E(I) * F
        ENDDO

Enabled by default at -O3. It's also enabled by -fprofile-use and -fauto-profile 👑

• -ftree-loop-distribute-patterns — does loop distribution of patterns that can be code generated with calls to a library. Enabled by default at -O2 and higher, and by -fprofile-use and -fauto-profile.

This pass distributes the initialization loops and generates a call to memset zero. So this loop:

        DO I = 1, N
          A(I) = 0
          B(I) = A(I) + I
        ENDDO

becomes this:

        DO I = 1, N
           A(I) = 0
        ENDDO
        DO I = 1, N
           B(I) = A(I) + I
        ENDDO

and then the initialization loop gets turned into a call to memset zero. Chef's kiss 🎀

• -floop-interchange — does loop interchange outside of graphite. Can improve cache performance on loop nests and let further loop optimizations, like vectorization, take place. Example: this loop

        for (int i = 0; i < N; i++)
          for (int j = 0; j < N; j++)
            for (int k = 0; k < N; k++)
              c[i][j] = c[i][j] + a[i][k]*b[k][j];

gets transformed into

        for (int i = 0; i < N; i++)
          for (int k = 0; k < N; k++)
            for (int j = 0; j < N; j++)
              c[i][j] = c[i][j] + a[i][k]*b[k][j];

Enabled by default at -O3. Also enabled by -fprofile-use and -fauto-profile 💖

• -floop-unroll-and-jam — applies unroll and jam transformations on feasible loops. In a loop nest, this unrolls the outer loop by some factor and then fuses the resulting multiple inner loops together. Enabled by default at -O3. Also enabled by -fprofile-use and -fauto-profile 🌸
• -ftree-loop-im — does loop invariant motion on trees. Lowkey picky though: this pass only moves invariants that are hard to handle at the RTL level — think function calls, and operations that expand into nontrivial sequences of insns. With -funswitch-loops it also moves operands of conditions that are invariant out of the loop, so loop unswitching can get away with just trivial invariantness analysis. The pass also includes store motion 🦋
• -ftree-loop-ivcanon — creates a canonical counter for the number of iterations in loops where figuring out the iteration count takes complicated analysis. Later optimizations can then determine the number easily. Especially useful in connection with unrolling ⭐
• -ftree-scev-cprop — does final value replacement. Here's the tea: if a variable is modified in a loop such that its value on exit can be determined using only its initial value and the number of loop iterations, then uses of that final value get replaced by that computation — provided it's sufficiently cheap. This reduces data dependencies and may allow further simplifications. Enabled by default at -O1 and higher 🫶
• -fivopts — does induction variable optimizations on trees: strength reduction, induction variable merging and induction variable elimination. Enabled by default at -O1 and higher 👑
• -ftree-parallelize-loops=n — parallelizes loops, i.e. it splits their iteration space so they run across n threads. Only possible for loops whose iterations are independent and can be arbitrarily reordered. And honestly it's only profitable on multiprocessor machines, for loops that are CPU-intensive — not loops constrained by stuff like memory bandwidth. Heads up: this option implies -pthread, so it's only supported on targets that have support for -pthread 🔥
• -ftree-pta — does function-local points-to analysis on trees. Enabled by default at -O1 and higher — except for -Og 🙈
• -ftree-sra — does scalar replacement of aggregates. This pass replaces structure references with scalars so structures don't get committed to memory too early. Enabled by default at -O1 and higher — again except for -Og 💅
• -fstore-merging — does merging of narrow stores to consecutive memory addresses. It takes contiguous stores of immediate values narrower than a word and merges them into fewer, wider stores to cut down the number of instructions. Enabled by default at -O2 and higher, as well as -Os ✨
• -ftree-ter — does temporary expression replacement during the SSA->normal phase. Single use/single def temporaries get replaced at their use location with their defining expression. Yes, this results in non-GIMPLE code, but it gives the expanders much more complex trees to work on, which results in better RTL generation. Enabled by default at -O1 and higher 🎀
• -ftree-slsr — does straight-line strength reduction on trees. It recognizes related expressions involving multiplications and swaps them for less expensive calculations when it can. Enabled by default at -O1 and higher 🌈
• -ftree-vectorize — does vectorization on trees. This flag enables -ftree-loop-vectorize and -ftree-slp-vectorize if they weren't explicitly specified 🔥
• -ftree-loop-vectorize — does loop vectorization on trees. Enabled by default at -O2, and by -ftree-vectorize, -fprofile-use, and -fauto-profile 💜
• -ftree-slp-vectorize — does basic block vectorization on trees. Enabled by default at -O2, and by -ftree-vectorize, -fprofile-use, and -fauto-profile 🍓
• -ftrivial-auto-var-init=choice — ok so bestie, this one either gives your automatic variables a pattern or zeroes them out, and the whole point is to make your program more secure and more predictable by preventing uninitialized memory disclosure and use 💖. But here's the thing — GCC still treats an automatic variable with no explicit initializer as uninitialized, so -Wuninitialized and -Wanalyzer-use-of-uninitialized-value will still hit you with warning messages on those automatic variables, and the compiler will optimize exactly as if the variable were never initialized. Period. With this option on, GCC will also zero out any padding of automatic variables that have structure or union types. However — and this is the messy corner, ngl — the current implementation cannot initialize automatic variables that are declared between the controlling expression and the first case of a switch statement. Use -Wtrivial-auto-var-init to report all such cases, k? 🙈

The three values of choice are:

• uninitialized — doesn't initialize any automatic variables. This is C and C++'s default, iconic behavior, no notes.
• pattern — initializes automatic variables with values which will likely transform logic bugs into crashes down the line, are easily recognized in a crash dump and without being values that programmers can rely on for useful program semantics. The current value is byte-repeatable pattern with byte 0xFE. The values used for pattern initialization might be changed in the future, so don't get attached 💅
• zero — initialize automatic variables with zeroes.

The default is uninitialized.

And like, a super important note: the initializer values, whether zero or pattern, refer to data representation (in memory or machine registers), rather than to their interpretation as numerical values. This distinction may be important in languages that support types with biases or implicit multipliers, and with such extensions as hardbool. For example, a variable that uses 8 bits to represent (biased) quantities in the "range 160..400" will be initialized with the bit patterns 0x00 or 0xFE, depending on choice, whether or not these representations stand for values in that range, and even if they do, the interpretation of the value held by the variable will depend on the bias. A hardbool variable that uses say "0X5A" and 0xA5 for "false" and "true", respectively, will trap with either choice of trivial initializer — i.e., zero initialization will not convert to the representation for "false", even if it would for a "static" variable of the same type 😭. This means the initializer pattern doesn't generally depend on the type of the initialized variable. One notable exception though: (non-hardened) boolean variables that fit in registers are initialized with "false" (zero), even when pattern is requested. Slay, I guess.

You can control this behavior for a specific variable by using the variable attribute uninitialized ✨

• -fvect-cost-model=model — this changes up the cost model used for vectorization, bestie. The model argument should be one of unlimited, dynamic, cheap or very-cheap. With the unlimited model the vectorized code-path is assumed to be profitable, while with the dynamic model a runtime check guards the vectorized code-path to enable it only for iteration counts that will likely execute faster than when executing the original scalar loop. The cheap model disables vectorization of loops where doing so would be cost prohibitive — for example due to required runtime checks for data dependence or alignment — but otherwise is equal to the dynamic model. The very-cheap model disables vectorization of loops when any runtime check for data dependence or alignment is required, and it also disables vectorization of epilogue loops, but otherwise is equal to the cheap model.

The default cost model depends on other optimization flags and is either dynamic or cheap. iykyk 🌈

• -fsimd-cost-model=model — same energy but for the vectorization of loops marked with the OpenMP simd directive 💫. The model argument should be one of unlimited, dynamic, cheap. All values of model have the same meaning as described in -fvect-cost-model, and by default a cost model defined with -fvect-cost-model is used.
• -ftree-vrp — Perform Value Range Propagation on trees. It's similar to the constant propagation pass, but instead of values, ranges of values are propagated. That lets the optimizers remove unnecessary range checks like array bound checks and null pointer checks — obsessed 🫶. This is enabled by default at -O2 and higher. Null pointer check elimination is only done if -fdelete-null-pointer-checks is enabled, so keep that in mind.
• -fsplit-paths — Split paths leading to loop backedges. This can improve dead code elimination and common subexpression elimination. Enabled by default at -O3 and above ✨
• -fsplit-ivs-in-unroller — Enables expression of values of induction variables in later iterations of the unrolled loop using the value in the first iteration. This breaks long dependency chains, thus improving efficiency of the scheduling passes 💅

A combination of -fweb and CSE is often sufficient to obtain the same effect. However, that is not reliable in cases where the loop body is more complicated than a single basic block. It also does not work at all on some architectures due to restrictions in the CSE pass. RIP to those architectures I guess 😭

This optimization is enabled by default.

• -fvariable-expansion-in-unroller — With this option, the compiler creates multiple copies of some local variables when unrolling a loop, which can result in superior code. Main character move fr fr ✨

This optimization is enabled by default for PowerPC targets, but disabled by default otherwise.

• -fpartial-inlining — Inline parts of functions, bestie. Just the parts. This option has any effect only when inlining itself is turned on by the -finline-functions or -finline-small-functions options.

Enabled at levels -O2, -O3, -Os.

• -fpredictive-commoning — Perform predictive commoning optimization, i.e., reusing computations (especially memory loads and stores) performed in previous iterations of loops. Recycling but make it compiler internals 🎀

This option is enabled at level -O3. It is also enabled by -fprofile-use and -fauto-profile.

• -fprefetch-loop-arrays — If supported by the target machine, generate instructions to prefetch memory to improve the performance of loops that access large arrays. Manifesting faster loops 🦋

This option may generate better or worse code; results are highly dependent on the structure of loops within the source code. So like, your mileage very much varies.

Disabled at level -Os.

• -fno-printf-return-value — Do not substitute constants for known return value of formatted output functions such as sprintf, snprintf, vsprintf, and vsnprintf (but not printf of fprintf). This transformation allows GCC to optimize or even eliminate branches based on the known return value of these functions called with arguments that are either constant, or whose values are known to be in a range that makes determining the exact return value possible. For example, when -fprintf-return-value is in effect, both the branch and the body of the if statement (but not the call to snprint) can be optimized away when i is a 32-bit or smaller integer because the return value is guaranteed to be at most 8. Like:

char buf[9];
if (snprintf (buf, "%08x", i) >= sizeof buf)
  ...

The -fprintf-return-value option relies on other optimizations and yields best results with -O2 and above. It works in tandem with the -Wformat-overflow and -Wformat-truncation options. The -fprintf-return-value option is enabled by default.

• -fno-peephole
• -fno-peephole2 — Disable any machine-specific peephole optimizations 🚫. The difference between -fno-peephole and -fno-peephole2 is in how they are implemented in the compiler; some targets use one, some use the other, a few use both. Cute.

-fpeephole is enabled by default. -fpeephole2 enabled at levels -O2, -O3, -Os.

• -fno-guess-branch-probability — Do not guess branch probabilities using heuristics, bestie. No vibes-based guessing today 💀

GCC uses heuristics to guess branch probabilities if they are not provided by profiling feedback (-fprofile-arcs). These heuristics are based on the control flow graph. If some branch probabilities are specified by __builtin_expect, then the heuristics are used to guess branch probabilities for the rest of the control flow graph, taking the __builtin_expect info into account. The interactions between the heuristics and __builtin_expect can be complex, and in some cases, it may be useful to disable the heuristics so that the effects of __builtin_expect are easier to understand. Fr, sometimes you just gotta turn off the guessing to see what's actually happening 👑

It is also possible to specify expected probability of the expression with __builtin_expect_with_probability built-in function.

The default is -fguess-branch-probability at levels -O, -O2, -O3, -Os.

• -freorder-blocks — Reorder basic blocks in the compiled function in order to reduce number of taken branches and improve code locality. Just shuffling the blocks around for maximum slay ✨

Enabled at levels -O1, -O2, -O3, -Os.

• -freorder-blocks-algorithm=algorithm — Use the specified algorithm for basic block reordering. The algorithm argument can be simple, which does not increase code size (except sometimes due to secondary effects like alignment), or stc, the "software trace cache" algorithm, which tries to put all often executed code together, minimizing the number of branches executed by making extra copies of code.

The default is simple at levels -O1, -Os, and stc at levels -O2, -O3.

• -freorder-blocks-and-partition — In addition to reordering basic blocks in the compiled function, in order to reduce number of taken branches, this partitions hot and cold basic blocks into separate sections of the assembly and .o files, to improve paging and cache locality performance. Hot girl code and cold girl code living in their own little neighborhoods 💖

This optimization is automatically turned off in the presence of exception handling or unwind tables (on targets using setjump/longjump or target specific scheme), for linkonce sections, for functions with a user-defined section attribute and on any architecture that does not support named sections. When -fsplit-stack is used this option is not enabled by default (to avoid linker errors), but may be enabled explicitly (if using a working linker).

Enabled for x86 at levels -O2, -O3, -Os.

---

• -freorder-functions — bestie, this one literally just shuffles your functions around inside the object file so the code that lives near each other actually hangs out together. 💅 And unlike -fipa-reorder-for-locality (its whole different situation), this one prioritises gluing every function in a category together — hot, normal, cold, never-executed, all in their own little cliques. How? By dropping them into special subsections: .text.hot for the most frequently executed functions and .text.unlikely for the ones that, respectfully, almost never run. The actual reordering is done by the linker, so your object file format has to support named sections and the linker has to place them somewhere reasonable. 👑

Heads up: this option isn't effective unless you either hand over profile feedback (peep -fprofile-arcs for the deets) or manually annotate your functions with "hot" or "cold" attributes. No data, no glow-up.

Enabled at levels -O2, -O3, -Os. ✨

• -fstrict-aliasing — okay so this lets the compiler assume the strictest aliasing rules that apply to whatever language you're compiling. For C (and C++), that means it turns on optimizations based on the type of your expressions. Specifically: an object of one type is assumed to NEVER live at the same address as an object of a different type — unless the types are almost the same. Example: an "unsigned int" can alias an "int", but it can NOT alias a "void*" or a "double". And a character type may alias any other type. 🦋

Pay special attention to code like this, bestie:

        union a_union {
          int i;
          double d;
        };

        int f() {
          union a_union t;
          t.d = 3.0;
          return t.i;
        }

The practice of reading from a different union member than the one most recently written to (we call that "type-punning") is, like, super common. And even with -fstrict-aliasing, type-punning is allowed — provided the memory is accessed through the union type. So the code above works exactly as expected. 🫶 However, this code might not:

        int f() {
          union a_union t;
          int* ip;
          t.d = 3.0;
          ip = &t.i;
          return *ip;
        }

Similarly (and this is the part where it gets messy 😭), access by taking the address, casting the resulting pointer and dereferencing the result has undefined behavior — even if the cast uses a union type, like:

        int f() {
          double d = 3.0;
          return ((union a_union *) &d)->i;
        }

The -fstrict-aliasing option is enabled at levels -O2, -O3, -Os.

• -fipa-strict-aliasing — this one controls whether the rules of -fstrict-aliasing get applied across function boundaries. Fun lil note: if multiple functions get inlined into a single function, the memory accesses are no longer considered to be crossing a function boundary. ✨

-fipa-strict-aliasing is enabled by default and is effective only in combination with -fstrict-aliasing.

---

• -falign-functions
• -falign-functions=n
• -falign-functions=n:m
• -falign-functions=n:m:n2
• -falign-functions=n:m:n2:m2 — align the start of functions to the next power-of-two greater than or equal to n, skipping up to m-1 bytes. This ensures that at least the first m bytes of the function can be fetched by the CPU without crossing an n-byte alignment boundary. It's a code-performance optimization, and alignment is ignored for functions considered cold. If alignment is required for all functions, use -fmin-function-alignment instead. 💅

If m is not specified, it defaults to n.

Examples, since examples are everything: -falign-functions=32 aligns functions to the next 32-byte boundary, -falign-functions=24 aligns to the next 32-byte boundary only if this can be done by skipping 23 bytes or less, -falign-functions=32:7 aligns to the next 32-byte boundary only if this can be done by skipping 6 bytes or less.

The second pair of n2:m2 values lets you specify a secondary alignment: -falign-functions=64:7:32:3 aligns to the next 64-byte boundary if this can be done by skipping 6 bytes or less, otherwise it aligns to the next 32-byte boundary if this can be done by skipping 2 bytes or less. If m2 is not specified, it defaults to n2. 🎀

Some assemblers only support this flag when n is a power of two; in that case, it's rounded up.

-fno-align-functions and -falign-functions=1 are equivalent and mean functions are not aligned.

If n is not specified or is zero, use a machine-dependent default. The maximum allowed n option value is 65536.

Enabled at levels -O2, -O3. ⭐

• -flimit-function-alignment — turn this on and the compiler tries to avoid unnecessarily overaligning your functions. It attempts to instruct the assembler to align by the amount specified by -falign-functions, but not to skip more bytes than the size of the function. Respectful. 👑
• -falign-labels
• -falign-labels=n
• -falign-labels=n:m
• -falign-labels=n:m:n2
• -falign-labels=n:m:n2:m2 — align all branch targets to a power-of-two boundary. ✨

The parameters of this option are analogous to the -falign-functions option. -fno-align-labels and -falign-labels=1 are equivalent and mean labels are not aligned.

If -falign-loops or -falign-jumps are applicable and are greater than this value, then their values are used instead.

If n is not specified or is zero, use a machine-dependent default which is very likely to be 1, meaning no alignment. The maximum allowed n option value is 65536.

Enabled at levels -O2, -O3. 🌸

• -falign-loops
• -falign-loops=n
• -falign-loops=n:m
• -falign-loops=n:m:n2
• -falign-loops=n:m:n2:m2 — aligns loops to a power-of-two boundary. If the loops get executed many times, that makes up for any execution of the dummy padding instructions. It's a code-performance optimization, and alignment is ignored for loops considered cold. 🥶

If -falign-labels is greater than this value, then its value is used instead.

Parameters of this option are analogous to the -falign-functions option. -fno-align-loops and -falign-loops=1 are equivalent and mean loops are not aligned. The maximum allowed n option value is 65536.

If n is not specified or is zero, use a machine-dependent default.

Enabled at levels -O2, -O3. 💜

• -falign-jumps
• -falign-jumps=n
• -falign-jumps=n:m
• -falign-jumps=n:m:n2
• -falign-jumps=n:m:n2:m2 — aligns branch targets to a power-of-two boundary, specifically for branch targets where the targets can only be reached by jumping. In that case, no dummy operations need be executed. It's a code-performance optimization, and alignment is ignored for jumps considered cold. 🦋

If -falign-labels is greater than this value, then its value is used instead.

Parameters of this option are analogous to the -falign-functions option. -fno-align-jumps and -falign-jumps=1 are equivalent and mean loops are not aligned.

If n is not specified or is zero, use a machine-dependent default. The maximum allowed n option value is 65536.

Enabled at levels -O2, -O3. ✨

• -fmin-function-alignment — specify the minimal alignment of functions to the next power-of-two greater than or equal to n. Unlike -falign-functions, this alignment is applied to all functions (yes, even the cold ones, no one gets left out 💖). The alignment is also not affected by -flimit-function-alignment.
• -fno-allocation-dce — do NOT remove unused C++ allocations (the ones using operator "new" and operator "delete") during dead code elimination. Leave those allocations alone. 🙈

See also -fmalloc-dce.

• -fallow-store-data-races — lets the compiler perform optimizations that may introduce new data races on stores, without proving the variable can't be concurrently accessed by other threads. It does not affect optimization of local data. It's safe to use this option if it's known that global data will not be accessed by multiple threads. 👑

Examples of optimizations enabled by -fallow-store-data-races include hoisting or if-conversions that may cause a value that was already in memory to be re-written with that same value. Such re-writing is safe in a single-threaded context but may be unsafe in a multi-threaded context. Note that on some processors, if-conversions may be required in order to enable vectorization. ⚠️

Enabled at level -Ofast.

• -funit-at-a-time — this option is left for compatibility reasons only, bestie. -funit-at-a-time has no effect, while -fno-unit-at-a-time implies -fno-toplevel-reorder and -fno-section-anchors. 💀

Enabled by default.

• -fno-toplevel-reorder — do not reorder top-level functions, variables, and "asm" statements. Output them in the same order they appear in the input file. When this option is used, unreferenced static variables are not removed. This option is intended to support existing code that relies on a particular ordering. For new code, it's better to use attributes when possible. 🎀

-ftoplevel-reorder is the default at -O1 and higher, and also at -O0 if -fsection-anchors is explicitly requested. Additionally, -fno-toplevel-reorder implies -fno-section-anchors.

• -funreachable-traps — with this option, the compiler turns calls to "__builtin_unreachable" into traps, instead of using them for optimization. This also affects any such calls implicitly generated by the compiler. 🪤

This option has the same effect as -fsanitize=unreachable -fsanitize-trap=unreachable, but does not affect the values of those options. If -fsanitize=unreachable is enabled, that option takes priority over this one. 👑

This option is enabled by default at -O0 and -Og.

• -fweb — constructs webs the way they're commonly used for register allocation purposes, and assigns each web its own individual pseudo register. This lets the register allocation pass operate on pseudos directly, but it also strengthens several other optimization passes — like CSE, the loop optimizer, and the trivial dead code remover. 💪 It can, however, make debugging impossible, since variables no longer stay in a "home register". 😭

Enabled by default with -funroll-loops.

• -fwhole-program — assume that the current compilation unit represents the whole program being compiled. All public functions and variables — with the exception of "main" and those merged by the attribute "externally_visible" — become static functions and, in effect, get optimized way more aggressively by interprocedural optimizers. 🌈

so, real talk — with -flto in the mix, this option right here has, like, limited usefulness 🙈 in most cases the precise list of symbols used or exported from the binary is already known from the resolution info that gets passed to the link-time optimizer by the linker plugin. it's still lowkey handy if no linker plugin is being used, or during the incremental link step when final code is produced (that's with -flto -flinker-output=nolto-rel).

---

💅 -flto[=n] ✨ — the whole link-time optimizer moment

okay so THIS one runs the standard link-time optimizer. 👑 when you invoke it with source code, it generates GIMPLE (that's one of GCC's internal representations, iykyk) and writes it to special ELF sections in the object file. then when the object files get linked together, all the function bodies are read back out of those ELF sections and instantiated as if they had been part of the same translation unit the whole time. it's giving one big file energy. 🦋

to actually use the link-time optimizer, -flto and your optimization options should be specified at compile time AND during the final link. it is recommended that you compile all the files participating in the same link with the same options, and also specify those options at link time. for example:

gcc -c -O2 -flto foo.c
gcc -c -O2 -flto bar.c
gcc -o myprog -flto -O2 foo.o bar.o

the first two invocations to GCC save a bytecode representation of GIMPLE into special ELF sections inside foo.o and bar.o. the final invocation reads the GIMPLE bytecode from foo.o and bar.o, merges the two files into a single internal image, and compiles the result as usual. since both foo.o and bar.o are merged into a single image, this causes all the interprocedural analyses and optimizations in GCC to work across the two files as if they were a single one. which means, for example, that the inliner is able to inline functions in bar.o into functions in foo.o and vice-versa. periodt. 💖

another (simpler) way to enable link-time optimization is:

gcc -o myprog -flto -O2 foo.c bar.c

the above generates bytecode for foo.c and bar.c, merges them together into a single GIMPLE representation and optimizes them as usual to produce myprog.

the important thing to keep in mind 🧠 (fr fr) is that to enable link-time optimizations you need to use the GCC driver to perform the link step. GCC automatically performs link-time optimization if any of the objects involved were compiled with the -flto command-line option. you can always override the automatic decision to do link-time optimization by passing -fno-lto to the link command.

to make whole program optimization effective, it is necessary to make certain whole program assumptions. the compiler needs to know what functions and variables can be accessed by libraries and runtime outside of the link-time optimized unit. when supported by the linker, the linker plugin (see -fuse-linker-plugin) passes information to the compiler about used and externally visible symbols. when the linker plugin is not available, -fwhole-program should be used to allow the compiler to make these assumptions, which leads to more aggressive optimization decisions. 🔥

when a file is compiled with -flto without -fuse-linker-plugin, the generated object file is larger than a regular object file, because it contains GIMPLE bytecodes and the usual final code (see -ffat-lto-objects). this means that object files with LTO information can be linked as normal object files; if -fno-lto is passed to the linker, no interprocedural optimizations are applied. note that when -fno-fat-lto-objects is enabled the compile stage is faster but you cannot perform a regular, non-LTO link on them. 😭 (pick your struggle, bestie)

when producing the final binary, GCC only applies link-time optimizations to those files that contain bytecode. therefore, you can mix and match object files and libraries with GIMPLE bytecodes and final object code. GCC automatically selects which files to optimize in LTO mode and which files to link without further processing. 🎀

generally, options specified at link time override those specified at compile time, although in some cases GCC attempts to infer link-time options from the settings used to compile the input files.

if you do not specify an optimization level option -O at link time, then GCC uses the highest optimization level used when compiling the object files. note that it is generally ineffective to specify an optimization level option only at link time and not at compile time, for two reasons. first, compiling without optimization suppresses compiler passes that gather information needed for effective optimization at link time. second, some early optimization passes can be performed only at compile time and not at link time. so like, do it at compile time too. 💅

---

there are some code generation flags preserved by GCC when generating bytecodes, as they need to be used during the final link. currently, the following options and their settings are taken from the first object file that explicitly specifies them: -fcommon, -fexceptions, -fnon-call-exceptions, -fgnu-tm and all the -m target flags.

the following options -fPIC, -fpic, -fpie and -fPIE are combined based on the following scheme:

B<-fPIC> + B<-fpic> = B<-fpic>
B<-fPIC> + B<-fno-pic> = B<-fno-pic>
B<-fpic/-fPIC> + (no option) = (no option)
B<-fPIC> + B<-fPIE> = B<-fPIE>
B<-fpic> + B<-fPIE> = B<-fpie>
B<-fPIC/-fpic> + B<-fpie> = B<-fpie>

certain ABI-changing flags are required to match in all compilation units, and trying to override this at link time with a conflicting value is ignored. this includes options such as -freg-struct-return and -fpcc-struct-return. 🚫

other options such as -ffp-contract, -fno-strict-overflow, -fwrapv, -fno-trapv or -fno-strict-aliasing are passed through to the link stage and merged conservatively for conflicting translation units. specifically -fno-strict-overflow, -fwrapv and -fno-trapv take precedence; and for example -ffp-contract=off takes precedence over -ffp-contract=fast. you can override them at link time.

diagnostic options such as -Wstringop-overflow are passed through to the link stage and their setting matches that of the compile-step at function granularity. note that this matters only for diagnostics emitted during optimization. note that code transforms such as inlining can lead to warnings being enabled or disabled for regions if code not consistent with the setting at compile time. 🙈

when you need to pass options to the assembler via -Wa or -Xassembler make sure to either compile such translation units with -fno-lto or consistently use the same assembler options on all translation units. you can alternatively also specify assembler options at LTO link time.

to enable debug info generation you need to supply -g at compile time. if any of the input files at link time were built with debug info generation enabled the link will enable debug info generation as well. any elaborate debug info settings like the dwarf level -gdwarf-5 need to be explicitly repeated at the linker command line and mixing different settings in different translation units is discouraged. (don't do it, bestie, it's not the vibe 😤)

---

if LTO encounters objects with C linkage declared with incompatible types in separate translation units to be linked together (undefined behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be issued. the behavior is still undefined at run time. similar diagnostics may be raised for other languages. 💀

another feature of LTO is that it is possible to apply interprocedural optimizations on files written in different languages:

gcc -c -flto foo.c
g++ -c -flto bar.cc
gfortran -c -flto baz.f90
g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran

notice that the final link is done with g++ to get the C++ runtime libraries and -lgfortran is added to get the Fortran runtime libraries. in general, when mixing languages in LTO mode, you should use the same link command options as when mixing languages in a regular (non-LTO) compilation. 🌈

if object files containing GIMPLE bytecode are stored in a library archive, say libfoo.a, it is possible to extract and use them in an LTO link if you are using a linker with plugin support. to create static libraries suitable for LTO, use gcc-ar and gcc-ranlib instead of ar and ranlib; to show the symbols of object files with GIMPLE bytecode, use gcc-nm. those commands require that ar, ranlib and nm have been compiled with plugin support. at link time, use the flag -fuse-linker-plugin to ensure that the library participates in the LTO optimization process:

gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo

with the linker plugin enabled, the linker extracts the needed GIMPLE files from libfoo.a and passes them on to the running GCC to make them part of the aggregated GIMPLE image to be optimized. slay. ✨

if you are not using a linker with plugin support and/or do not enable the linker plugin, then the objects inside libfoo.a are extracted and linked as usual, but they do not participate in the LTO optimization process. in order to make a static library suitable for both LTO optimization and usual linkage, compile its object files with -flto -ffat-lto-objects.

---

link-time optimizations do not require the presence of the whole program to operate. if the program does not require any symbols to be exported, it is possible to combine -flto and -fwhole-program to allow the interprocedural optimizers to use more aggressive assumptions which may lead to improved optimization opportunities. use of -fwhole-program is not needed when linker plugin is active (see -fuse-linker-plugin).

the current implementation of LTO makes no attempt to generate bytecode that is portable between different types of hosts. the bytecode files are versioned and there is a strict version check, so bytecode files generated in one version of GCC do not work with an older or newer version of GCC. ⚠️ like, zero backwards or forwards compatibility, it's giving strict.

link-time optimization does not work well with generation of debugging information on systems other than those using a combination of ELF and DWARF.

Okay so like, if you pass the optional n, the optimization and code generation that happens at link time runs in parallel — n parallel jobs, bestie, actual multitasking main character energy ✨ — by utilizing an installed make program. And the environment variable MAKE may be used to override the program that gets used. 🫶

You can also say -flto=jobserver and use GNU make's job server mode to figure out how many parallel jobs to run. This is so useful when the Makefile that's calling GCC is already executing in parallel. But — and hear me out — you must prepend a + to the command recipe in the parent Makefile for this to work. Also this option likely only works if MAKE is GNU make. And even without the option value, GCC tries to automatically detect a running GNU make's job server. Iconic behavior tbh. 💅

Use -flto=auto to use GNU make's job server if one is available, or otherwise fall back to autodetection of the number of CPU threads present in your system. Slay. 🌸

• -flto-partition=alg — ok so this one specifies the partitioning algorithm used by the link-time optimizer 👑. The value is one of:
  • 1to1 — partitioning that mirrors the original source files.
  • balanced — partitioning into equally sized chunks, whenever possible.
  • max — create a new partition for every symbol, where possible.
  • cache — balance chunk sizes while keeping related symbols together, for better caching in incremental LTO.
  • Specifying none as an algorithm disables partitioning and streaming completely.

The default value is balanced. While 1to1 can be used as a workaround for various code ordering issues, the max partitioning is intended for internal testing only — do not come for it, it's not for you. The value one specifies that exactly one partition should be used, while the value none bypasses partitioning and executes the link-time optimization step directly from the WPA phase. Wild, I know. 😭

• -flto-incremental=path — enable incremental LTO, with its cache living in the given existing directory. Can significantly shorten edit-compile cycles with LTO. Fr fr. When used with LTO (-flto), the output of translation units inside LTO gets cached. Cached translation units are likely to be encountered again when you recompile with small code changes, which leads to recompile time reduction. And multiple GCC instances can use the same cache in parallel — no cap. 🧚
• -flto-incremental-cache-size=n — specifies the number of cache entries in incremental LTO after which it prunes old entries. Heads up though, this is a soft limit, so temporarily there may be more entries. 🙈
• -flto-compression-level=n — this option specifies the level of compression used for the intermediate language written to LTO object files, and it's only meaningful in conjunction with LTO mode (-flto). GCC currently supports two LTO compression algorithms. For zstd, valid values are 0 (no compression) to 19 (maximum compression), while zlib supports values from 0 to 9. Values outside this range are clamped to either the minimum or maximum of the supported values. If the option isn't given, a default balanced compression setting is used. Balance, bestie. ⭐
• -fuse-linker-plugin — enables the use of a linker plugin during link-time optimization. This option relies on plugin support in the linker, which is available in gold or in GNU ld 2.21 or newer. This option enables the extraction of object files with GIMPLE bytecode out of library archives. This improves the quality of optimization by exposing more code to the link-time optimizer 🔥. This information specifies what symbols can be accessed externally (by non-LTO object or during dynamic linking). Resulting code quality improvements on binaries (and shared libraries that use hidden visibility) are similar to -fwhole-program. See -flto for a description of the effect of this flag and how to use it. This option is enabled by default when LTO support in GCC is enabled and GCC was configured for use with a linker supporting plugins (GNU ld 2.21 or newer or gold).
• -ffat-lto-objects — Fat LTO objects are object files that contain both the intermediate language and the object code 💖. This makes them usable for both LTO linking and normal linking. This option is effective only when compiling with -flto and is ignored at link time.
  • -fno-fat-lto-objects improves compilation time over plain LTO, but it requires the complete toolchain to be aware of LTO. It requires a linker with linker plugin support for basic functionality. Additionally, nm, ar and ranlib need to support linker plugins to allow a full-featured build environment (capable of building static libraries etc). GCC provides the gcc-ar, gcc-nm, gcc-ranlib wrappers to pass the right options to these tools. With non fat LTO, makefiles need to be modified to use them.
  • Note that modern binutils provide a plugin auto-load mechanism. Installing the linker plugin into $libdir/bfd-plugins has the same effect as usage of the command wrappers (gcc-ar, gcc-nm and gcc-ranlib).
  • The default is -fno-fat-lto-objects on targets with linker plugin support.
• -fcompare-elim — after register allocation and post-register allocation instruction splitting, identify arithmetic instructions that compute processor flags similar to a comparison operation based on that arithmetic. If possible, eliminate the explicit comparison operation. This pass only applies to certain targets that cannot explicitly represent the comparison operation before register allocation is complete. Enabled at levels -O1, -O2, -O3, -Os.
• -ffold-mem-offsets, -fno-fold-mem-offsets — try to eliminate add instructions by folding them in memory loads/stores. Enabled at levels -O2, -O3.

NOTE: On Debian/Ubuntu this is disabled by default on m68k-linux-gnu. 👀

• -fcprop-registers — after register allocation and post-register allocation instruction splitting, perform a copy-propagation pass to try to reduce scheduling dependencies and occasionally eliminate the copy. Enabled at levels -O1, -O2, -O3, -Os.
• -fprofile-correction — profiles collected using an instrumented binary for multi-threaded programs may be inconsistent because of missed counter updates 😵. When this option is specified, GCC uses heuristics to correct or smooth out such inconsistencies. By default, GCC emits an error message when an inconsistent profile is detected. This option is enabled by -fauto-profile.
• -fprofile-partial-training — with -fprofile-use all portions of programs not executed during the train run are optimized aggressively for size rather than speed. In some cases it's just not practical to train all possible hot paths in the program. (For example, the program may contain functions specific for a given hardware, and the training may not cover all hardware configurations the program is run on.) With -fprofile-partial-training, profile feedback is ignored for all functions not executed during the train run, leading them to be optimized as if they were compiled without profile feedback. This leads to better performance when the train run isn't representative, but it also leads to significantly bigger code. Tradeoffs, bestie. 😭
• -fprofile-use, -fprofile-use=path — enable profile feedback-directed optimizations, and the following optimizations, many of which are generally profitable only with profile feedback available:
  • -fbranch-probabilities, -fprofile-values, -funroll-loops, -fpeel-loops, -ftracer, -fvpt, -finline-functions, -fipa-cp, -fipa-cp-clone, -fipa-bit-cp, -fpredictive-commoning, -fsplit-loops, -funswitch-loops, -fgcse-after-reload, -ftree-loop-vectorize, -ftree-slp-vectorize, -fvect-cost-model=dynamic, -ftree-loop-distribute-patterns, -fprofile-reorder-functions

Before you can use this option, you must first generate profiling information. By default, GCC emits an error message if the feedback profiles do not match the source code. This error can be turned into a warning by using -Wno-error=coverage-mismatch. Note that this may result in poorly optimized code. Additionally, by default, GCC also emits a warning message if the feedback profiles do not exist (see -Wmissing-profile). If path is specified, GCC looks at path to find the profile feedback data files. See -fprofile-dir.

• -fauto-profile, -fauto-profile=path — enable sampling-based feedback-directed optimizations, and the following optimizations, many of which are generally profitable only with profile feedback available:
  • -fbranch-probabilities, -fprofile-values, -funroll-loops, -fpeel-loops, -ftracer, -fvpt, -finline-functions, -fipa-cp, -fipa-cp-clone, -fipa-bit-cp, -fpredictive-commoning, -fsplit-loops, -funswitch-loops, -fgcse-after-reload, -ftree-loop-vectorize, -ftree-slp-vectorize, -fvect-cost-model=dynamic, -ftree-loop-distribute-patterns, -fprofile-correction

path is the name of a file containing AutoFDO profile information. If omitted, it defaults to fbdata.afdo in the current directory.

Producing an AutoFDO profile data file requires running your program with the perf utility on a supported GNU/Linux target system. For more information, see https://perf.wiki.kernel.org/. 🐧

E.g.

perf record -e br_inst_retired:near_taken -b -o perf.data \
    -- your_program

Then use the create_gcov tool to convert the raw profile data to a format that can be used by GCC. You must also supply the unstripped binary for your program to this tool. See https://github.com/google/autofdo.

E.g.

create_gcov --binary=your_program.unstripped --profile=perf.data \
    --gcov=profile.afdo

---

The following options control compiler behavior regarding floating-point arithmetic. These options trade off between speed and correctness. And listen — all of them must be specifically enabled. None of them just show up on their own, okay? 💅

• -fexcess-precision=style — this option allows control over excess precision on machines where floating-point operations occur in a format with more precision or range than the IEEE standard and interchange floating-point types. An example of such a target is x87 floating point on x86 processors, which uses an 80-bit representation internally instead of the 64-bit IEEE format. For most programs, the excess precision is harmless, but some programs may rely on the requirements of the C or C++ language standards for handling IEEE values. So like, it depends on your vibe. 🦋

By default, -fexcess-precision=fast is in effect, bestie. What that means: operations can be carried out in a wider precision than the types specified in the source if that would result in faster code, and it is unpredictable when rounding to the types specified in the source code takes place. When you're compiling C or C++, if -fexcess-precision=standard is specified then excess precision follows the rules specified in ISO C99 or C++, and in particular, both casts and assignments cause values to be rounded to their semantic types — whereas -ffloat-store only affects assignments, not casts, no cap. This option is enabled by default for C or C++ if a strict conformance option such as -std=c99 or -std=c++17 is used. And -ffast-math enables -fexcess-precision=fast by default regardless of whether a strict conformance option is used. If -fexcess-precision=16 is specified, constants and the results of expressions with types _Float16 and __bf16 are computed without excess precision. ✨

-fexcess-precision=standard is, unfortunately, not implemented for languages other than C or C++. And on the x86, it has no effect if -mfpmath=sse or -mfpmath=sse+387 is specified — in the former case, IEEE semantics apply without excess precision, and in the latter, rounding is unpredictable. So like. choose wisely 💅

• -ffloat-store — Do not store floating-point variables in registers, and inhibit other options that might change whether a floating-point value is taken from a register or memory. This option has generally been subsumed by -fexcess-precision=standard, which is more general. If you do use -ffloat-store, you may need to modify your program to explicitly store intermediate computations in temporary variables, since -ffloat-store handles rounding to IEEE format only on assignments and not casts, the way -fexcess-precision=standard does. 🫠
• -ffast-math — Sets the options -fno-math-errno, -funsafe-math-optimizations, -ffinite-math-only, -fno-rounding-math, -fno-signaling-nans, -fcx-limited-range and -fexcess-precision=fast.

This option causes the preprocessor macro __FAST_MATH__ to be defined.

This option is not turned on by any -O option besides -Ofast, since it can result in incorrect output for programs that depend on an exact implementation of IEEE or ISO rules/specifications for math functions. It may, however, yield faster code for programs that do not require the guarantees of these specifications. Speed vs. correctness, the eternal love story 🔥

• -fno-math-errno — Do not set errno after calling math functions that are executed with a single instruction, e.g., sqrt. A program that relies on IEEE exceptions for math error handling may want to use this flag for speed while maintaining IEEE arithmetic compatibility.

This option is not turned on by any -O option besides -Ofast, since it can result in incorrect output for programs that depend on an exact implementation of IEEE or ISO rules/specifications for math functions. It may, however, yield faster code for programs that do not require the guarantees of these specifications.

The default is -fmath-errno.

On Darwin systems, the math library never sets errno. There is therefore no reason for the compiler to consider the possibility that it might, and -fno-math-errno is the default. 🍎

• -funsafe-math-optimizations — Allow optimizations for floating-point arithmetic that (a) assume that arguments and results are valid and (b) may violate IEEE or ANSI standards. When used at link time, it may include libraries or startup files that change the default FPU control word or other similar optimizations.

babes the name literally has "unsafe" in it, i'm just saying 🙈

This option is not turned on by any -O option besides -Ofast, since it can result in incorrect output for programs that depend on an exact implementation of IEEE or ISO rules/specifications for math functions. It may, however, yield faster code for programs that do not require the guarantees of these specifications. Enables -fno-signed-zeros, -fno-trapping-math, -fassociative-math and -freciprocal-math.

The default is -fno-unsafe-math-optimizations.

• -fassociative-math — Allow re-association of operands in series of floating-point operations. This violates the ISO C and C++ language standard by possibly changing computation result. NOTE: re-ordering may change the sign of zero as well as ignore NaNs and inhibit or create underflow or overflow (and thus cannot be used on code that relies on rounding behavior like (x + 2**52) - 2**52). May also reorder floating-point comparisons and thus may not be used when ordered comparisons are required. This option requires that both -fno-signed-zeros and -fno-trapping-math be in effect. Moreover, it doesn't make much sense with -frounding-math. For Fortran the option is automatically enabled when both -fno-signed-zeros and -fno-trapping-math are in effect.

The default is -fno-associative-math.

• -freciprocal-math — Allow the reciprocal of a value to be used instead of dividing by the value if this enables optimizations. For example x / y can be replaced with x * (1/y), which is useful if (1/y) is subject to common subexpression elimination. Note that this loses precision and increases the number of flops operating on the value.

The default is -fno-reciprocal-math.

• -ffinite-math-only — Allow optimizations for floating-point arithmetic that assume that arguments and results are not NaNs or +-Infs. like. what if they ARE tho 😭

This option is not turned on by any -O option besides -Ofast, since it can result in incorrect output for programs that depend on an exact implementation of IEEE or ISO rules/specifications for math functions. It may, however, yield faster code for programs that do not require the guarantees of these specifications.

The default is -fno-finite-math-only.

• -fno-signed-zeros — Allow optimizations for floating-point arithmetic that ignore the signedness of zero. IEEE arithmetic specifies the behavior of distinct +0.0 and -0.0 values, which then prohibits simplification of expressions such as x+0.0 or 0.0*x (even with -ffinite-math-only). This option implies that the sign of a zero result isn't significant.

The default is -fsigned-zeros.

• -fno-trapping-math — Compile code assuming that floating-point operations cannot generate user-visible traps. These traps include division by zero, overflow, underflow, inexact result and invalid operation. This option requires that -fno-signaling-nans be in effect. Setting this option may allow faster code if one relies on "non-stop" IEEE arithmetic, for example.

This option is not turned on by any -O option besides -Ofast, since it can result in incorrect output for programs that depend on an exact implementation of IEEE or ISO rules/specifications for math functions.

The default is -ftrapping-math.

Future versions of GCC may provide finer control of this setting using C99's FENV_ACCESS pragma. This command-line option will be used along with -frounding-math to specify the default state for FENV_ACCESS.

• -frounding-math — Disable transformations and optimizations that assume default floating-point rounding behavior (round-to-nearest). This option should be specified for programs that change the FP rounding mode dynamically, or that may be executed with a non-default rounding mode. This option disables constant folding of floating-point expressions at compile time (which may be affected by rounding mode) and arithmetic transformations that are unsafe in the presence of sign-dependent rounding modes.

The default is -fno-rounding-math.

This option is experimental and does not currently guarantee to disable all GCC optimizations that are affected by rounding mode. 🙈

Future versions of GCC may provide finer control of this setting using C99's FENV_ACCESS pragma. This command-line option will be used along with -ftrapping-math to specify the default state for FENV_ACCESS.

• -fsignaling-nans — Compile code assuming that IEEE signaling NaNs may generate user-visible traps during floating-point operations. Setting this option disables optimizations that may change the number of exceptions visible with signaling NaNs. This option implies -ftrapping-math.

This option causes the preprocessor macro __SUPPORT_SNAN__ to be defined. ✨

The default is -fno-signaling-nans.

Also experimental, bestie — it does not currently guarantee to disable all GCC optimizations that affect signaling NaN behavior.

• -fno-fp-int-builtin-inexact — Do not allow the built-in functions ceil, floor, round and trunc, and their float and long double variants, to generate code that raises the inexact floating-point exception for noninteger arguments. ISO C99 and C11 allow these functions to raise the inexact exception, but ISO/IEC TS 18661-1:2014, the C bindings to IEEE 754-2008, as integrated into ISO C23, does not allow these functions to do so.

The default is -ffp-int-builtin-inexact, allowing the exception to be raised, unless C23 or a later C standard is selected. This option does nothing unless -ftrapping-math is in effect.

Even if -fno-fp-int-builtin-inexact is used, if the functions generate a call to a library function then the inexact exception may be raised if the library implementation does not follow TS 18661. iykyk 🫶

• -fsingle-precision-constant — Treat floating-point constants as single precision instead of implicitly converting them to double-precision constants.
• -fcx-limited-range — When enabled, this option states that a range reduction step is not needed when performing complex division. Also, there is no checking whether the result of a complex multiplication or division is NaN + I*NaN, with an attempt to rescue the situation in that case. The default is -fno-cx-limited-range, but is enabled by -ffast-math.

This option controls the default setting of the ISO C99 CX_LIMITED_RANGE pragma. Nevertheless, the option applies to all languages.

• -fcx-fortran-rules — Complex multiplication and division follow Fortran rules. Range reduction is done as part of complex division, but there is no checking whether the result of a complex multiplication or division is NaN + I*NaN, with an attempt to rescue the situation in that case.

The default is -fno-cx-fortran-rules.

So these are the ✨extra credit✨ optimization options — the ones that can make your code faster, but that no -O level turns on for you. And heads up bestie: some of these are experimental, meaning they can literally hand you broken code. Not a vibe. Read the fine print, okay?

---

• -fbranch-probabilities — okay so picture this: you compile your program with -fprofile-arcs, run it, and then you compile it a second time using -fbranch-probabilities so GCC can optimize based on how many times each branch actually got taken. When a program compiled with -fprofile-arcs exits, it saves arc execution counts to a file called sourcename.gcda for each source file. The tea ☕ here is that this data file is super dependent on the structure of the generated code, so you must use the same source code and the same optimization options for both compilations. (Details on the file naming live over in -fprofile-arcs.)

With -fbranch-probabilities, GCC puts a REG_BR_PROB note on each JUMP_INSN and CALL_INSN. These can be used to improve optimization. Currently they're only used in one place: in reorg.cc, instead of guessing which path a branch is most likely to take, the REG_BR_PROB values are used to exactly determine which path is taken more often. 💅

Enabled by -fprofile-use and -fauto-profile.

• -fprofile-values — if you combine it with -fprofile-arcs, it adds code so that some data about values of expressions in the program gets gathered.

With -fbranch-probabilities, it reads back the data gathered from profiling values of expressions for usage in optimizations.

Enabled by -fprofile-generate, -fprofile-use, and -fauto-profile.

• -fprofile-reorder-functions — function reordering based on profile instrumentation collects the first time of execution of a function and orders these functions in ascending order, aiming to optimize program startup through more efficient loading of text segments. 🚀

Enabled with -fprofile-use.

• -fvpt — if you combine it with -fprofile-arcs, this option tells the compiler to add code to gather information about values of expressions.

With -fbranch-probabilities, it reads back the data gathered and actually performs the optimizations based on them. Currently the optimizations include specialization of division operations using the knowledge about the value of the denominator. (dividing by a number it already knows about? slay.)

Enabled with -fprofile-use and -fauto-profile.

• -frename-registers — attempts to avoid false dependencies in scheduled code by making use of registers left over after register allocation. This optimization most benefits processors with lots of registers. BUT — depending on the debug information format adopted by the target, it can make debugging impossible, since variables no longer stay in a "home register". yeah. impossible. no cap. 😭

Enabled by default with -funroll-loops.

• -fschedule-fusion — performs a target dependent pass over the instruction stream to schedule instructions of the same type together, because the target machine can execute them more efficiently if they're adjacent to each other in the instruction flow. 🫶

Enabled at levels -O2, -O3, -Os.

• -ftracer — performs tail duplication to enlarge superblock size. This transformation simplifies the control flow of the function, allowing other optimizations to do a better job.

Enabled by -fprofile-use and -fauto-profile.

• -funroll-loops — unrolls loops whose number of iterations can be determined at compile time or upon entry to the loop. -funroll-loops implies -frerun-cse-after-loop, -fweb and -frename-registers. It also turns on complete loop peeling (i.e. complete removal of loops with a small constant number of iterations). This option makes code larger, and may or may not make it run faster. (gambling, basically 🎲)

Enabled by -fprofile-use and -fauto-profile.

• -funroll-all-loops — unrolls all the loops, even if their number of iterations is uncertain when the loop is entered. This usually makes programs run more slowly. ma'am. -funroll-all-loops implies the same options as -funroll-loops.
• -fpeel-loops — peels loops for which there's enough information that they don't roll much (from profile feedback or static analysis). It also turns on complete loop peeling (i.e. complete removal of loops with a small constant number of iterations).

Enabled by -O3, -fprofile-use, and -fauto-profile.

• -fmalloc-dce — controls whether malloc (and its variants such as calloc or strdup) can be optimized away, provided its return value is only used as a parameter of a free call or compared with NULL. If -fmalloc-dce=1 is used, only calls to free are allowed, while with -fmalloc-dce=2 comparisons with NULL pointer are also considered safe to remove.

The default is -fmalloc-dce=2. See also -fallocation-dce.

• -fmove-loop-invariants — enables the loop invariant motion pass in the RTL loop optimizer. Enabled at level -O1 and higher, except for -Og.
• -fmove-loop-stores — enables the loop store motion pass in the GIMPLE loop optimizer. This moves invariant stores to after the end of the loop, in exchange for carrying the stored value in a register across the iteration. Note: for this option to actually do anything, -ftree-loop-im has to be enabled as well. Enabled at level -O1 and higher, except for -Og.
• -fsplit-loops — splits a loop into two if it contains a condition that's always true for one side of the iteration space and false for the other. clean split. iconic.

Enabled by -fprofile-use and -fauto-profile.

• -funswitch-loops — moves branches with loop invariant conditions out of the loop, with duplicates of the loop on both branches (modified according to the result of the condition).

Enabled by -fprofile-use and -fauto-profile.

• -fversion-loops-for-strides — if a loop iterates over an array with a variable stride, it creates another version of the loop that assumes the stride is always one. For example:

for (int i = 0; i < n; ++i)
  x[i * stride] = ...;

becomes:

if (stride == 1)
  for (int i = 0; i < n; ++i)
    x[i] = ...;
else
  for (int i = 0; i < n; ++i)
    x[i * stride] = ...;

This is particularly useful for assumed-shape arrays in Fortran, where (for example) it allows better vectorization assuming contiguous accesses. This flag is enabled by default at -O3. It is also enabled by -fprofile-use and -fauto-profile. 🌈

• -ffunction-sections
• -fdata-sections — these two place each function or data item into its own section in the output file, if the target supports arbitrary sections. The name of the function or the name of the data item determines the section's name in the output file.

Use these options on systems where the linker can perform optimizations to improve locality of reference in the instruction space. Most systems using the ELF object format have linkers with such optimizations. On AIX, the linker rearranges sections (CSECTs) based on the call graph. The performance impact varies.

Together with a linker garbage collection (linker --gc-sections option) these options may lead to smaller statically-linked executables (after stripping). smol. cute. efficient.

On ELF/DWARF systems these options do not degenerate the quality of the debug information. There could be issues with other object files/debug info formats. 🙈

Only use these options when there are significant benefits from doing so. When you specify these options, the assembler and linker create larger object and executable files and are also slower. These options affect code generation. They prevent optimizations by the compiler and assembler using relative locations inside a translation unit, since the locations are unknown until link time. An example of such an optimization is relaxing calls to short call instructions.

• -fstdarg-opt — optimizes the prologue of variadic argument functions with respect to usage of those arguments.
• -fsection-anchors — tries to reduce the number of symbolic address calculations by using shared "anchor" symbols to address nearby objects. This transformation can help to reduce the number of GOT entries and GOT accesses on some targets.

For example, the implementation of the following function foo:

static int a, b, c;
int foo (void) { return a + b + c; }

usually calculates the addresses of all three variables, but if you compile it with -fsection-anchors, it accesses the variables from a common anchor point instead. The effect is similar to the following pseudocode (which isn't valid C):

int foo (void)
{
  register int *xr = &x;
  return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
}

Not all targets support this option. 💀

• -fzero-call-used-regs=*choice* — zeroes call-used registers at function return to increase program security, either by mitigating Return-Oriented Programming (ROP) attacks or by preventing information leakage through registers. security queen behavior.

The possible values of choice are the same as for the zero_call_used_regs attribute. The default is skip.

You can control this behavior for a specific function by using the function attribute zero_call_used_regs.

• --param *name*=*value* — okay so in some places, GCC uses various constants to control the amount of optimization that gets done. Like, for example, GCC does not inline functions that contain more than a certain number of instructions. You can control some of these constants on the command line using the --param option.

The names of specific parameters, and the meaning of the values, are tied to the internals of the compiler, and are subject to change without notice in future releases. ⚠️ (so like, don't build your whole personality around them)

In order to get the minimal, maximal and default values of a parameter, use the --help=param -Q options.

In each case, the value is an integer. The following choices of name are recognized for all targets:

• phiopt-factor-max-stmts-live — when factoring statements out of if/then/else, this is the max # of statements after the defining statement to be allowed to extend the lifetime of a name.
• predictable-branch-outcome — when a branch is predicted to be taken with probability lower than this threshold (in percent), then it's considered well predictable. 🔮

continuing the --param saga, bestie 💅 these are the tuning knobs that make GCC go brrrr or absolutely not, depending on your vibe:

• max-rtl-if-conversion-insns — okay so RTL if-conversion is out here trying to ghost conditional branches around a block and swap them for conditionally executed instructions. This parameter sets the max number of instructions in a block that should even be considered for if-conversion. The compiler also uses other heuristics to decide whether if-conversion is likely to be profitable.
• file-cache-files — Max number of files in the file cache. The file cache is used to print source lines in diagnostics and do some source checks like -Wmisleading-indentation.
• file-cache-lines — Max number of lines to index into file cache. When 0, this is automatically sized. Same file cache agenda: print source lines in diagnostics and do source checks like -Wmisleading-indentation.
• max-rtl-if-conversion-predictable-cost — RTL if-conversion will try to remove conditional branches around a block and replace them with conditionally executed instructions. These parameters give the maximum permissible cost for the sequence that would be generated by if-conversion, depending on whether the branch is statically determined to be predictable or not. The units for this parameter are the same as those for the GCC internal seq_cost metric. The compiler will try to provide a reasonable default for this parameter using the BRANCH_COST target macro.
• max-crossjump-edges — Maximum number of incoming edges to consider for cross-jumping. The algorithm used by -fcrossjumping is O(N^2) in the number of edges incoming to each block. Increasing values mean more aggressive optimization, making the compilation time increase with probably small improvement in executable size.
• min-crossjump-insns — The minimum number of instructions that must be matched at the end of two blocks before cross-jumping is performed on them. This value is ignored in the case where all instructions in the block being cross-jumped from are matched.
• max-grow-copy-bb-insns — The maximum code size expansion factor when copying basic blocks instead of jumping. The expansion is relative to a jump instruction.
• max-goto-duplication-insns — The maximum number of instructions to duplicate to a block that jumps to a computed goto. To avoid O(N^2) behavior in a number of passes, GCC factors computed gotos early in the compilation process, and unfactors them as late as possible. Only computed jumps at the end of a basic block with no more than max-goto-duplication-insns are unfactored.
• max-delay-slot-insn-search — The maximum number of instructions to consider when looking for an instruction to fill a delay slot. If more than this arbitrary number of instructions are searched, the time savings from filling the delay slot are minimal, so stop searching. Increasing values mean more aggressive optimization, making the compilation time increase with probably small improvement in execution time.
• max-delay-slot-live-search — When trying to fill delay slots, the maximum number of instructions to consider when searching for a block with valid live register information. Increasing this arbitrarily chosen value means more aggressive optimization, increasing the compilation time. This parameter should be removed when the delay slot code is rewritten to maintain the control-flow graph.
• max-gcse-memory — The approximate maximum amount of memory in "kB" that can be allocated in order to perform the global common subexpression elimination optimization. If more memory than specified is required, the optimization is not done.
• max-gcse-insertion-ratio — If the ratio of expression insertions to deletions is larger than this value for any expression, then RTL PRE inserts or removes the expression and thus leaves partially redundant computations in the instruction stream.
• max-pending-list-length — The maximum number of pending dependencies scheduling allows before flushing the current state and starting over. Large functions with few branches or calls can create excessively large lists which needlessly consume memory and resources.
• max-modulo-backtrack-attempts — The maximum number of backtrack attempts the scheduler should make when modulo scheduling a loop. Larger values can exponentially increase compilation time.
• max-inline-functions-called-once-loop-depth — Maximal loop depth of a call considered by inline heuristics that tries to inline all functions called once.
• max-inline-functions-called-once-insns — Maximal estimated size of functions produced while inlining functions called once.
• max-inline-insns-single — Several parameters control the tree inliner used in GCC. This number sets the maximum number of instructions (counted in GCC's internal representation) in a single function that the tree inliner considers for inlining. This only affects functions declared inline and methods implemented in a class declaration (C++).
• max-inline-insns-auto — When you use -finline-functions (included in -O3), a lot of functions that would otherwise not be considered for inlining by the compiler are investigated. To those functions, a different (more restrictive) limit compared to functions declared inline can be applied (--param max-inline-insns-auto).
• max-inline-insns-small — This is the bound applied to calls that are considered relevant with -finline-small-functions.
• max-inline-insns-size — This is the bound applied to calls that are optimized for size. Small growth may be desirable to anticipate optimization opportunities exposed by inlining.
• uninlined-function-insns — Number of instructions accounted by inliner for function overhead such as function prologue and epilogue.
• uninlined-function-time — Extra time accounted by inliner for function overhead such as time needed to execute function prologue and epilogue.
• inline-heuristics-hint-percent — The scale (in percents) applied to inline-insns-single, inline-insns-single-O2, inline-insns-auto when inline heuristics hints that inlining is very profitable (will enable later optimizations).
• uninlined-thunk-insns, uninlined-thunk-time — Same as --param uninlined-function-insns and --param uninlined-function-time but applied to function thunks.
• inline-min-speedup — When estimated performance improvement of caller + callee runtime exceeds this threshold (in percent), the function can be inlined regardless of the limit on --param max-inline-insns-single and --param max-inline-insns-auto.
• large-function-insns — The limit specifying really large functions. For functions larger than this limit after inlining, inlining is constrained by --param large-function-growth. This parameter is useful primarily to avoid extreme compilation time caused by non-linear algorithms used by the back end.
• large-function-growth — Specifies maximal growth of large functions caused by inlining in percents. For example, parameter value 100 limits large function growth to 2.0 times the original size.
• large-unit-insns — The limit specifying large translation unit. Growth caused by inlining of units larger than this limit is limited by --param inline-unit-growth. For small units this might be too tight. For example, consider a unit consisting of function A that is inline and B that just calls A three times. If B is small relative to A, the growth of unit is 300\% and yet such inlining is very sane. For very large units consisting of small inlineable functions, however, the overall unit growth limit is needed to avoid exponential explosion of code size. Thus for smaller units, the size is increased to --param large-unit-insns before applying --param inline-unit-growth.
• lazy-modules — Maximum number of concurrently open C++ module files when lazy loading.
• inline-unit-growth — Specifies maximal overall growth of the compilation unit caused by inlining. For example, parameter value 20 limits unit growth to 1.2 times the original size. Cold functions (either marked cold via an attribute or by profile feedback) are not accounted into the unit size.
• ipa-cp-unit-growth — Specifies maximal overall growth of the compilation unit caused by interprocedural constant propagation. For example, parameter value 10 limits unit growth to 1.1 times the original size.
• ipa-cp-large-unit-insns — The size of translation unit that IPA-CP pass considers large.
• large-stack-frame — The limit specifying large stack frames. While inlining the algorithm is trying to not grow past this limit too much.
• large-stack-frame-growth — Specifies maximal growth of large stack frames caused by inlining in percents. For example, parameter value 1000 limits large stack frame growth to 11 times the original size.
• max-inline-insns-recursive, max-inline-insns-recursive-auto — Specifies the maximum number of instructions an out-of-line copy of a self-recursive inline function can grow into by performing recursive inlining.

--param max-inline-insns-recursive applies to functions declared inline. For functions not declared inline, recursive inlining happens only when -finline-functions (included in -O3) is enabled; --param max-inline-insns-recursive-auto applies instead.

• max-inline-recursive-depth, max-inline-recursive-depth-auto — Specifies the maximum recursion depth used for recursive inlining.
• --param max-inline-recursive-depth — this one's for functions that are actually declared inline, okay? For functions that did not get declared inline, the recursive inlining tea only gets spilled when -finline-functions is enabled (and yes bestie, that's baked into -O3); and in that case it's --param max-inline-recursive-depth-auto that runs the show instead. 💅
• min-inline-recursive-probability — so recursive inlining is only worth it for functions that have deep recursion on average, and it can literally hurt functions with shallow recursion depth by bloating the prologue size or making the function body more complex for the other optimizers to deal with. 😭 When profile feedback is available (peep -fprofile-generate), you can lowkey guess the actual recursion depth from the probability that a function recurses via a given call expression. This parameter keeps inlining to only the call expressions whose probability exceeds the threshold you give it (in percents, bestie). ✨
• early-inlining-insns — sets the growth that the early inliner is allowed to make. The way it works, it just increases how much inlining happens for code with a big abstraction penalty. 🫶
• max-early-inliner-iterations — caps how many iterations the early inliner does. Basically it bounds how many nested indirect calls the early inliner can resolve. Deeper chains? Not abandoned, they get handled by late inlining. 💜
• comdat-sharing-probability — the probability (in percent) that a C++ inline function with comdat visibility gets shared across multiple compilation units.
• modref-max-bases, modref-max-refs, modref-max-accesses — these three set the maximal number of base pointers, references and accesses that the mod/ref analysis stores for a single function. 📦
• modref-max-tests — the maxmal number of tests the alias oracle can run to disambiguate memory locations using the mod/ref info. Heads up: this one ought to be bigger than --param modref-max-bases and --param modref-max-refs. 👑
• modref-max-depth — the maximum depth of the DFS walk that modref escape analysis uses. Set it to 0 and the analysis is completely disabled, no cap. 💀
• modref-max-escape-points — the maximum number of escape points modref tracks per SSA-name.
• modref-max-adjustments — the maximum number of times the access range gets enlarged during modref dataflow analysis.
• profile-func-internal-id — controls whether the function internal id is used in the profile database lookup. If the value is 0, the compiler uses an id based on the function assembler name and filename, which makes old profile data more tolerant to source changes like function reordering etc. 🌈
• min-vect-loop-bound — the minimum number of iterations under which loops do not get vectorized when -ftree-vectorize is in play. For vectorization to be allowed, the number of iterations after vectorization has to be greater than whatever value you set here.
• gcse-cost-distance-ratio — the scaling factor used when calculating the maximum distance an expression can be moved by GCSE optimizations. Right now this is only supported in the code hoisting pass. The bigger the ratio, the more aggressive code hoisting gets with simple expressions — meaning the ones that cost less than gcse-unrestricted-cost. Set it to 0 and hoisting of simple expressions is off. ✋
• gcse-unrestricted-cost — the cost, roughly measured as the cost of a single typical machine instruction, at which GCSE optimizations stop constraining how far an expression can travel. Also currently only supported in the code hoisting pass. The lesser the cost, the more aggressive code hoisting is. Set it to 0 and every expression gets to travel unrestricted distances, periodt. 🦋
• max-hoist-depth — how deep the search goes in the dominator tree for expressions to hoist. It's there to avoid quadratic behavior in the hoisting algorithm. A value of 0 doesn't limit the search at all, but ngl it may slow down compilation of huge functions. 😬
• max-tail-merge-comparisons — the maximum amount of similar bbs to compare a bb with. Used to dodge quadratic behavior in tree tail merging.
• max-tail-merge-iterations — the maximum amount of iterations of the pass over the function. This one keeps compilation time under control in tree tail merging. ⏳
• store-merging-allow-unaligned — lets the store merging pass introduce unaligned stores, but only if it's legal to do so. 👀
• max-stores-to-merge — the maximum number of stores to attempt to merge into wider stores in the store merging pass.
• max-store-chains-to-track — the maximum number of store chains to track at the same time while trying to merge them into wider stores in the store merging pass.
• max-stores-to-track — the maximum number of stores to track at the same time in the attemt to to merge them into wider stores in the store merging pass. (yes, that's exactly how it's spelled, we don't fix it, we just vibe with it 💀)
• max-unrolled-insns — the maximum number of instructions a loop may have to be unrolled. And if a loop does get unrolled, this parameter also decides how many times the loop code gets unrolled. 🔁
• max-average-unrolled-insns — the maximum number of instructions, biased by the probabilities of their execution, that a loop may have to be unrolled. If a loop is unrolled, this parameter also determines how many times the loop code is unrolled.
• max-unroll-times — the maximum number of unrollings of a single loop.
• max-peeled-insns — the maximum number of instructions that a loop may have to be peeled. If a loop is peeled, this parameter also determines how many times the loop code is peeled. 🍌
• max-peel-times — the maximum number of peelings of a single loop.
• max-peel-branches — the maximum number of branches on the hot path through the peeled sequence.
• max-completely-peeled-insns — the maximum number of insns of a completely peeled loop.
• max-completely-peel-times — the maximum number of iterations of a loop for it to be suitable for complete peeling.
• max-completely-peel-loop-nest-depth — the maximum depth of a loop nest that's suitable for complete peeling.
• max-unswitch-insns — the maximum number of insns of an unswitched loop.
• max-unswitch-depth — the maximum depth of a loop nest to be unswitched.
• lim-expensive — the minimum cost of an expensive expression in the loop invariant motion.
• min-loop-cond-split-prob — when FDO profile information is available, min-loop-cond-split-prob sets the minimum threshold for the probability of a semi-invariant condition statement to trigger a loop split.
• iv-consider-all-candidates-bound — the bound on the number of candidates for induction variables, below which all candidates get considered for each use in induction variable optimizations. If there are more candidates than this, only the most relevant ones get considered, so we avoid quadratic time complexity. 🧠
• iv-max-considered-uses — the induction variable optimizations give up on loops that contain more induction variable uses than this.
• iv-always-prune-cand-set-bound — if the number of candidates in the set is smaller than this value, always try to remove unnecessary ivs from the set when adding a new one. ✂️
• avg-loop-niter — average number of iterations of a loop.
• dse-max-object-size — the maximum size (in bytes) of objects tracked bytewise by dead store elimination. Heads up: larger values may result in larger compilation times. 🐌
• dse-max-alias-queries-per-store — the maximum number of queries into the alias oracle per store. Larger values mean larger compilation times, and may also mean more removed dead stores. Trade-offs, bestie.
• scev-max-expr-size — the bound on the size of expressions used in the scalar evolutions analyzer. Large expressions slow the analyzer down. 😮‍💨
• scev-max-expr-complexity — the bound on the complexity of the expressions in the scalar evolutions analyzer. Complex expressions also slow the analyzer down.
• max-tree-if-conversion-phi-args — the maximum number of arguments in a PHI supported by TREE if conversion, unless the loop is marked with the simd pragma. 🎀
• vect-max-layout-candidates — the maximum number of possible vector layouts (like permutations) to consider when optimizing to-be-vectorized code.
• vect-max-version-for-alignment-checks — the maximum number of run-time checks that can be performed when doing loop versioning for alignment in the vectorizer.
• vect-max-version-for-alias-checks — the maximum number of run-time checks that can be performed when doing loop versioning for alias in the vectorizer.
• vect-max-peeling-for-alignment — the maximum number of loop peels to enhance access alignment for the vectorizer. Value -1 means no limit, go wild. 🔥
• max-iterations-to-track — the maximum number of iterations of a loop that the brute-force algorithm for analysis of the number of iterations of the loop tries to evaluate.
• hot-bb-count-fraction — the denominator n of fraction 1/n of the maximal execution count of a basic block in the entire program that a basic block needs to at least have in order to be considered hot. The default is 10000, which means a basic block is considered hot if its execution count is greater than 1/10000 of the maximal execution count. 0 means it is never considered hot. Used in non-LTO mode. 🔥
• hot-bb-count-ws-permille — the number of most executed permilles, ranging from 0 to 1000, of the profiled execution of the entire program that the execution count of a basic block must be part of in order to be considered hot. The default is 990, which means a basic block is considered hot if its execution count contributes to the upper 990 permilles, or 99.0%, of the profiled execution of the entire program. 0 means it is never considered hot. Used in LTO mode. 👑

hot-bb-frequency-fraction 🔥

Okay so imagine the denominator n of the fraction 1/n of how often a function's entry block runs. A basic block in that function has to hit at least that much to be considered hot 💅.

Default is 1000, so a basic block counts as hot if it runs more than 1/1000 of the entry block's frequency. And 0? Yeah, that means it is never considered hot. Ever. Cold queen behavior 🧊.

unlikely-bb-count-fraction 🙈

The denominator n of the fraction 1/n of the total profiled runs of the whole program — if a basic block's execution count is below that, it gets flagged as unlikely executed.

Default is 20, so a basic block is unlikely executed if it runs in fewer than 1/20, aka 5%, of the program's runs. Set it to 0 and it is always considered unlikely executed. no cap 😭.

max-predicted-iterations 🔁

The max number of loop iterations GCC predicts statically. Super handy when a function has one loop with a known bound and another with an unknown bound — the known one gets predicted correctly while the unknown one averages out to roughly 10. Which means the unbounded loop ends up looking artificially cold next to its sibling 🙃.

builtin-expect-probability 🎲

Controls the probability that the expression has the specified value. This parameter takes a percentage (i.e. 0 ... 100) as input.

builtin-string-cmp-inline-length 🧵

The max length of a constant string for a builtin string cmp call that's eligible for inlining.

align-threshold ✨

Pick the fraction of the maximal execution frequency of a basic block in a function to align that basic block. It's giving alignment energy 👑

align-loop-iterations 🌀

A loop that's expected to iterate at least the selected number of iterations gets aligned.

tracer-dynamic-coverage / tracer-dynamic-coverage-feedback 🕵️

This value limits superblock formation once a given percentage of executed instructions is covered. Basically it puts a cap on unnecessary code size expansion, bestie ✋.

tracer-dynamic-coverage-feedback only kicks in when profile feedback is available. And real profiles (the opposite of statically estimated ones) are way less balanced, which lets the threshold go to a larger value. iykyk 📊.

tracer-max-code-growth 📈

Stops tail duplication once code growth hits the given percentage. Honestly this is a pretty artificial limit — most of the duplicates get eliminated later during cross jumping — so you can set it to way higher values than the code growth you actually want.

tracer-min-branch-ratio 🥀

Stops reverse growth when the reverse probability of the best edge drops below this threshold (in percent).

tracer-min-branch-probability / tracer-min-branch-probability-feedback 🛑

Stops forward growth if the best edge has a probability lower than this threshold.

Just like with tracer-dynamic-coverage, you get two parameters: tracer-min-branch-probability-feedback is for compiling with profile feedback, and tracer-min-branch-probability is for compiling without it. The feedback value needs to be more conservative (higher) for the tracer to actually be effective, periodt 🌸.

stack-clash-protection-guard-size 🛡️

Specify the size of the operating system provided stack guard as 2 raised to num bytes. Higher values may cut down the number of explicit probes — but a value larger than the OS-provided guard will leave your code vulnerable to stack clash style attacks. Like, actually vulnerable 💀.

stack-clash-protection-probe-interval 📏

Stack clash protection means probing stack space as it gets allocated. This param controls the maximum distance between probes into the stack, as 2 raised to num bytes. Higher values may reduce the explicit probes, but again — bigger than the OS-provided guard and your code is vulnerable to stack clash style attacks 😬.

max-cse-path-length 🛤️

The max number of basic blocks on a path that CSE considers.

max-cse-insns 🧮

The max number of instructions CSE processes before flushing.

ggc-min-expand 🗑️

GCC uses a garbage collector to manage its own memory allocation, cute right? This parameter sets the minimum percentage by which the GC's heap is allowed to expand between collections. Tuning it may improve compilation speed — and it has no effect on code generation at all 💖.

The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when RAM >= 1GB. If getrlimit is available, "RAM" means the smallest of actual RAM and RLIMIT_DATA or RLIMIT_AS. If GCC can't calculate RAM on a particular platform, the lower bound of 30% is used. And if you set this parameter and ggc-min-heapsize to zero, a full collection happens at every opportunity — which is extremely slow, but can be useful for debugging 🐛.

ggc-min-heapsize 🧺

The minimum size of the GC's heap before it even bothers collecting garbage. The first collection happens after the heap expands by ggc-min-expand% beyond ggc-min-heapsize. Again, tuning this may improve compilation speed, and has no effect on code generation ✨.

The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that tries to make sure RLIMIT_DATA or RLIMIT_AS aren't exceeded — with a lower bound of 4096 (four megabytes) and an upper bound of 131072 (128 megabytes). If GCC can't figure out RAM on a particular platform, the lower bound gets used. Setting this parameter very large effectively disables garbage collection entirely. Setting this parameter and ggc-min-expand to zero causes a full collection at every opportunity.

max-reload-search-insns 🔄

The max number of instructions reload should look backward for an equivalent register. Increasing values mean more aggressive optimization — compilation time goes up, with probably slightly better performance. tradeoff era 😌.

max-cselib-memory-locations 🗂️

The max number of memory locations cselib should take into account. Same deal: higher values mean more aggressive optimization, longer compilation time, and probably slightly better performance.

max-sched-ready-insns 📋

The max number of instructions ready to be issued that the scheduler should consider at any given moment during the first scheduling pass. Increasing values mean more thorough searches, which makes compilation time increase with probably little benefit. 🙃

max-sched-region-blocks 🧱

The max number of blocks in a region to be considered for interblock scheduling.

max-pipeline-region-blocks 🚰

The max number of blocks in a region to be considered for pipelining in the selective scheduler.

max-sched-region-insns 🧩

The max number of insns in a region to be considered for interblock scheduling.

max-pipeline-region-insns 🎀

The max number of insns in a region to be considered for pipelining in the selective scheduler.

min-spec-prob 🎯

The minimum probability (in percents) of reaching a source block for interblock speculative scheduling.

max-sched-extend-regions-iters 🔂

The max number of iterations through CFG to extend regions. A value of 0 disables region extensions, fyi.

max-sched-insn-conflict-delay ⏳

The max conflict delay for an insn to be considered for speculative motion.

sched-spec-prob-cutoff ✂️

The minimal probability of speculation success (in percents), so that speculative insns actually get scheduled.

sched-state-edge-prob-cutoff 🌉

The minimum probability an edge has to have for the scheduler to save its state across it.

sched-mem-true-dep-cost 💾

Minimal distance (in CPU cycles) between a store and a load targeting the same memory locations.

selsched-max-lookahead 🔭

The max size of the lookahead window of selective scheduling. It's basically a depth of search for available instructions.

selsched-max-sched-times 🔁

The max number of times an instruction gets scheduled during selective scheduling. This is the limit on the number of iterations the instruction may be pipelined through.

selsched-insns-to-rename 🏷️

The max number of best instructions in the ready list that are considered for renaming in the selective scheduler.

sms-min-sc 🎡

The minimum value of stage count that swing modulo scheduler generates.

max-last-value-rtl 🗃️

The max size, measured as number of RTLs, that can be recorded in an expression in combiner for a pseudo register as the last known value of that register.

max-combine-insns 🧬

The max number of instructions the RTL combiner tries to combine.

max-combine-search-insns 🔍

The max number of instructions the RTL combiner searches through to find the next use of a given register definition. If it hits that limit without finding such a use, the combiner just stops trying to optimize the definition, shrug 💅.

Right now this limit only applies after certain successful combination attempts — but it could be extended to other cases in future. So like, watch this space 👀.

integer-share-limit 🔢

Small integer constants can share a data structure, which reduces the compiler's memory usage and bumps its speed. This sets the maximum value of a shared integer constant.

ssp-buffer-size 🛡️

The minimum size of buffers (i.e. arrays) that receive stack smashing protection when -fstack-protector is used.

Fun lil history moment: this default was "8" before Ubuntu 10.10, and currently it's "4", to increase the number of functions protected by the stack protector. glow-up 💖✨

---

• min-size-for-stack-sharing — the tiniest size a variable can be and still get invited to the stack slot sharing party when we're not optimizing. below that? it's not sharing, bestie.
• max-jump-thread-duplication-stmts — the cap on how many statements can live in a block that needs duplicating when we're threading jumps. block's too chunky? we're not duplicating it. 💅
• max-jump-thread-paths — the most paths we'll consider while out here hunting for jump threading opportunities. the vibe check: when we roll up to a block, we only even look at the incoming edges if the number of paths we've searched so far, multiplied by the number of incoming edges, doesn't blow past this max. math first, always. 🔥
• max-fields-for-field-sensitive — the cap on how many fields in a struct get treated in a field-sensitive manner during pointer analysis. huge struct? okay girl, we're not tracking every single field.
• prefetch-latency — our estimate of the average number of instructions that get executed before a prefetch finishes. the distance we prefetch ahead is proportional to this constant. heads up though: bumping this number up may also mean less streams being prefetched (peep simultaneous-prefetches).
• simultaneous-prefetches — the max number of prefetches that can be running at the exact same time. it's giving limited seating at the function. ✨
• l1-cache-line-size — the size of a cache line in the L1 data cache, in bytes. periodt.
• l1-cache-size — the size of your L1 data cache, in kilobytes.
• l2-cache-size — the size of your L2 data cache, also in kilobytes.
• prefetch-dynamic-strides — do we want the loop array prefetch pass issuing software prefetch hints for strides that are non-constant? sometimes that's a win, but ngl, a non-constant stride makes it genuinely hard to predict whether there's a clear benefit to sending those hints out.

Set it to 1 if the prefetch hints should be issued for non-constant strides. Set it to 0 if prefetch hints should only go out for strides that are known to be constant and below prefetch-minimum-stride.

• prefetch-minimum-stride — the minimum constant stride, in bytes, where we start using prefetch hints at all. stride under this threshold? no hints, bestie, we said what we said.

this setting is lowkey perfect for processors that already have hardware prefetchers, because in that case the hardware and the software prefetchers can end up beefing with each other. if the hardware prefetchers have a maximum stride they can handle, that value should go here — it makes the software prefetchers way more useful.

a value of -1 means we have no threshold at all, so prefetch hints can be issued for any constant stride.

just know this setting only does anything for strides that are known and constant. non-constant strides, this is not for you 💀

• destructive-interference-size, constructive-interference-size — the values for the C++17 variables std::hardware_destructive_interference_size and std::hardware_constructive_interference_size. okay so: the destructive interference size is the minimum recommended offset between two independent, concurrently-accessed objects. the constructive interference size is the maximum recommended size of contiguous memory accessed together. typically both of them are just the size of an L1 cache line for the target, in bytes. and for a generic target covering a whole range of L1 cache line sizes, usually the constructive interference size is the small end of that range and the destructive one is the large end. 🦋

the destructive interference size is meant for layout, which means it has ABI impact — real consequences. the default value is not expected to be stable, and on some targets it varies with -mtune, so using this variable somewhere ABI stability actually matters — like the public interface of a library — is strongly discouraged. if you're gonna do it anyway, users can stabilize the value using this option. not me warning you twice. 🙈

the constructive interference size is way less sensitive, since it's typically only used in a static_assert to make sure a type fits inside a cache line. cute, low-stakes, we love her.

see also -Winterference-size.

• loop-interchange-max-num-stmts — the max number of stmts in a loop that we're willing to interchange.
• loop-interchange-stride-ratio — the minimum ratio between the strides of two loops for an interchange to actually be profitable. no ratio, no swap. ✨
• min-insn-to-prefetch-ratio — the minimum ratio between the number of instructions and the number of prefetches needed to enable prefetching in a loop.
• prefetch-min-insn-to-mem-ratio — the minimum ratio between the number of instructions and the number of memory references needed to enable prefetching in a loop.
• use-canonical-types — should the compiler use the "canonical" type system? bestie, this should always be 1, which uses a more efficient internal mechanism for comparing types in C++ and Objective-C++. but if bugs in the canonical type system are causing compilation failures, set this value to 0 to disable canonical types. emergency lever! 🚨
• switch-conversion-max-branch-ratio — switch initialization conversion straight up refuses to create arrays that are bigger than switch-conversion-max-branch-ratio times the number of branches in the switch. she has boundaries. 💅
• switch-lower-slow-alg-max-cases — the maximum number of cases where the slow switch lowering algorithms get used.
• max-partial-antic-length — the maximum length of the partial antic set computed during the tree partial redundancy elimination optimization (-ftree-pre) when we're optimizing at -O3 and above. and here's why this matters: for some sorts of source code, the enhanced partial redundancy elimination optimization can basically run away and eat all of the memory on the host machine 😭. this parameter puts a limit on the length of the sets that get computed, which stops that runaway behavior from happening. setting it to 0? that allows an unlimited set length. so like... that's your choice to make, i guess. 🙈
• rpo-vn-max-loop-depth — the maximum loop depth that gets value-numbered optimistically. when the limit hits, the innermost rpo-vn-max-loop-depth loops and the outermost loop in the loop nest get value-numbered optimistically, and the remaining ones don't. the middle children are on their own fr fr.
• sccvn-max-alias-queries-per-access — the max number of alias-oracle queries we perform when we're out here looking for redundancies in loads and stores. hit this limit and the search gets aborted, and the load or store isn't considered redundant. also, the number of queries is algorithmically limited to the number of stores on all paths from the load to the function entry, so there's a natural ceiling anyway. 🌈
• ira-max-loops-num — IRA uses regional register allocation by default. if a function has more loops than the number you give this parameter, then only at most that given number of the most frequently-executed loops will form regions for regional register allocation. popular loops only, sorry to the rest.
• ira-max-conflict-table-size — even though IRA uses a sophisticated algorithm to compress the conflict table, for huge functions that table can still demand an excessive amount of memory. so: if the conflict table for a function could be more than the size in MB given by this parameter, the register allocator switches to a faster, simpler, and lower-quality algorithm that doesn't need to build a pseudo-register conflict table at all. speed over beauty. 💅
• ira-loop-reserved-regs — IRA can be used to evaluate more accurate register pressure in loops, which informs decisions about moving loop invariants (see -O3). this parameter says how many available registers get reserved for some other purposes. the default is literally the best value found from numerous experiments, so maybe don't touch it unless you know something. ✨
• ira-consider-dup-in-all-alts — makes IRA consider the matching constraint (that's the duplicated operand number) heavily in all the available alternatives for the preferred register class. if it's set to zero, IRA only respects the matching constraint when it's in the only available alternative with an appropriate register class. otherwise, IRA checks all available alternatives for the preferred register class, even if it already found some choice with an appropriate register class, and respects the qualified matching constraint it found. she's thorough like that. 👑
• ira-simple-lra-insn-threshold — approximate function insn number, in 1K units, that triggers simple local RA.
• lra-inheritance-ebb-probability-cutoff — LRA tries to reuse values that were reloaded into registers in subsequent insns. this optimization is called inheritance (iconic name, no notes). EBB is used as the region to do this optimization in. this parameter defines the minimal fall-through edge probability, in percentage, used to add a BB to the inheritance EBB in LRA. the default value was chosen from numerous runs of SPEC2000 on x86-64. science! 🧚
• loop-invariant-max-bbs-in-loop — loop invariant motion can get very expensive, both in compilation time and in the amount of compile-time memory it needs, once loops get very large. so loops with more basic blocks than this parameter simply won't have loop invariant motion optimization performed on them. big loop, no motion. 🍓
• loop-max-datarefs-for-datadeps — building data dependencies is expensive for very large loops, so this parameter limits the number of data references in loops that get considered for data dependence analysis. these large loops just aren't handled by the optimizations that use loop data dependencies.
• max-vartrack-size — sets a maximum number of hash table slots to use during variable tracking dataflow analysis of any function. if this limit is exceeded with variable tracking at assignments enabled, analysis for that function is retried without it, after removing all debug insns from the function. and if the limit is exceeded even without debug insns? then var tracking analysis is completely disabled for that function. setting the parameter to 0 makes it unlimited. 🎀
• max-vartrack-expr-depth — okay so this sets the max recursion levels when you're trying to map variable names or debug temporaries to value expressions. it's a whole trade-off: more complete debug info vs. compile time. if you set it too low, value expressions that are literally available and could be represented in the debug info just... won't be used 💀. set it higher and the compiler might find more complex debug expressions, but bestie, compile time and memory use are gonna grow. periodt.
• max-debug-marker-count — this sets a threshold on how many debug markers (like begin stmt markers) you can have to avoid a complexity explosion at inlining or expanding to RTL. if a function has more of those gimple stmts than the set limit, they get dropped from the inlined copy of a function, and from its RTL expansion. no cap.
• min-nondebug-insn-uid — use uids starting at this parameter for nondebug insns. the range below the parameter is reserved exclusively for debug insns created by -fvar-tracking-assignments, but debug insns may get (non-overlapping) uids above it if the reserved range is exhausted. it's giving overflow energy ✨.
• ipa-sra-deref-prob-threshold — IPA-SRA replaces a pointer known not to be NULL with one or more new parameters only when the probability (in percent, relative to function entry) of it being dereferenced is higher than this parameter. so if it's not likely to be dereferenced, no replacement for you.
• ipa-sra-ptr-growth-factor — IPA-SRA replaces a pointer to an aggregate with one or more new parameters only when their cumulative size is less than or equal to ipa-sra-ptr-growth-factor times the size of the original pointer parameter. math is mathing 🧮.
• ipa-sra-ptrwrap-growth-factor — additional maximum allowed growth of total size of new parameters that ipa-sra replaces a pointer to an aggregate with, if it points to a local variable that the caller only writes to and passes it as an argument to other functions. very specific, i know.
• ipa-sra-max-replacements — maximum pieces of an aggregate that IPA-SRA tracks. as a consequence, it's also the maximum number of replacements of a formal parameter. two for one 🫶.
• sra-max-scalarization-size-Ospeed, sra-max-scalarization-size-Osize — the two Scalar Reduction of Aggregates passes (SRA and IPA-SRA) aim to replace scalar parts of aggregates with uses of independent scalar variables. these parameters control the maximum size, in storage units, of aggregate which is considered for replacement when compiling for speed (sra-max-scalarization-size-Ospeed) or size (sra-max-scalarization-size-Osize) respectively. pick your fighter 💅.
• sra-max-propagations — the maximum number of artificial accesses that Scalar Replacement of Aggregates (SRA) will track, per one local variable, in order to facilitate copy propagation. it's giving stalker energy but for compiler optimizations.
• tm-max-aggregate-size — when making copies of thread-local variables in a transaction, this parameter specifies the size in bytes after which variables are saved with the logging functions as opposed to save/restore code sequence pairs. this option only applies when using -fgnu-tm. noted 📝.
• graphite-max-nb-scop-params — to avoid exponential effects in the Graphite loop transforms, the number of parameters in a Static Control Part (SCoP) is bounded. a value of zero can be used to lift the bound. a variable whose value is unknown at compilation time and defined outside a SCoP is a parameter of the SCoP. so now you know 💖.
• hardcfr-max-blocks — disable -fharden-control-flow-redundancy for functions with a larger number of blocks than the specified value. zero removes any limit. so if you want it to run on everything, set it to zero and let it cook 🍳.
• hardcfr-max-inline-blocks — force -fharden-control-flow-redundancy to use out-of-line checking for functions with a larger number of basic blocks than the specified value. so it stops inlining and goes out-of-line. slay 💅.
• loop-block-tile-size — loop blocking or strip mining transforms, enabled with -floop-block or -floop-strip-mine, strip mine each loop in the loop nest by a given number of iterations. the strip length can be changed using the loop-block-tile-size parameter. iconic 🌈.
• ipa-jump-function-lookups — specifies number of statements visited during jump function offset discovery. that's it. that's the tweet 🐦.
• ipa-cp-value-list-size — IPA-CP attempts to track all possible values and types passed to a function's parameter in order to propagate them and perform devirtualization. ipa-cp-value-list-size is the maximum number of values and types it stores per one formal parameter of a function. so it doesn't go overboard 🌊.
• ipa-cp-eval-threshold — IPA-CP calculates its own score of cloning profitability heuristics and performs those cloning opportunities with scores that exceed ipa-cp-eval-threshold. if you don't meet the threshold, you don't get cloned 🙅‍♀️.
• ipa-cp-max-recursive-depth — maximum depth of recursive cloning for self-recursive function. so it doesn't go infinite. because that would be a mess 😵‍💫.
• ipa-cp-min-recursive-probability — recursive cloning only when the probability of call being executed exceeds the parameter. so it's gotta be likely.
• ipa-cp-profile-count-base — when using -fprofile-use option, IPA-CP will consider the measured execution count of a call graph edge at this percentage position in their histogram as the basis for its heuristics calculation. so it's all about that histogram placement 📊.
• ipa-cp-recursive-freq-factor — the number of times interprocedural copy propagation expects recursive functions to call themselves. so it's like, how recursive do you think they'll be? 🤔
• ipa-cp-recursion-penalty — percentage penalty the recursive functions will receive when they are evaluated for cloning. so they get docked points for being recursive.
• ipa-cp-single-call-penalty — percentage penalty functions containing a single call to another function will receive when they are evaluated for cloning. so if you're just one call, you get penalized.
• ipa-max-agg-items — IPA-CP is also capable to propagate a number of scalar values passed in an aggregate. ipa-max-agg-items controls the maximum number of such values per one parameter. so it doesn't get too greedy 🤑.
• ipa-cp-loop-hint-bonus — when IPA-CP determines that a cloning candidate would make the number of iterations of a loop known, it adds a bonus of ipa-cp-loop-hint-bonus to the profitability score of the candidate. bonus points, bestie ✨.
• ipa-max-loop-predicates — the maximum number of different predicates IPA will use to describe when loops in a function have known properties. so it doesn't use all the predicates.
• ipa-max-aa-steps — during its analysis of function bodies, IPA-CP employs alias analysis in order to track values pointed to by function parameters. in order not spend too much time analyzing huge functions, it gives up and consider all memory clobbered after examining ipa-max-aa-steps statements modifying memory. so it taps out after a while 😮‍💨.
• ipa-max-switch-predicate-bounds — maximal number of boundary endpoints of case ranges of switch statement. for switch exceeding this limit, IPA-CP will not construct cloning cost predicate, which is used to estimate cloning benefit, for default case of the switch statement. so it just skips it.
• ipa-max-param-expr-ops — IPA-CP will analyze conditional statement that references some function parameter to estimate benefit for cloning upon certain constant value. but if number of operations in a parameter expression exceeds ipa-max-param-expr-ops, the expression is treated as complicated one, and is not handled by IPA analysis. so keep it simple, maybe 🍃.
• lto-partitions — specify desired number of partitions produced during WHOPR compilation. the number of partitions should exceed the number of CPUs used for compilation. so don't skimp.
• lto-min-partition — size of minimal partition for WHOPR (in estimated instructions). this prevents expenses of splitting very small programs into too many partitions. so it's a floor.
• lto-max-partition — size of max partition for WHOPR (in estimated instructions). to provide an upper bound for individual size of partition. meant to be used only with balanced partitioning. so it's a ceiling.
• lto-partition-locality-frequency-cutoff — the denominator n of fraction 1/n of the execution frequency of callee to be cloned for a particular caller. special value of 0 dictates to always clone without a cut-off. so 0 is the 'yolo' setting 🎢.
• lto-partition-locality-size-cutoff — size cut-off for callee including inlined calls to be cloned for a particular caller. that's the size limit.
• lto-max-locality-partition — maximal size of a locality partition for LTO (in estimated instructions). value of 0 results in default value being used. so 0 is 'automatic' 🤖.
• lto-max-streaming-parallelism — maximal number of parallel processes used for LTO streaming. so it's like, how many cores do you want to throw at it? 💥
• cxx-max-namespaces-for-diagnostic-help — the maximum number of namespaces to consult for suggestions when C++ name lookup fails for an identifier. so when you typo, it doesn't search all of them.
• sink-frequency-threshold — the maximum relative execution frequency (in percents) of the target block relative to a statement's original block to allow statement sinking of a statement. larger numbers result in more aggressive statement sinking. a small positive adjustment is applied for statements with memory operands as those are even more profitable so sink. so they get a little boost 🚀.
• max-stores-to-sink — the maximum number of conditional store pairs that can be sunk. set to 0 if either vectorization (-ftree-vectorize) or if-conversion (-ftree-loop-if-convert) is disabled. so if you turn those off, sink no more 🛑.
• case-values-threshold — the smallest number of different values for which it is best to use a jump-table instead of a tree of conditional branches. if the value is 0, use the default for the machine. so it's like, when do we switch to a jump table? 🤷‍♀️
• jump-table-max-growth-ratio-for-size — the maximum code size growth ratio when expanding into a jump table (in percent). the parameter is used when optimizing for size. so it keeps growth in check 📏.
• jump-table-max-growth-ratio-for-speed — the maximum code size growth ratio when expanding into a jump table (in percent). the parameter is used when optimizing for speed. so it's the speed version 💨.
• tree-reassoc-width — set the maximum number of instructions executed in parallel in reassociated tree. this parameter overrides target dependent heuristics used by default if has non zero value. so it's an override 🎮.
• sched-pressure-algorithm — okay so -fsched-pressure has two implementations and this is where you pick 🎀 Algorithm 1 is the O.G. one, and it's the one more likely to stop instructions from getting reordered. Algorithm 2 was designed as a lil compromise between algorithm 1's pretty conservative vibe and the kinda aggressive vibe of the default scheduler. Heads up tho: algorithm 2 leans way harder on having a regular register file and accurate register pressure classes. Want the full lore? It's in haifa-sched.cc in the GCC sources 💅 The default choice? Depends on the target, bestie.
• max-slsr-cand-scan — sets the maximum number of existing candidates that get considered when we're out here seeking a basis for a new straight-line strength reduction candidate ✨
• asan-globals — turns on buffer overflow detection for global objects. This kind of protection is on by default if you're using -fsanitize=address. Wanna turn global objects protection off? Then:

--param asan-globals=0

• asan-stack — turns on buffer overflow detection for stack objects. On by default when you use -fsanitize=address. Disable stack protection with:

--param asan-stack=0

• asan-instrument-reads — buffer overflow detection for memory reads 🔍 On by default with -fsanitize=address. Turn memory reads protection off like:

--param asan-instrument-reads=0

• asan-instrument-writes — buffer overflow detection for memory writes ✍️ On by default with -fsanitize=address. Disable memory writes protection with:

--param asan-instrument-writes=0

• asan-memintrin — turns on detection for built-in functions. On by default when you're using -fsanitize=address. Disable built-in functions protection via:

--param asan-memintrin=0

• asan-use-after-return — turns on detection of use-after-return. On by default when you use the -fsanitize=address option. Disable it with:

--param asan-use-after-return=0

Note: by default this check is disabled at run time. To actually turn it on, add "detect_stack_use_after_return=1" to the environment variable ASAN_OPTIONS 👑

• asan-instrumentation-with-call-threshold — if the number of memory accesses in the function being instrumented is greater or equal to this number, use callbacks instead of inline checks. E.g. to disable inline code, use:

--param asan-instrumentation-with-call-threshold=0

• asan-kernel-mem-intrinsic-prefix — if nonzero, prefix calls to "memcpy", "memset" and "memmove" with __asan_ or __hwasan_ for -fsanitize=kernel-address or -fsanitize=kernel-hwaddress, respectively 💖
• hwasan-instrument-stack — turns on hwasan instrumentation of statically sized stack-allocated variables. This kind of instrumentation is on by default with -fsanitize=hwaddress and off by default with -fsanitize=kernel-hwaddress. To disable stack instrumentation use --param hwasan-instrument-stack=0, and to enable it use --param hwasan-instrument-stack=1 🦋
• hwasan-random-frame-tag — when using stack instrumentation, this decides tags for stack variables using a deterministic sequence beginning at a random tag for each frame. With this parameter unset, tags are chosen using the same sequence but beginning from 1. Enabled by default for -fsanitize=hwaddress and unavailable for -fsanitize=kernel-hwaddress. Disable it with:

--param hwasan-random-frame-tag=0

• hwasan-instrument-allocas — turns on hwasan instrumentation of dynamically sized stack-allocated variables. On by default with -fsanitize=hwaddress, off by default with -fsanitize=kernel-hwaddress. To disable instrumentation of such variables use --param hwasan-instrument-allocas=0, and to enable it use --param hwasan-instrument-allocas=1 ✨
• hwasan-instrument-reads — turns on hwasan checks on memory reads. Instrumentation of reads is on by default for both -fsanitize=hwaddress and -fsanitize=kernel-hwaddress. To disable checking memory reads use:

--param hwasan-instrument-reads=0

• hwasan-instrument-writes — turns on hwasan checks on memory writes. Instrumentation of writes is on by default for both -fsanitize=hwaddress and -fsanitize=kernel-hwaddress. To disable checking memory writes use:

--param hwasan-instrument-writes=0

• hwasan-instrument-mem-intrinsics — turns on hwasan instrumentation of builtin functions. Instrumentation of these builtin functions is on by default for both -fsanitize=hwaddress and -fsanitize=kernel-hwaddress. To disable instrumentation of builtin functions use:

--param hwasan-instrument-mem-intrinsics=0

• use-after-scope-direct-emission-threshold — if the size of a local variable in bytes is smaller or equal to this number, directly poison (or unpoison) shadow memory instead of using run-time callbacks 🌸
• tsan-distinguish-volatile — emit special instrumentation for accesses to volatiles 👀
• tsan-instrument-func-entry-exit — emit instrumentation calls to __tsan_func_entry() and __tsan_func_exit() 💜
• max-fsm-thread-path-insns — maximum number of instructions to copy when duplicating blocks on a finite state automaton jump thread path 🔥
• threader-debug — threader-debug=[none|all] enables verbose dumping of the threader solver. That's the whole vibe 💅
• parloops-chunk-size — chunk size of omp schedule for loops parallelized by parloops ✨
• parloops-schedule — schedule type of omp schedule for loops parallelized by parloops (static, dynamic, guided, auto, runtime) 🎀
• parloops-min-per-thread — the minimum number of iterations per thread of an innermost parallelized loop for which the parallelized variant is preferred over the single threaded one. Quick note tho: for a parallelized loop nest, the minimum number of iterations of the outermost loop per thread is two 🦋
• max-ssa-name-query-depth — maximum depth of recursion when querying properties of SSA names in things like fold routines. One level of recursion = following a use-def chain, iykyk 🔗
• max-speculative-devirt-maydefs — the maximum number of may-defs we analyze when we're looking for a must-def that specifies the dynamic type of an object that invokes a virtual call we may be able to devirtualize speculatively 👑
• ranger-debug — specifies the type of debug output to be issued for ranges 🌈
• unroll-jam-min-percent — the minimum percentage of memory references that must be optimized away for the unroll-and-jam transformation to be considered profitable 💰
• unroll-jam-max-unroll — the maximum number of times the outer loop should be unrolled by the unroll-and-jam transformation 🌀
• max-rtl-if-conversion-unpredictable-cost — maximum permissible cost for the sequence that would be generated by the RTL if-conversion pass for a branch that's considered unpredictable 🎲
• max-variable-expansions-in-unroller — if -fvariable-expansion-in-unroller is used, the maximum number of times that an individual variable will be expanded during loop unrolling ✨
• partial-inlining-entry-probability — maximum probability of the entry BB of split region (in percent relative to entry BB of the function) to make partial inlining happen 📊
• max-tracked-strlens — maximum number of strings for which the strlen optimization pass will track string lengths 🧵
• gcse-after-reload-partial-fraction — the threshold ratio for performing partial redundancy elimination after reload 💅
• gcse-after-reload-critical-fraction — the threshold ratio of critical edges execution count that permit performing redundancy elimination after reload 🔥
• max-loop-header-insns — the maximum number of insns in a loop header duplicated by the copy loop headers pass 👯
• vect-epilogues-nomask — enable loop epilogue vectorization using a smaller vector size 🐣
• vect-partial-vector-usage — controls when the loop vectorizer considers using partial vector loads and stores as an alternative to falling back to scalar code. Okay so: 0 stops the vectorizer from ever using partial vector loads and stores. 1 allows partial vector loads and stores if vectorization removes the need for the code to iterate. 2 allows partial vector loads and stores in all loops. Real talk: this parameter only has an effect on targets that support partial vector loads and stores 🦋
• vect-inner-loop-cost-factor — the maximum factor which the loop vectorizer applies to the cost of statements in an inner loop relative to the loop being vectorized. The factor applied is the maximum of the estimated number of iterations of the inner loop and this parameter. Default value of this parameter is 50 💯
• vect-induction-float — enable loop vectorization of floating point inductions 🌊
• vect-force-slp — force the use of SLP when vectorizing, and fail if not possible 💀
• vrp-block-limit — maximum number of basic blocks before VRP switches to a lower memory algorithm 🧠
• vrp-sparse-threshold — maximum number of basic blocks before VRP uses a sparse bitmap cache 🗺️
• vrp-switch-limit — maximum number of outgoing edges in a switch before VRP will not process it 🚫
• vrp-vector-threshold — maximum number of basic blocks for VRP to use a basic cache vector ⚡
• avoid-fma-max-bits — maximum number of bits for which we avoid creating FMAs 🙈
• fully-pipelined-fma — is FMA fully pipelined on this target or nah? If it's non-zero, reassociation will clock the glow-up from parallelizing FMA's multiplication part and its addition part, assuming FMUL and FMA share the same units that can also do FADD. Bestie is doing the math 💅
• sms-loop-average-count-threshold — a threshold on the average loop count that the swing modulo scheduler cares about.
• sms-dfa-history — how many cycles the swing modulo scheduler looks back on when it's checking conflicts with DFA.
• graphite-allow-codegen-errors — should codegen errors be ICEs when -fchecking is on? Yes/no, you decide.
• sms-max-ii-factor — a factor for tuning that upper bound the swing modulo scheduler uses when it schedules a loop.
• lra-max-considered-reload-pseudos — the max number of reload pseudos that get considered while spilling a non-reload pseudo.
• max-pow-sqrt-depth — maximum depth of sqrt chains to flex when synthesizing exponentiation by a real constant 🌸
• max-dse-active-local-stores — maximum number of active local stores in RTL dead store elimination.
• asan-instrument-allocas — turns on asan allocas/VLAs protection. Protect the vibes 🫶
• max-iterations-computation-cost — the bound on the cost of an expression to compute the number of iterations.
• max-isl-operations — maximum number of isl operations, and 0 means unlimited. Iconic, tbh.
• graphite-max-arrays-per-scop — maximum number of arrays per scop.
• max-vartrack-reverse-op-size — max size of loc list for which reverse ops should be added.
• fsm-scale-path-stmts — the scale factor applied to the number of statements in a threading path crossing a loop backedge, when comparing to --param=max-jump-thread-duplication-stmts.
• uninit-control-dep-attempts — maximum number of nested calls when searching for control dependencies during uninitialized variable analysis.
• uninit-max-chain-len — maximum number of predicates anded for each predicate ored in the normalized predicate chain.
• uninit-max-num-chains — maximum number of predicates ored in the normalized predicate chain.
• uninit-max-prune-work — maximum amount of work done to prune paths where the variable is always initialized.
• sched-autopref-queue-depth — hardware autoprefetcher scheduler model control flag. It's the number of lookahead cycles the model looks into; at ' ' it only enables the instruction sorting heuristic. (Yes, a literal space character. She's quirky like that 💀)
• loop-versioning-max-inner-insns — the maximum number of instructions an inner loop can have before the loop versioning pass decides it's too big to copy.
• loop-versioning-max-outer-insns — the maximum number of instructions an outer loop can have before the loop versioning pass decides it's too big to copy — and it discounts any instructions in inner loops that directly benefit from versioning. Bestie gets a discount card 🎀
• ssa-name-def-chain-limit — the maximum number of SSA_NAME assignments to follow when determining a property of a variable, like its value. This limits the number of iterations or recursive calls GCC performs when optimizing certain statements, or when determining their validity before it issues diagnostics.
• store-merging-max-size — maximum size of a single store merging region, in bytes.
• store-forwarding-max-distance — maximum instruction distance that a small store forwarded to a larger load may stall. A value of '0' disables the cost checks for the avoid-store-forwarding pass.
• hash-table-verification-limit — the number of elements for which hash table verification is done for each searched element.
• max-find-base-term-values — maximum number of VALUEs handled during a single find_base_term call.
• analyzer-max-enodes-per-program-point — the maximum number of exploded nodes per program point within the analyzer, before it terminates analysis of that point. She has a limit, and she will enforce it 😤
• analyzer-max-constraints — the maximum number of constraints per state.
• analyzer-min-snodes-for-call-summary — the minimum number of supernodes within a function before the analyzer even considers summarizing its effects at call sites.
• analyzer-max-enodes-for-full-dump — the maximum depth of exploded nodes that should show up in a dot dump before it switches to a less verbose format.
• analyzer-max-recursion-depth — the maximum number of times a callsite can appear in a call stack within the analyzer, before it terminates analysis of a call that would recurse deeper.
• analyzer-max-svalue-depth — the maximum depth of a symbolic value, before it approximates the value as unknown.
• analyzer-max-infeasible-edges — the maximum number of infeasible edges to reject before declaring a diagnostic as infeasible.
• gimple-fe-computed-hot-bb-threshold — the number of executions of a basic block that counts as hot. This parameter is used only in GIMPLE FE.
• analyzer-bb-explosion-factor — the maximum number of 'after supernode' exploded nodes within the analyzer per supernode, before analysis gets terminated.
• analyzer-text-art-string-ellipsize-threshold — wait, it's analyzer-text-art-string-ellipsis-threshold — the number of bytes at which to ellipsize string literals in analyzer text art diagrams. Art requires limits, periodt ✨
• analyzer-text-art-ideal-canvas-width — the ideal width, in characters, of text art diagrams generated by the analyzer. It's giving Pinterest board 📌
• analyzer-text-art-string-ellipsis-head-len — the number of literal bytes to show at the head of a string literal in text art when ellipsizing it.
• analyzer-text-art-string-ellipsis-tail-len — the number of literal bytes to show at the tail of a string literal in text art when ellipsizing it.
• ranger-logical-depth — maximum depth of logical expression evaluation that ranger will look through when it's evaluating outgoing edge ranges.
• ranger-recompute-depth — maximum depth of instruction chains to consider for recomputation in the outgoing range calculator.
• relation-block-limit — maximum number of relations the oracle will register in a basic block.
• transitive-relations-work-bound — the work bound when discovering transitive relations from existing relations.
• min-pagesize — minimum page size, for warning and early break vectorization purposes.
• openacc-kernels — specify the mode of OpenACC kernels constructs handling. With --param=openacc-kernels=decompose, OpenACC kernels constructs get decomposed into parts — a sequence of compute constructs, each then handled individually. Heads up: this is work in progress 🚧. With --param=openacc-kernels=parloops, OpenACC kernels constructs are handled by the parloops pass, en bloc. This is the current default.
• openacc-privatization — control whether the -fopt-info-omp-note and applicable -fdump-tree-*-details options emit OpenACC privatization diagnostics. With --param=openacc-privatization=quiet, don't diagnose. This is the current default. With --param=openacc-privatization=noisy, do diagnose. Noisy is a personality trait bestie, own it 🔊
• cycle-accurate-model — specifies whether GCC should assume that the scheduling description is mostly a cycle-accurate model of the target processor the code is intended to run on, in the absence of cache misses. Nonzero means the selected scheduling model is accurate and likely describes an in-order processor, and that scheduling should aggressively spill to try and fill any pipeline bubbles. This is the current default. Zero means the scheduling description might not be available/accurate, or maybe not applicable at all — such as for modern out-of-order processors. 🤖

---

The following choices of name are available on AArch64 targets:

• aarch64-vect-compare-costs — when vectorizing, consider using multiple different approaches and let the cost model pick the cheapest one. This includes:
  • Trying both SVE and Advanced SIMD, when SVE is available.
  • Trying to use 64-bit Advanced SIMD vectors for the smallest data elements, rather than using 128-bit vectors for everything.
  • Trying to use "unpacked" SVE vectors for smaller elements. This includes storing smaller elements in larger containers and accessing elements with extending loads and truncating stores.
• aarch64-float-recp-precision — the number of Newton iterations for calculating the reciprocal for float type. The precision of division is proportional to this param when division approximation is enabled. The default value is 1.
• aarch64-double-recp-precision — the number of Newton iterations for calculating the reciprocal for double type. The precision of division is proportional to this param when division approximation is enabled. The default value is 2.
• aarch64-autovec-preference — force an ISA selection strategy for auto-vectorization. The options are:
  • default — use the default heuristics.
  • asimd-only — use only Advanced SIMD for auto-vectorization.
  • sve-only — use only SVE for auto-vectorization.
  • prefer-asimd — use both Advanced SIMD and SVE. Prefer Advanced SIMD when the costs are deemed equal.
  • prefer-sve — use both Advanced SIMD and SVE. Prefer SVE when the costs are deemed equal.
• aarch64-ldp-policy — fine-grained policy for load pairs. With --param=aarch64-ldp-policy=default, use the policy of the tuning structure. This is the current default. With --param=aarch64-ldp-policy=always, emit ldp regardless of alignment. With --param=aarch64-ldp-policy=never, do not emit ldp. With --param=aarch64-ldp-policy=aligned, emit ldp only if the source pointer is aligned to at least double the alignment of the type. Alignment check is literally checking if you're worthy 💅
• --param=aarch64-stp-policy=... — the fine-grained policy for store pairs, bestie. With --param=aarch64-stp-policy=default, use the policy of the tuning structure — this is the current default, periodt. With --param=aarch64-stp-policy=always, emit stp regardless of alignment. With --param=aarch64-stp-policy=never, do not emit stp. With --param=aarch64-stp-policy=aligned, emit stp only if the source pointer is aligned to at least double the alignment of the type. 💅
• --param=aarch64-ldp-alias-check-limit — the limit on the number of alias checks performed by the AArch64 load/store pair fusion pass when it's attempting to form an ldp/stp. Higher values make the pass more aggressive at re-ordering loads over stores, at the expense of increased compile time. So like, more power but slower, you know how it is. 🔥
• --param=aarch64-ldp-writeback — a param to control which writeback opportunities we try to handle in the AArch64 load/store pair fusion pass. A value of zero disables writeback handling. One means we try to form pairs involving one or more existing individual writeback accesses where possible. A value of two means we also try to opportunistically form writeback opportunities by folding in trailing destructive updates of the base register used by a pair. Numbers go up, ambition goes up, slay. ✨
• --param=aarch64-loop-vect-issue-rate-niters — okay so the tuning for some AArch64 CPUs tries to take both latencies and issue rates into account when deciding whether a loop should be vectorized using SVE, vectorized using Advanced SIMD, or not vectorized at all. If this parameter is set to n, GCC will not use this heuristic for loops that are known to execute in fewer than n Advanced SIMD iterations.
• --param=aarch64-vect-unroll-limit — the vectorizer will use available tuning information to determine whether it would be beneficial to unroll the main vectorized loop and by how much. This parameter sets the upper bound of how much the vectorizer will unroll the main loop. The default value is four.

The following choices of name are available on GCN targets:

• --param=gcn-preferred-vectorization-factor — the preferred vectorization factor: default, 32, 64.

The following choices of name are available on i386 and x86_64 targets:

• --param=x86-stlf-window-ninsns — the instructions number above which STFL stall penalty can be compensated.
• --param=x86-stv-max-visits — the maximum number of use and def visits when discovering a STV chain before the discovery is aborted.

---

🌸 Program Instrumentation Options 🌸

GCC supports a number of command-line options that control adding run-time instrumentation to the code it normally generates. Like, it literally goes in and sprinkles little detectors all over your program. For example, one purpose of instrumentation is collecting profiling statistics for use in finding program hot spots, code coverage analysis, or profile-guided optimizations. Another class of program instrumentation is adding run-time checking to detect programming errors like invalid pointer dereferences or out-of-bounds array accesses, as well as deliberately hostile attacks such as stack smashing or C++ vtable hijacking. There is also a general hook which can be used to implement other forms of tracing or function-level instrumentation for debug or program analysis purposes. We love a versatile hook. 🫶

• -p
• -pg — generate extra code to write profile information suitable for the analysis program prof (for -p) or gprof (for -pg). You must use this option when compiling the source files you want data about, and you must also use it when linking. Both things, bestie, no shortcuts.

You can use the function attribute "no_instrument_function" to suppress profiling of individual functions when compiling with these options.

• -fprofile-arcs — add code so that program flow arcs are instrumented. During execution the program records how many times each branch and call is executed and how many times it is taken or returns. On targets that support constructors with priority support, profiling properly handles constructors, destructors and C++ constructors (and destructors) of classes which are used as a type of a global variable.

When the compiled program exits it saves this data to a file called auxname.gcda for each source file. The data may be used for profile-directed optimizations (-fbranch-probabilities), or for test coverage analysis (-ftest-coverage). Each object file's auxname is generated from the name of the output file, if explicitly specified and it is not the final executable, otherwise it is the basename of the source file. In both cases any suffix is removed (e.g. foo.gcda for input file dir/foo.c, or dir/foo.gcda for output file specified as -o dir/foo.o).

Note that if a command line directly links source files, the corresponding .gcda files will be prefixed with the unsuffixed name of the output file. E.g. gcc a.c b.c -o binary would generate binary-a.gcda and binary-b.gcda files. Wild, right? 👀

• -fcondition-coverage — add code so that program conditions are instrumented. During execution the program records what terms in a conditional contributes to a decision, which can be used to verify that all terms in a Boolean function are tested and have an independent effect on the outcome of a decision. The result can be read with gcov --conditions.
• -fpath-coverage — add code so that the paths taken are tracked. During execution the program records the prime paths taken. The number of paths grows very fast with complexity, and to avoid exploding compile times GCC will give up instrumentation if the approximate number of paths exceeds the limit controlled by -fpath-coverage-limit. The result can be read with gcov --prime-paths --prime-paths-lines --prime-paths-source,
• -fpath-coverage-limit=limit — the threshold at which point -fpath-coverage gives up on instrumenting a function. This limit is approximate and conservative, as GCC uses a pessimistic heuristic which slightly overcounts the running number of paths, and gives up if the threshold is reached before finding all the paths. This option is not for fine grained control over which functions to instrument — rather it is intended to limit the effect of path explosion and keep compile times reasonable. The default is 250000.
• --coverage — this option is used to compile and link code instrumented for coverage analysis. The option is a synonym for -fprofile-arcs -ftest-coverage (when compiling) and -lgcov (when linking). See the documentation for those options for more details.
  • Compile the source files with -fprofile-arcs plus optimization and code generation options. For test coverage analysis, use the additional -ftest-coverage option. You do not need to profile every source file in a program.
  • Compile the source files additionally with -fprofile-abs-path to create absolute path names in the .gcno files. This allows gcov to find the correct sources in projects where compilations occur with different working directories.
  • Link your object files with -lgcov or -fprofile-arcs (the latter implies the former).
  • Run the program on a representative workload to generate the arc profile information. This may be repeated any number of times. You can run concurrent instances of your program, and provided that the file system supports locking, the data files will be correctly updated. Unless a strict ISO C dialect option is in effect, "fork" calls are detected and correctly handled without double counting.

Moreover, an object file can be recompiled multiple times and the corresponding .gcda file merges as long as the source file and the compiler options are unchanged.

• For profile-directed optimizations, compile the source files again with the same optimization and code generation options plus -fbranch-probabilities.
• For test coverage analysis, use gcov to produce human readable information from the .gcno and .gcda files. Refer to the gcov documentation for further information.

With -fprofile-arcs, for each function of your program GCC creates a program flow graph, then finds a spanning tree for the graph. Only arcs that are not on the spanning tree have to be instrumented: the compiler adds code to count the number of times that these arcs are executed. When an arc is the only exit or only entrance to a block, the instrumentation code can be added to the block; otherwise, a new basic block must be created to hold the instrumentation code. Efficient queen behaviour, honestly. 👑

With -fcondition-coverage, for each conditional in your program GCC creates a bitset and records the exercised boolean values that have an independent effect on the outcome of that expression.

With -fpath-coverage, GCC finds and enumerates and records the taken prime paths of each function, unless the number of paths would exceed the limit controlled by -fpath-coverage-limit. If the limit is exceeded the function is not instrumented as if -fpath-coverage was not used. A prime path is the longest sequence of unique blocks, except possibly the first and last, which is not a subpath of any other path.

• -ftest-coverage — produce a notes file that the gcov code-coverage utility can use to show program coverage. Each source file's note file is called auxname.gcno. Refer to the -fprofile-arcs option above for a description of auxname and instructions on how to generate test coverage data. Coverage data matches the source files more closely if you do not optimize. (ngl, optimizing and coverage are kinda beefing. 😭)
• -fprofile-abs-path — automatically convert relative source file names to absolute path names in the .gcno files. This allows gcov to find the correct sources in projects where compilations occur with different working directories.
• -fprofile-dir=path — bestie this one tells GCC where to go looking for your profile data files, okay? path is the spot. Now listen, it only matters for profile data made by -fprofile-generate, -ftest-coverage, -fprofile-arcs, and the stuff used by -fprofile-use and -fbranch-probabilities plus all its related options. Both absolute and relative paths work, no biggie. Defaults? current directory. So your profile data file just vibes in the same dir as the object file. To stop file names from beefing with each other, if the object file name isn't an absolute path, GCC mangles the absolute path of the sourcename.gcda file and uses that as the name of the .gcda file. Full deets on the naming live in -fprofile-arcs, and there's a similar option called -fprofile-note. 🎀

Oh and when your executable is running in a huge parallel environment, it's (highly) recommended you save profiles to different folders. You can do that with these variables in path that get exported at run-time:

• %p — the process ID 💅
• %q{VAR} — the value of environment variable VAR

• -fprofile-generate / -fprofile-generate=path — turns on the options usually used to instrument an app so it spits out a profile you can later recompile with profile-feedback-based optimization. Important, no cap: you HAVE to use -fprofile-generate both when compiling and when linking your program. ✨

These options get enabled: -fprofile-arcs, -fprofile-values, -finline-functions, and -fipa-bit-cp.

If you pass path, GCC looks at that path to find the profile feedback data files — see -fprofile-dir for that whole situation.

To actually optimize based on the profile info you collected, use -fprofile-use. 🫶

• -fprofile-info-section / -fprofile-info-section=name — registers the profile info in the section you pick instead of using a constructor/destructor. The section name is name if you give one, otherwise it defaults to ".gcov_info". A pointer to the profile info generated by -fprofile-arcs gets placed in that section for each translation unit. Heads up: this disables the profile info registration through a constructor and disables the profile info processing through a destructor. This is not meant for hosted environments like GNU/Linux — it targets freestanding environments (think embedded systems ngl) with limited resources that don't support constructors/destructors or the C library file I/O. 🦋

The linker could collect the input sections in a continuous memory block and define start/end symbols. Here's a GNU linker script example that defines a linker output section:

.gcov_info      :
{
  PROVIDE (__gcov_info_start = .);
  KEEP (*(.gcov_info))
  PROVIDE (__gcov_info_end = .);
}

And the program could dump the profiling info registered in this linker set, for example like this:

#include <gcov.h>
#include <stdio.h>
#include <stdlib.h>

extern const struct gcov_info *const __gcov_info_start[];
extern const struct gcov_info *const __gcov_info_end[];

static void
dump (const void *d, unsigned n, void *arg)
{
  const unsigned char *c = d;

  for (unsigned i = 0; i < n; ++i)
    printf ("%02x", c[i]);
}

static void
filename (const char *f, void *arg)
{
  __gcov_filename_to_gcfn (f, dump, arg );
}

static void *
allocate (unsigned length, void *arg)
{
  return malloc (length);
}

static void
dump_gcov_info (void)
{
  const struct gcov_info *const *info = __gcov_info_start;
  const struct gcov_info *const *end = __gcov_info_end;

  /* Obfuscate variable to prevent compiler optimizations.  */
  __asm__ ("" : "+r" (info));

  while (info != end)
  {
    void *arg = NULL;
    __gcov_info_to_gcda (*info, filename, dump, allocate, arg);
    putchar ('\n');
    ++info;
  }
}

int
main (void)
{
  dump_gcov_info ();
  return 0;
}

The merge-stream subcommand of gcov-tool may be used to deserialize the data stream generated by the "__gcov_filename_to_gcfn" and "__gcov_info_to_gcda" functions and merge the profile info into .gcda files on the host filesystem. 💖

• -fprofile-note=path — if you give it a path, GCC saves the .gcno file into that path location. But watch out bestie — if you combine this option with multiple source files, the .gcno file will get overwritten. 💀
• -fprofile-prefix-path=path — this one works together with profile-generate=profile_dir and profile-use=profile_dir to tell GCC where the base directory of the built source tree is. By default profile_dir will contain files with mangled absolute paths of all object files in the built project. That's annoying when the directory used to build the instrumented binary is different from the directory used to build the binary that's optimized with profile feedback — because then the profile data won't be found during the optimized build. 😭 In those setups, -fprofile-prefix-path=path with path pointing to the base directory of the build can be used to strip the irrelevant part of the path and keep all file names relative to the main build directory. slay.
• -fprofile-prefix-map=old=new — when compiling files that live in directory old, record profiling info (with --coverage) describing them as if they lived in directory new instead. Also peep -ffile-prefix-map and -fcanon-prefix-map. ✨
• -fprofile-update=method — changes the update method for an app instrumented for profile-feedback-based optimization. The method argument should be one of single, atomic, or prefer-atomic. The first one is great for single-threaded apps, while the second one stops profile corruption by emitting thread-safe code. 🔥

Warning: when an app does not properly join all threads (or creates a detached thread), a profile file can still be corrupted. 🙈

Using prefer-atomic gets transformed into either atomic (when the target supports it) or single otherwise. The GCC driver automatically picks prefer-atomic when -pthread is in the command line; otherwise the default method is single.

If atomic is selected, the profile info is updated using atomic operations on a best-effort basis. Ideally it's updated through atomic operations in hardware. But if the target platform doesn't support the required atomic ops in hardware and libatomic is available, then it updates through calls to libatomic. And if the target platform supports neither the required hardware atomic ops nor libatomic, then the profile info is not atomically updated and a warning is issued — in this case the profiling info you get may be corrupt for multi-threaded apps. 😭

For performance reasons, if 64-bit counters are used for the profiling info and the target platform only supports 32-bit atomic operations in hardware, then the performance-critical profiling updates are done using two 32-bit atomic operations per counter update. And if a signal interrupts those two operations while updating a counter, the profiling info may end up in an inconsistent state. 💀

• -fprofile-filter-files=regex — instrument only functions from files whose name matches any of the regular expressions (separated by semi-colons).

For example, -fprofile-filter-files=main\.c;module.*\.c will instrument only main.c and all C files starting with 'module'. 🎀

• -fprofile-exclude-files=regex — instrument only functions from files whose name does not match any of the regular expressions (separated by semi-colons).

For example, -fprofile-exclude-files=/usr/.* will prevent instrumentation of all files located in the /usr/ folder. 🌸

• -fprofile-reproducible=[multithreaded|parallel-runs|serial] — controls the level of reproducibility of the profile gathered by "-fprofile-generate". This makes it possible to rebuild a program with the same outcome, which is useful for stuff like distribution packages fr fr.

With -fprofile-reproducible=serial, the profile gathered by -fprofile-generate is reproducible provided the trained program behaves the same at each invocation of the train run, it's not multi-threaded, and profile data streaming is always done in the same order. Quick note: profile streaming happens at the end of a program run, but also before the "fork" function is invoked. 👑

And ngl, it's pretty common that execution counts of some parts of programs depend on things like the length of temporary file names or memory space randomization (which may affect hash-table collision rate). Those non-reproducible parts can be annotated with the "no_instrument_function" function attribute. You can use gcov-dump with -l to dump the gathered data and verify it's actually reproducible. ✨

With -fprofile-reproducible=parallel-runs, the collected profile stays reproducible regardless of the order the data is streamed into the gcda files. This setting is what makes it possible to run multiple instances of an instrumented program in parallel (like with "make -j"). Lowkey it does reduce the quality of the gathered data though, especially for indirect call profiling. 💅

• -fsanitize=address — ok bestie, this one turns on AddressSanitizer, aka the fast memory error detector, aka the reason you're about to find out your code was cursed all along 💅. Memory access instructions get instrumented so it can catch out-of-bounds and use-after-free bugs. Turning this on also enables -fsanitize-address-use-after-scope. More deets at <https://github.com/google/sanitizers/wiki/AddressSanitizer>. You can vibe-check the run-time behavior with the ASAN_OPTIONS environment variable — set it to "help=1" and the available options get shown at startup of the instrumented program. Full list of supported options lives at <https://github.com/google/sanitizers/wiki/AddressSanitizerFlags#run-time-flags>. This option cannot be combined with -fsanitize=thread or -fsanitize=hwaddress. Also note that the only targets -fsanitize=hwaddress is currently supported on are x86-64 (only with -mlam=u48 or -mlam=u57 options) and AArch64 — and in both cases only in ABIs with 64-bit pointers.

When compiling with -fsanitize=address, you should also use -g so the output actually means something. To get more accurate stack traces, you can use options such as -O0, -O1, or -Og (which, for instance, prevent most function inlining), -fno-optimize-sibling-calls (which prevents optimizing sibling and tail recursive calls; this option is implicit for -O0, -O1, or -Og), or -fno-ipa-icf (which disables Identical Code Folding for functions). Using -fno-omit-frame-pointer also improves stack traces. And since multiple runs of the program may yield backtraces with different addresses due to ASLR (Address Space Layout Randomization), it may be desirable to turn ASLR off. On Linux, this can be achieved with setarch \uname -m` -R ./prog`.

• -fsanitize=kernel-address — turns on AddressSanitizer but for the Linux kernel 🐧✨. More details at <https://github.com/google/kernel-sanitizers>.
• -fsanitize=hwaddress — enables Hardware-assisted AddressSanitizer, which uses a hardware ability to ignore the top byte of a pointer so it can detect memory errors with a low memory overhead 🦋. Memory access instructions are instrumented to detect out-of-bounds and use-after-free bugs. The option enables -fsanitize-address-use-after-scope. More details at <https://clang.llvm.org/docs/HardwareAssistedAddressSanitizerDesign.html>. Run-time behavior is influenced using the HWASAN_OPTIONS environment variable — set it to "help=1" and the available options are shown at startup of the instrumented program. This option cannot be combined with -fsanitize=thread or -fsanitize=address, and is currently only available on AArch64.
• -fsanitize=kernel-hwaddress — enables Hardware-assisted AddressSanitizer for compilation of the Linux kernel. Similar to -fsanitize=kernel-address but using an alternate instrumentation method, and similar to -fsanitize=hwaddress but with instrumentation differences necessary for compiling the Linux kernel. Those differences are to avoid hwasan library initialization calls and to account for the stack pointer having a different value in its top byte.

Note: this option has different defaults to -fsanitize=hwaddress. Instrumenting the stack and alloca calls are not on by default but are still possible by specifying the command-line options --param hwasan-instrument-stack=1 and --param hwasan-instrument-allocas=1 respectively. Using a random frame tag is not implemented for kernel instrumentation.

• -fsanitize=pointer-compare — instruments comparison operation (<, <=, >, >=) with pointer operands 👀. The option must be combined with either -fsanitize=kernel-address or -fsanitize=address. The option cannot be combined with -fsanitize=thread. Note: by default the check is disabled at run time. To enable it, add "detect_invalid_pointer_pairs=2" to the environment variable ASAN_OPTIONS. Using "detect_invalid_pointer_pairs=1" detects invalid operation only when both pointers are non-null.
• -fsanitize=pointer-subtract — instruments subtraction with pointer operands. The option must be combined with either -fsanitize=kernel-address or -fsanitize=address. The option cannot be combined with -fsanitize=thread. Note: by default the check is disabled at run time. To enable it, add "detect_invalid_pointer_pairs=2" to the environment variable ASAN_OPTIONS. Using "detect_invalid_pointer_pairs=1" detects invalid operation only when both pointers are non-null.
• -fsanitize=shadow-call-stack — enables ShadowCallStack, a security enhancement mechanism used to protect programs against return address overwrites (e.g. stack buffer overflows) 🛡️✨. It works by saving a function's return address to a separately allocated shadow call stack in the function prologue and restoring the return address from the shadow call stack in the function epilogue. Instrumentation only occurs in functions that need to save the return address to the stack.

Currently it only supports the aarch64 platform. It is specifically designed for linux kernels that enable the CONFIG_SHADOW_CALL_STACK option. For user space programs, runtime support is not currently provided in libc and libgcc. Users who want to use this feature in user space need to provide their own support for the runtime. It should be noted that this may cause the ABI rules to be broken 😭.

On aarch64, the instrumentation makes use of the platform register "x18". This generally means that any code that may run on the same thread as code compiled with ShadowCallStack must be compiled with the flag -ffixed-x18, otherwise functions compiled without -ffixed-x18 might clobber "x18" and so corrupt the shadow stack pointer 💀.

Also, because there is no userspace runtime support, code compiled with ShadowCallStack cannot use exception handling. Use -fno-exceptions to turn off exceptions.

See <https://clang.llvm.org/docs/ShadowCallStack.html> for more details.

• -fsanitize=thread — enables ThreadSanitizer, a fast data race detector 🏃♀️💨. Memory access instructions are instrumented to detect data race bugs. See <https://github.com/google/sanitizers/wiki#threadsanitizer> for more details. The run-time behavior can be influenced using the TSAN_OPTIONS environment variable; see <https://github.com/google/sanitizers/wiki/ThreadSanitizerFlags> for a list of supported options. The option cannot be combined with -fsanitize=address, -fsanitize=leak.

When compiling with -fsanitize=thread, you should also use -g to produce more meaningful output.

Note that sanitized atomic builtins cannot throw exceptions when operating on invalid memory addresses with non-call exceptions (-fnon-call-exceptions).

• -fsanitize=leak — enables LeakSanitizer, a memory leak detector 🫧. This option only matters for linking of executables. The executable is linked against a library that overrides "malloc" and other allocator functions. See <https://github.com/google/sanitizers/wiki/AddressSanitizerLeakSanitizer> for more details. The run-time behavior can be influenced using the LSAN_OPTIONS environment variable. The option cannot be combined with -fsanitize=thread.
• -fsanitize=undefined — enables UndefinedBehaviorSanitizer, a fast undefined behavior detector 🔥. Various computations are instrumented to detect undefined behavior at runtime. See <https://clang.llvm.org/docs/UndefinedBehaviorSanitizer.html> for more details. The run-time behavior can be influenced using the UBSAN_OPTIONS environment variable. Current suboptions are, and this list is a whole mood:
  • -fsanitize=shift — this option enables checking that the result of a shift operation is not undefined. Note that what exactly is considered undefined differs slightly between C and C++, as well as between ISO C90 and C99, etc. This option has two suboptions, -fsanitize=shift-base and -fsanitize=shift-exponent.
  • -fsanitize=shift-exponent — this option enables checking that the second argument of a shift operation is not negative and is smaller than the precision of the promoted first argument.
  • -fsanitize=shift-base — if the second argument of a shift operation is within range, check that the result of a shift operation is not undefined. Note that what exactly is considered undefined differs slightly between C and C++, as well as between ISO C90 and C99, etc.
  • -fsanitize=integer-divide-by-zero — detect integer division by zero. Yeah, that one. 🚫➗0️⃣
  • -fsanitize=unreachable — with this option, the compiler turns the "__builtin_unreachable" call into a diagnostics message call instead. When reaching the "__builtin_unreachable" call, the behavior is undefined. So, like, don't 🙈.
  • -fsanitize=vla-bound — this option instructs the compiler to check that the size of a variable length array is positive.
  • -fsanitize=null — this option enables pointer checking. Particularly, the application built with this option turned on will issue an error message when it tries to dereference a NULL pointer, or if a reference (possibly an rvalue reference) is bound to a NULL pointer, or if a method is invoked on an object pointed by a NULL pointer.
  • -fsanitize=return — this option enables return statement checking. Programs built with this option turned on will issue an error message when the end of a non-void function is reached without actually returning a value. This option works in C++ only.
  • -fsanitize=signed-integer-overflow — this option enables signed integer overflow checking. We check that the result of "+", "*", and both unary and binary "-" does not overflow in the signed arithmetics. This also detects "INT_MIN / -1" signed division. Note, integer promotion rules must be taken into account. That is, the following is not an overflow:

signed char a = SCHAR_MAX;
a++;

• -fsanitize=bounds — okay so this one turns on instrumentation of array bounds ✨ and various out of bounds accesses get detected. But listen, bestie: flexible array members, flexible array member-like arrays, and initializers of variables with static storage are not instrumented 🚫 — with one exception, being flexible array member-like arrays for which -fstrict-flex-arrays or -fstrict-flex-arrays= options or strict_flex_array attributes say they shouldn't be treated like flexible array member-like arrays. iykyk 💅
• -fsanitize=bounds-strict — the strict glow-up of the above 🎀 it enables strict instrumentation of array bounds, so most out of bounds accesses are detected, including flexible array member-like arrays. Initializers of variables with static storage are still not instrumented though, no cap.
• -fsanitize=alignment — this one is the vibe check for pointers 🔍 it enables checking of alignment of pointers when they are dereferenced, or when a reference is bound to an insufficiently aligned target, or when a method or constructor is invoked on an insufficiently aligned object.
• -fsanitize=object-size — enables instrumentation of memory references using the __builtin_dynamic_object_size function, and various out of bounds pointer accesses are detected. lowkey iconic 🌸
• -fsanitize=float-divide-by-zero — detects floating-point division by zero. Unlike other similar options, -fsanitize=float-divide-by-zero is not enabled by -fsanitize=undefined, since floating-point division by zero can be a legitimate way of obtaining infinities and NaNs. so it's giving "it's not a bug, it's a feature" 💀
• -fsanitize=float-cast-overflow — enables floating-point type to integer conversion checking, and we check that the result of the conversion does not overflow. Just like the float-divide one, -fsanitize=float-cast-overflow is not enabled by -fsanitize=undefined. Also note: this option does not work well with FE_INVALID exceptions enabled 😭
• -fsanitize=nonnull-attribute — enables instrumentation of calls, checking whether null values are not passed to arguments marked as requiring a non-null value by the nonnull function attribute.
• -fsanitize=returns-nonnull-attribute — enables instrumentation of return statements in functions marked with the returns_nonnull function attribute, to detect returning of null values from such functions. main character functions only, no nulls allowed 👑
• -fsanitize=bool — enables instrumentation of loads from bool. If a value other than 0/1 is loaded, a run-time error is issued.
• -fsanitize=enum — enables instrumentation of loads from an enum type. If a value outside the range of values for the enum type is loaded, a run-time error is issued.
• -fsanitize=vptr — enables instrumentation of C++ member function calls, member accesses and some conversions between pointers to base and derived classes, to verify the referenced object has the correct dynamic type ✨ very "are you who you say you are" of it.
• -fsanitize=pointer-overflow — enables instrumentation of pointer arithmetics. If the pointer arithmetics overflows, a run-time error is issued.
• -fsanitize=builtin — enables instrumentation of arguments to selected builtin functions. If an invalid value is passed to such arguments, a run-time error is issued 🚨 e.g. passing 0 as the argument to __builtin_ctz or __builtin_clz invokes undefined behavior and is diagnosed by this option.

Note that sanitizers tend to increase the rate of false positive warnings, most notably those around -Wmaybe-uninitialized. We recommend against combining -Werror and [the use of] sanitizers. 🙈

While -ftrapv causes traps for signed overflows to be emitted, -fsanitize=undefined gives a diagnostic message. This currently works only for the C family of languages.

• -fno-sanitize=all — this one disables all previously enabled sanitizers 🛑 careful bestie. -fsanitize=all is not allowed, as some sanitizers cannot be used together.
• -fasan-shadow-offset=number — forces GCC to use a custom shadow offset in AddressSanitizer checks. It is useful for experimenting with different shadow memory layouts in Kernel AddressSanitizer. science vibes 🧪
• -fsanitize-sections=s1,s2,... — sanitize global variables in selected user-defined sections. si may contain wildcards.
• -fsanitize-recover[=opts] — -fsanitize-recover= controls error recovery mode for sanitizers mentioned in the comma-separated list of opts. Enabling this option for a sanitizer component causes it to attempt to continue running the program as if no error happened 😌 which means multiple runtime errors can be reported in a single program run, and the exit code of the program may indicate success even when errors have been reported 💀. The -fno-sanitize-recover= option can be used to alter this behavior: only the first detected error is reported and the program then exits with a non-zero exit code.

Currently this feature only works for -fsanitize=undefined (and its suboptions except for -fsanitize=unreachable and -fsanitize=return), -fsanitize=float-cast-overflow, -fsanitize=float-divide-by-zero, -fsanitize=bounds-strict, -fsanitize=kernel-address and -fsanitize=address. For these sanitizers error recovery is turned on by default, except -fsanitize=address, for which this feature is experimental. -fsanitize-recover=all and -fno-sanitize-recover=all is also accepted, the former enables recovery for all sanitizers that support it, the latter disables recovery for all sanitizers that support it.

Even if a recovery mode is turned on the compiler side, it needs to be also enabled on the runtime library side, otherwise the failures are still fatal 🫠 The runtime library defaults to halt_on_error=0 for ThreadSanitizer and UndefinedBehaviorSanitizer, while the default value for AddressSanitizer is halt_on_error=1. This can be overridden through setting the halt_on_error flag in the corresponding environment variable.

Syntax without an explicit opts parameter is deprecated. It is equivalent to specifying an opts list of:

undefined,float-cast-overflow,float-divide-by-zero,bounds-strict

• -fsanitize-address-use-after-scope — enables sanitization of local variables to detect use-after-scope bugs. The option sets -fstack-reuse to none.
• -fsanitize-trap[=opts] — the -fsanitize-trap= option instructs the compiler to report for sanitizers mentioned in the comma-separated list of opts undefined behavior using __builtin_trap rather than a libubsan library routine. If this option is enabled for a certain sanitizer, it takes precedence over the -fsanitizer-recover= for that sanitizer, __builtin_trap will be emitted and be fatal regardless of whether recovery is enabled or disabled using -fsanitize-recover=. ⚡

The advantage of this is that the libubsan library is not needed and is not linked in, so this is usable even in freestanding environments. slay 🫶

Currently this feature works with -fsanitize=undefined (and its suboptions except for -fsanitize=vptr), -fsanitize=float-cast-overflow, -fsanitize=float-divide-by-zero and -fsanitize=bounds-strict. -fsanitize-trap=all can be also specified, which enables it for undefined suboptions, -fsanitize=float-cast-overflow, -fsanitize=float-divide-by-zero and -fsanitize=bounds-strict. If -fsanitize-trap=undefined or -fsanitize-trap=all is used and -fsanitize=vptr is enabled on the command line, the instrumentation is silently ignored as the instrumentation always needs libubsan support, -fsanitize-trap=vptr is not allowed 🚫

• -fsanitize-undefined-trap-on-error — the -fsanitize-undefined-trap-on-error option is a deprecated equivalent of -fsanitize-trap=all.
• -fsanitize-coverage=trace-pc — enables coverage-guided fuzzing code instrumentation. Inserts a call to __sanitizer_cov_trace_pc into every basic block. every. single. one. 💅
• -fsanitize-coverage=trace-cmp — enables dataflow guided fuzzing code instrumentation. Inserts a call to __sanitizer_cov_trace_cmp1, __sanitizer_cov_trace_cmp2, __sanitizer_cov_trace_cmp4 or __sanitizer_cov_trace_cmp8 for integral comparison with both operands variable, or __sanitizer_cov_trace_const_cmp1, __sanitizer_cov_trace_const_cmp2, __sanitizer_cov_trace_const_cmp4 or __sanitizer_cov_trace_const_cmp8 for integral comparison with one operand constant, __sanitizer_cov_trace_cmpf or __sanitizer_cov_trace_cmpd for float or double comparisons, and __sanitizer_cov_trace_switch for switch statements ✨
• -fcf-protection=[full|branch|return|none|check], -fcf-protection — enables code instrumentation to increase program security by checking that target addresses of control-flow transfer instructions (such as indirect function call, function return, indirect jump) are valid 🔒 this prevents diverting the flow of control to an unexpected target. It's intended to protect against such threats as Return-oriented Programming (ROP), and similarly call/jmp-oriented programming (COP/JOP). no sneaky redirects allowed bestie.

The -fcf-protection= keywords are interpreted as follows.

• The value branch tells the compiler to implement checking of validity of control-flow transfer at the point of indirect branch instructions, i.e. call/jmp instructions.
• The value return implements checking of validity at the point of returning from a function.
• The value full is an alias for specifying both branch and return.
• The value check is used for the final link with link-time optimization (LTO). An error is issued if LTO object files are compiled with different -fcf-protection values. The value check is ignored at the compile time.
• The value none turns off instrumentation.

-fcf-protection is an alias for -fcf-protection=full. To override a previous -fcf-protection option on the command line, add -fcf-protection=none and then -fcf-protection=kind.

The macro __CET__ is defined when -fcf-protection is used. The first bit of __CET__ is set to 1 for the value branch and the second bit of __CET__ is set to 1 for the return. 🌈

You can also use the nocf_check attribute to identify which functions and calls should be skipped from instrumentation.

Currently the x86 GNU/Linux target ships an implementation based on Intel Control-flow Enforcement Technology (CET) that works for i686 processor or newer. 🔥

NOTE: In Ubuntu 19.10 and later versions, -fcf-protection is enabled by default for C, C++, ObjC, ObjC++, if none of -fno-cf-protection nor -fcf-protection=* are found. 💅

---

• -fharden-compares — okay so for every single logical test that survives gimple optimizations and is not the condition in a conditional branch (think: conditions tested for conditional moves, or conditions you're storing in boolean variables), GCC emits extra code that computes and verifies the reversed condition, and calls __builtin_trap if the results don't match. 💀 Use it with -fharden-conditional-branches and you've covered all the conditionals, periodt.
• -fharden-conditional-branches — for every non-vectorized conditional branch that survives gimple optimizations, emit extra code to compute and verify the reversed condition, and to call __builtin_trap if the result is unexpected. Pair it with -fharden-compares to cover all conditionals. 💖
• -fharden-control-flow-redundancy — emit extra code to set booleans when entering basic blocks, and to verify and trap, at function exits, when the booleans do not form an execution path that is compatible with the control flow graph. This is giving paranoid but make it fashion ✨

The verification happens before returns, before mandatory tail calls (see below) and, optionally, before escaping exceptions with -fhardcfr-check-exceptions, before returning calls with -fhardcfr-check-returning-calls, and before noreturn calls with -fhardcfr-check-noreturn-calls. There's tuning to be done too: --param hardcfr-max-blocks and --param hardcfr-max-inline-blocks are available. 🎛️

So here's the tea ☕: tail call optimization takes place too late to affect control flow redundancy. But calls annotated as mandatory tail calls by language front-ends, and any calls marked early enough as potential tail calls, would also have verification issued before the call — but bestie, these possibilities are merely theoretical, because these conditions can only be met when using custom compiler plugins. 🙈

• -fhardcfr-skip-leaf — disables -fharden-control-flow-redundancy in leaf functions. bye leaf functions, this one's not for you 🍃
• -fhardcfr-check-exceptions — when -fharden-control-flow-redundancy is active, check the recorded execution path against the control flow graph at exception escape points, like the function body was wrapped with a cleanup handler that did the check and reraised. This one is enabled by default; use -fno-hardcfr-check-exceptions to disable it. 🫶
• -fhardcfr-check-returning-calls — when -fharden-control-flow-redundancy is active, check the recorded execution path against the control flow graph before any function call that's immediately followed by a return of its result, if any — all so it doesn't prevent tail-call optimization, whether or not it's ultimately optimized to a tail call. 💅

This option is enabled by default whenever sibling call optimizations are enabled (see -foptimize-sibling-calls), but you can enable it (or disable it, using its negated form) explicitly, regardless of the optimizations. 👑

• -fhardcfr-check-noreturn-calls=[always|no-xthrow|nothrow|never] — when -fharden-control-flow-redundancy is active, check the recorded execution path against the control flow graph before noreturn calls. Which ones? Your call: all of them (always), the ones that aren't expected to return control to the caller through an exception (no-xthrow, which is the default), the ones that may not return control to the caller through an exception either (nothrow), or none of them (never). ✨

Checking before a noreturn function that may return control to the caller through an exception can cause checking to be performed more than once, if the exception is caught in the caller — whether by a handler or a cleanup. 😭 When -fhardcfr-check-exceptions is also enabled, the compiler will avoid associating a noreturn call with the implicitly-added cleanup handler, since that would be redundant with the check performed before the call. But other handlers or cleanups in the function, if activated, will modify the recorded execution path and check it again when another checkpoint is hit. And the checkpoint may even be another noreturn call, so checking may end up performed multiple times. 💀

Various optimizers may cause calls to be marked as noreturn and/or nothrow, even in the absence of the corresponding attributes, which may affect the placement of checks before calls, as well as the addition of implicit cleanup handlers for them. This unpredictability — plus the fact that raising and reraising exceptions frequently amounts to implicitly calling noreturn functions — has made no-xthrow the default setting for this option: it excludes from the noreturn treatment only internal functions used to (re)raise exceptions, that are not affected by these optimizations. 🌈

• -fhardened — enable a whole set of flags for C and C++ that improve the security of the generated code without affecting its ABI. The precise flags enabled may change between major releases of GCC, but currently they are:

-D_FORTIFY_SOURCE=3 -D_GLIBCXX_ASSERTIONS
-ftrivial-auto-var-init=zero -fPIE  -pie  -Wl,-z,relro,-z,now
-fstack-protector-strong -fstack-clash-protection
-fcf-protection=full (x86 GNU/Linux only)

You can generate the list of options enabled by -fhardened using the --help=hardened option. 📝

When the system glibc is older than 2.35, -D_FORTIFY_SOURCE=2 is used instead. 🧚

This option is meant for production builds, not merely debug builds. fr fr.

Currently, -fhardened is only supported on GNU/Linux targets.

Also! -fhardened only enables a particular option if it wasn't already specified anywhere on the command line. For instance, -fhardened -fstack-protector will only enable -fstack-protector, but not -fstack-protector-strong. Sneaky little detail, iykyk. 💅

• -fstack-protector — emit extra code to check for buffer overflows, like stack smashing attacks. It works by adding a guard variable to functions with vulnerable objects — that includes functions that call alloca, and functions with buffers larger than or equal to 8 bytes. The guards are initialized when a function is entered and then checked when the function exits. If a guard check fails, an error message is printed and the program exits. 💀 Only variables that are actually allocated on the stack count — optimized away variables or variables allocated in registers don't. 🎀
• -fstack-protector-all — like -fstack-protector except that all functions are protected. all of them. every single one. 👑
• -fstack-protector-strong — like -fstack-protector but it includes additional functions to be protected — those that have local array definitions, or have references to local frame addresses. Only variables that are actually allocated on the stack are considered; optimized away variables or variables allocated in registers don't count. 🌸

NOTE: In Ubuntu 14.10 and later versions, -fstack-protector-strong is enabled by default for C, C++, ObjC, ObjC++, if none of -fno-stack-protector, -nostdlib, nor -ffreestanding are found. ✨

• -fstack-protector-explicit — like -fstack-protector but it only protects those functions which have the stack_protect attribute. picky, and honestly? iconic for it 💅
• -fstack-check — generate code to verify that you do not go beyond the boundary of the stack. You should specify this flag if you're running in an environment with multiple threads, but you only rarely need to specify it in a single-threaded environment, since stack overflow is automatically detected on nearly all systems if there is only one stack. 🧵

Heads up bestie: this switch does not actually cause checking to be done — the operating system or the language runtime has to do that. The switch causes generation of code to ensure that they see the stack being extended. 👀

You can additionally specify a string parameter: no means no checking, generic means force the use of old-style checking, specific means use the best checking method and is equivalent to bare -fstack-check.

Old-style checking is a generic mechanism that requires no specific target support in the compiler, but it comes with the following drawbacks: 😭

1. Modified allocation strategy for large objects: they are always allocated dynamically if their size exceeds a fixed threshold. Note this may change the semantics of some code.
2. Fixed limit on the size of the static frame of functions: when it is topped by a particular function, stack checking is not reliable and a warning is issued by the compiler.
3. Inefficiency: because of both the modified allocation strategy and the generic implementation, code performance is hampered.

Note that old-style stack checking is also the fallback method for specific if no target support has been added in the compiler. 🫶

-fstack-check= is designed for Ada's needs to detect infinite recursion and stack overflows. specific is an excellent choice when compiling Ada code. But — and this matters — it is not generally sufficient to protect against stack-clash attacks. To protect against those you want -fstack-clash-protection. 💥

• -fstack-clash-protection — generate code to prevent stack clash style attacks. When this option is enabled, the compiler will only allocate one page of stack space at a time, and each page is accessed immediately after allocation. Thus, it prevents allocations from jumping over any stack guard page provided by the operating system. 🛡️✨

Most targets do not fully support stack clash protection. However, on those targets -fstack-clash-protection will protect dynamic stack allocations. -fstack-clash-protection may also provide limited protection for static stack allocations if the target supports -fstack-check=specific. 🧚

NOTE: In Ubuntu 19.10 and later versions, -fstack-clash-protection is enabled by default for C, C++, ObjC, ObjC++, unless -fno-stack-clash-protection is found. 💖

• -fstack-limit-register=reg, -fstack-limit-symbol=sym, -fno-stack-limit — generate code to ensure that the stack does not grow beyond a certain value, either the value of a register or the address of a symbol. If a larger stack is required, a signal is raised at run time. 🚨 For most targets, the signal is raised before the stack overruns the boundary, so it's possible to catch the signal without taking special precautions. 🌸

For instance, if the stack starts at absolute address 0x80000000 and grows downwards, you can use the flags -fstack-limit-symbol=__stack_limit and -Wl,--defsym,__stack_limit=0x7ffe0000 to enforce a stack limit of 128KB. Note that this may only work with the GNU linker. 🫶

You can locally override stack limit checking by using the "no_stack_limit" function attribute. lowkey iconic that you can just... opt out per-function 💅

• -fsplit-stack — bestie this one generates code to automatically split the stack before it overflows 🌊✨ The resulting program has a discontiguous stack which can only overflow if the program is unable to allocate any more memory. It's most useful when you're running threaded programs, because it's no longer necessary to calculate a good stack size to use for each thread — no more doing math you don't wanna do fr fr. This is currently only implemented for the x86 targets running GNU/Linux. heads up on that one 👀

When code compiled with -fsplit-stack calls code compiled without -fsplit-stack, there may not be much stack space available for the latter code to run 😭 If compiling all code, including library code, with -fsplit-stack is not an option, then the linker can fix up these calls so that the code compiled without -fsplit-stack always has a large stack. Support for this is implemented in the gold linker in GNU binutils release 2.21 and later. 🏆

• -fstrub=disable — Disable stack scrubbing entirely, ignoring any "strub" attributes. See 🫧
• -fstrub=strict — Functions default to "strub" mode "disabled", and apply strictly the restriction that only functions associated with "strub"-"callable" modes ("at-calls", "callable" and "always_inline" "internal") are "callable" by functions with "strub"-enabled modes ("at-calls" and "internal"). strict mode said no exceptions and meant it 💀
• -fstrub=relaxed — Restore the default stack scrub ("strub") setting, namely, "strub" is only enabled as required by "strub" attributes associated with function and data types. "Relaxed" means that strub contexts are only prevented from calling functions explicitly associated with "strub" mode "disabled". This option is only useful to override other -fstrub=* options that precede it in the command line. ✨ (yes it's a "last one wins" situation, iykyk)
• -fstrub=at-calls — Enable "at-calls" "strub" mode where viable. The primary use of this option is for testing 🧪 It exercises the "strub" machinery in scenarios strictly local to a translation unit. This "strub" mode modifies function interfaces, so any function that is visible to other translation units, or that has its address taken, will not be affected by this option. Optimization options may also affect viability. See the "strub" attribute documentation for details on viability and eligibility requirements. 📖
• -fstrub=internal — Enable "internal" "strub" mode where viable. The primary use of this option is for testing 🧪 This option is intended to exercise thoroughly parts of the "strub" machinery that implement the less efficient, but interface-preserving "strub" mode. Functions that would not be affected by this option are quite uncommon. rare edge case spotted 👀
• -fstrub=all — Enable some "strub" mode where viable. When both strub modes are viable, "at-calls" is preferred. -fdump-ipa-strubm adds function attributes that tell which mode was selected for each function. The primary use of this option is for testing, to exercise thoroughly the "strub" machinery. 🔬
• -fvtable-verify=[std|preinit|none] — This option is only available when compiling C++ code 💜 It turns on (or off, if using -fvtable-verify=none) the security feature that verifies at run time, for every virtual call, that the vtable pointer through which the call is made is valid for the type of the object, and has not been corrupted or overwritten. If an invalid vtable pointer is detected at run time, an error is reported and execution of the program is immediately halted. like, full stop, no mercy 🛑

This option causes run-time data structures to be built at program startup, which are used for verifying the vtable pointers. The options std and preinit control the timing of when these data structures are built. In both cases the data structures are built before execution reaches main. Using -fvtable-verify=std causes the data structures to be built after shared libraries have been loaded and initialized. -fvtable-verify=preinit causes them to be built before shared libraries have been loaded and initialized. ⏰

If this option appears multiple times in the command line with different values specified, none takes highest priority over both std and preinit; preinit takes priority over std. priority order, do not @ me 🎀

• -fvtv-debug — When used in conjunction with -fvtable-verify=std or -fvtable-verify=preinit, causes debug versions of the runtime functions for the vtable verification feature to be called. This flag also causes the compiler to log information about which vtable pointers it finds for each class. 📝 This information is written to a file named vtv_set_ptr_data.log in the directory named by the environment variable VTV_LOGS_DIR if that is defined or the current working directory otherwise.

Note: This feature appends data to the log file. If you want a fresh log file, be sure to delete any existing one. 🗑️ (so real — stale logs are such a vibe killer)

• -fvtv-counts — This is a debugging flag 🐛 When used in conjunction with -fvtable-verify=std or -fvtable-verify=preinit, this causes the compiler to keep track of the total number of virtual calls it encounters and the number of verifications it inserts. It also counts the number of calls to certain run-time library functions that it inserts and logs this information for each compilation unit. The compiler writes this information to a file named vtv_count_data.log in the directory named by the environment variable VTV_LOGS_DIR if that is defined or the current working directory otherwise. It also counts the size of the vtable pointer sets for each class, and writes this information to vtv_class_set_sizes.log in the same directory. 📊

Note: This feature appends data to the log files. To get fresh log files, be sure to delete any existing ones. 🗑️✨

• -finstrument-functions — Generate instrumentation calls for entry and exit to functions 🎬 Just after function entry and just before function exit, the following profiling functions are called with the address of the current function and its call site. (On some platforms, "__builtin_return_address" does not work beyond the current function, so the call site information may not be available to the profiling functions otherwise.)

void __cyg_profile_func_enter (void *this_fn,
                               void *call_site);
void __cyg_profile_func_exit  (void *this_fn,
                               void *call_site);

The first argument is the address of the start of the current function, which may be looked up exactly in the symbol table. 🔍

This instrumentation is also done for functions expanded inline in other functions. The profiling calls indicate where, conceptually, the inline function is entered and exited. This means that addressable versions of such functions must be available. If all your uses of a function are expanded inline, this may mean an additional expansion of code size 📈 If you use "extern inline" in your C code, an addressable version of such functions must be provided. (This is normally the case anyway, but if you get lucky and the optimizer always expands the functions inline, you might have gotten away without providing static copies.) lucky you, bestie 🍀

A function may be given the attribute "no_instrument_function", in which case this instrumentation is not done. This can be used, for example, for the profiling functions listed above, high-priority interrupt routines, and any functions from which the profiling functions cannot safely be called (perhaps signal handlers, if the profiling routines generate output or allocate memory). 🛡️

• -finstrument-functions-once — This is similar to -finstrument-functions, but the profiling functions are called only once per instrumented function, i.e. the first profiling function is called after the first entry into the instrumented function and the second profiling function is called before the exit corresponding to this first entry. 1️⃣

The definition of "once" for the purpose of this option is a little vague because the implementation is not protected against data races 😬 As a result, the implementation only guarantees that the profiling functions are called at least once per process and at most once per thread, but the calls are always paired, that is to say, if a thread calls the first function, then it will call the second function, unless it never reaches the exit of the instrumented function. pair them up, they're a set 💞

• -finstrument-functions-exclude-file-list=file,file,... — Set the list of functions that are excluded from instrumentation (see the description of -finstrument-functions). If the file that contains a function definition matches with one of file, then that function is not instrumented. The match is done on substrings: if the file parameter is a substring of the file name, it is considered to be a match. 🧩

For example:

-finstrument-functions-exclude-file-list=/bits/stl,include/sys

excludes any inline function defined in files whose pathnames contain /bits/stl or include/sys. ✂️

If, for some reason, you want to include letter , in one of sym, write ,. For example, -finstrument-functions-exclude-file-list=',,tmp' (note the single quote surrounding the option). the quoting is doing so much heavy lifting here, respect 🙌

• -finstrument-functions-exclude-function-list=sym,sym,... — This is similar to -finstrument-functions-exclude-file-list, but this option sets the list of function names to be excluded from instrumentation. The function name to be matched is its user-visible name, such as "vector<int> blah(const vector<int> &)", not the internal mangled name (e.g., "_Z4blahRSt6vectorIiSaIiEE"). user-visible name bestie, not the cursed mangled one 💀 The match is done on substrings: if the sym parameter is a substring of the function name, it is considered to be a match. For C99 and C++ extended identifiers, the function name must be given in UTF-8, not using universal character names. 🌈

---

• -fpatchable-function-entry=N[,M] — okay bestie, this one sprinkles N NOPs right at the start of every single function, and the function's entry point sits before the Mth NOP. Skip M entirely? It just defaults to 0, so the function entry points to the address right at the first NOP. Those NOP instructions are basically ✨reserved parking spots✨ — they hold extra space you can patch any instrumentation you want into at run time, as long as the code segment is writable. The amount of space is controlled indirectly through how many NOPs you ask for, and the actual NOP instruction used matches whatever the internal GCC back-end interface gen_nop emits. Heads up though: this behavior is target-specific and it may also depend on the architecture variant and/or other compilation options. 🦋

For run-time identification, the starting addresses of these areas — which correspond to their respective function entries minus M — get collected in the __patchable_function_entries section of the resulting binary. Cute, right?

Lowkey important, bestie: the value of __attribute__ ((patchable_function_entry (N,M))) takes precedence over the command-line option -fpatchable-function-entry=N,M. Use that to bulk up the area size or to remove it completely on a single function. And if N=0, no pad location is recorded at all.

The NOP instructions are inserted at — and maybe before, depending on M — the function entry address, even before the prologue. On PowerPC with the ELFv2 ABI, for a function with dual entry points, the local entry point is this function entry address by default. See the -msplit-patch-nops option if you wanna change that.

The max value of N and M is 65535. And on PowerPC with the ELFv2 ABI, for a function with dual entry points, the supported values for M are 0, 2, 6 and 14 when you're not using -msplit-patch-nops. 👑

---

🧚 Options Controlling the Preprocessor 🧚

These options are all about the C preprocessor, which gets run on every C source file before the actual compilation happens. It's the opening act before the main character (the compiler) shows up. 💅

If you use the -E option, literally nothing happens except preprocessing — that's it, that's the whole show. And some of these options only make sense together with -E, because they make the preprocessor output unsuitable for actual compilation. So like, vibe check your use case first. 🙈

On top of the options down below, there are also a bunch of options controlling search paths for include files — those live in Directory Options. And options that control preprocessor diagnostics are over in Warning Options.

• -D name — predefine name as a macro, with definition 1.
• -D name=definition — the contents of definition get tokenized and processed exactly as if they'd shown up during translation phase three in a #define directive. In particular, the definition is truncated by embedded newline characters (sneaky little things).

If you're invoking the preprocessor from a shell or shell-like program, you may need to use the shell's quoting syntax to protect characters like spaces that mean something to the shell syntax.

Wanna define a function-like macro on the command line? Write its argument list with surrounding parentheses before the equals sign (if any). Parentheses are meaningful to most shells, so quote the option. With sh and csh, -D'name(args...)=definition' works. ✨

Periodt: -D and -U options are processed in the order they're given on the command line. All -imacros file and -include file options are processed after all -D and -U options.

• -U name — cancel any previous definition of name, whether it was built in or handed to you via a -D option. bye felicia 💅
• -include file — process file as if #include "file" appeared as the very first line of the primary source file. Main character energy, first line of the show. But plot twist: the first directory searched for file is the preprocessor's working directory instead of the directory containing the main source file. If it's not found there, it gets searched for in the remainder of the #include "..." search chain as normal.

Got multiple -include options? The files are included in the order they appear on the command line. 🎀

• -imacros file — exactly like -include, except any output produced by scanning file is thrown away. Macros it defines stay defined though. This lets you grab all the macros from a header without also processing its declarations — very "just here for the vibes, not the drama".

All files specified by -imacros are processed before all files specified by -include. ⭐

• -undef — do not predefine any system-specific or GCC-specific macros. The standard predefined macros remain defined, so, like, not a total blank slate.
• -pthread — define the additional macros required for using the POSIX threads library. Use this option consistently for both compilation and linking — consistency is key, bestie, we don't do half-measures. This option is supported on GNU/Linux targets, most other Unix derivatives, and also on x86 Cygwin and MinGW targets. 🌈
• -M — instead of outputting the result of preprocessing, output a rule suitable for make describing the dependencies of the main source file. The preprocessor outputs one make rule containing the object file name for that source file, a colon, and the names of all the included files — including those coming from -include or -imacros command-line options. The receipts, if you will. 🧾

Unless specified explicitly (with -MT or -MQ), the object file name consists of the name of the source file with any suffix replaced with the object file suffix, and with any leading directory parts removed. If there are many included files, the rule gets split into several lines using \-newline. The rule has no commands.

This option does not suppress the preprocessor's debug output, such as -dM. To avoid mixing that debug output up with the dependency rules, you should explicitly specify the dependency output file with -MF, or use an environment variable like DEPENDENCIES_OUTPUT. Debug output still gets sent to the regular output stream as normal.

Passing -M to the driver implies -E, and suppresses warnings with an implicit -w. Sneaky!

• -MM — like -M, but it does not mention header files found in system header directories, nor header files included, directly or indirectly, from such a header.

This implies that the choice of angle brackets or double quotes in an #include directive does not in itself determine whether that header appears in -MM dependency output. It's giving "don't judge a header by its brackets". 🙈

• -MF file — when used with -M or -MM, specifies a file to write the dependencies to. If no -MF switch is given, the preprocessor sends the rules to the same place it would send preprocessed output.

When used with the driver options -MD or -MMD, -MF overrides the default dependency output file.

If file is -, then the dependencies are written to stdout. 🫶

• -MG — used alongside an option like -M that requests dependency generation, -MG assumes missing header files are generated files and adds them to the dependency list without raising an error. The dependency filename is taken directly from the #include directive without prepending any path. -MG also suppresses preprocessed output, since a missing header file renders that useless. No point, no service. 💀

This feature is used in automatic updating of makefiles.

• -Mno-modules — disable dependency generation for compiled module interfaces. That's it, that's the whole thing.
• -MP — this instructs CPP to add a phony target for each dependency other than the main file, making each one depend on nothing. These dummy rules work around the errors make throws at you if you remove header files without updating the Makefile to match. Preventative queen behavior. 👑

This is typical output:

test.o: test.c test.h

test.h:

• -MT target — change the target of the rule emitted by dependency generation. By default, CPP takes the name of the main input file, deletes any directory components and any file suffix such as .c, and appends the platform's usual object suffix. That result is the target.

An -MT option sets the target to be exactly the string you specify. Want multiple targets? Specify them as a single argument to -MT, or use multiple -MT options. Dealer's choice. ✨

For example, -MT '$(objpfx)foo.o' might give:

$(objpfx)foo.o: foo.c

• -MQ target — same as -MT, but it quotes any characters which are special to Make. So -MQ '$(objpfx)foo.o' gives:

$$(objpfx)foo.o: foo.c

The default target is automatically quoted, as if it were given with -MQ. It came pre-quoted, no notes.

• -MD — -MD is equivalent to -M -MF file, except that -E is not implied. The driver determines file based on whether an -o option is given. If it is, the driver uses its argument but with a suffix of .d. Otherwise it takes the name of the input file, removes any directory components and suffix, and applies a .d suffix. 🍓

If -MD is used in conjunction with -E, any -o switch is understood to specify the dependency output file — but if used without -E, each -o is understood to specify a target object file. Context matters, bestie.

Since -E is not implied, -MD can be used to generate a dependency output file as a side effect of the compilation process. Sneaky productive.

• -MMD — like -MD except it mentions only user header files, not system header files. Stay in your lane, /usr/include. 💅
• -fpreprocessed — tells the preprocessor that the input file has already been preprocessed. This suppresses things like macro expansion, trigraph conversion, escaped newline splicing, and processing of most directives. The preprocessor still recognizes and removes comments though — so you can pass it a file preprocessed with -C without any problems. In this mode the integrated preprocessor is little more than a tokenizer for the front ends. Just here to tokenize and look cute. 🎀

-fpreprocessed is implicit if the input file has one of the extensions .i, .ii or .mi. Those are the extensions GCC uses for preprocessed files created by -save-temps.

• -fdirectives-only — when preprocessing, handle directives, but do not expand macros. The option's behavior depends on the -E and -fpreprocessed options, so let's break it down:
  • With -E, preprocessing is limited to the handling of directives such as #define, #ifdef, and #error. Other preprocessor operations, such as macro expansion and trigraph conversion, are not performed. In addition, the -dD option is implicitly enabled. ✨
  • With -fpreprocessed, predefinition of command line and most builtin macros is disabled. Macros such as __LINE__, which are contextually dependent, are handled normally. This enables compilation of files previously preprocessed with -E -fdirectives-only.
  • With both -E and -fpreprocessed, the rules for -fpreprocessed take precedence. This enables full preprocessing of files previously preprocessed with -E -fdirectives-only. Full circle, iconic. 🌸
• -fdollars-in-identifiers — so bestie, this one says "yes, $ can live inside identifiers." Let the dollar signs in. 💸
• -fextended-identifiers — accepts universal character names and extended characters inside identifiers. And heads up: this one's on by default for C99 (and all the later C standard versions) plus C++ — so you're probably already living this life and didn't even know. ✨
• -fno-canonical-system-headers — when preprocessing, don't shorten system header paths with canonicalization. That's it, that's the whole vibe. 🙈
• -fmax-include-depth=depth — sets the maximum depth of nested #include. The default is 200, so try not to stack your includes into the actual stratosphere. 🌀
• -fsearch-include-path[=kind] — tells GCC to look for input files on the #include path, not just the current directory. This is especially the move with C++20 modules, where both header units and module interface units need to be compiled directly. Like this:

g++ -c -std=c++20 -fmodules -fsearch-include-path bits/stdc++.h bits/std.cc

And the tea on kind: it defaults to user, which looks on the #include "..." search path. You can also explicitly pass system to look on the "#include <...>" search path instead. Both valid, both slay. 🫶

• -ftabstop=width — sets the distance between tab stops. Why do you care? Because it helps the preprocessor report the correct column numbers in warnings or errors even when tabs are sitting on the line like "hi I'm here to ruin your column math." If the value is less than 1 or greater than 100, the option gets straight-up ignored (no warning, no drama, just ignored). The default is 8. 📏
• -ftrack-macro-expansion[=level] — tracks the locations of tokens across macro expansions. The glow-up here: when a compilation error happens inside a macro expansion, the compiler can actually emit a diagnostic about the current macro expansion stack instead of just pointing at a mystery. 😭 Downside? Using this option makes the preprocessor and the compiler eat more memory. And the level parameter is how you dial the precision of token location tracking up or down to save memory if you need to:
  • 0 — de-activates this option entirely. bye.
  • 1 — tracks token locations in a degraded mode for minimal memory overhead. In this mode, all tokens resulting from the expansion of an argument of a function-like macro have the same location.
  • 2 — tracks token locations completely. This value is the most memory hungry. 🍽️

When you give this option with no argument, the default parameter value is 2.

Heads up bestie: -ftrack-macro-expansion=2 is activated by default. So you're already in the fancy mode unless you say otherwise.

• -fmacro-prefix-map=old=new — when preprocessing files that live in directory old, expand the __FILE__ and __BASE_FILE__ macros as if the files lived in directory new instead. Big brain use case: change an absolute path into a relative one by setting new to . — which gives you more reproducible builds that don't care where you built them. This option also affects __builtin_FILE() during compilation. See also -ffile-prefix-map and -fcanon-prefix-map. 🔁
• -fexec-charset=charset — sets the execution character set, used for string and character constants. The default is UTF-8. And charset can be any encoding supported by the system's iconv library routine. 🌈
• -fwide-exec-charset=charset — sets the wide execution character set, used for wide string and character constants. The default is one of UTF-32BE, UTF-32LE, UTF-16BE, or UTF-16LE — whichever one corresponds to the width of wchar_t and the big-endian or little-endian byte order being used for code generation. Same deal as -fexec-charset: charset can be any encoding supported by the system's iconv library routine. BUT — and this is the disclaimer, not me being dramatic — you will have problems with encodings that don't fit exactly in wchar_t. 💀
• -finput-charset=charset — sets the input character set, used to translate from the character set of the input file into the source character set GCC uses. If the locale doesn't specify it, or GCC just can't get that info from the locale, the default is UTF-8. This can be overridden by either the locale or by this command-line option — and currently the command-line option wins if there's a conflict. charset can be any encoding supported by the system's iconv library routine. 👑
• -fpch-deps — when you're using precompiled headers, this flag makes the dependency-output flags also list the files from the precompiled header's dependencies. If you don't specify it, only the precompiled header itself gets listed — not the files that were used to create it — because those files aren't consulted at all when a precompiled header is used. The way that makes sense though. 🎀
• -fpch-preprocess — this option lets you use a precompiled header together with -E. It inserts a special #pragma, specifically #pragma GCC pch_preprocess "filename", into the output to mark where the precompiled header was found, plus its filename. And when -fpreprocessed is in use, GCC recognizes this #pragma and loads the PCH.

This option is off by default — why? Because the resulting preprocessed output is only really suitable as input to GCC. It gets switched on by -save-temps.

You should not write this #pragma in your own code. However, it is safe to edit the filename if the PCH file is available in a different location. The filename may be absolute, or it may be relative to GCC's current directory. 🗂️

• -fworking-directory — enables generation of linemarkers in the preprocessor output that let the compiler know the current working directory at the time of preprocessing. When this is enabled, the preprocessor emits — right after the initial linemarker — a second linemarker with the current working directory followed by two slashes. GCC then uses that directory, when it's present in the preprocessed input, as the directory emitted as the current working directory in some debugging information formats. This option is implicitly enabled if debugging information is enabled, but you can inhibit it with the negated form -fno-working-directory. And if the -P flag is present on the command line, this option has no effect, since no #line directives are emitted whatsoever. 🌸
• -A predicate=answer — makes an assertion with the predicate predicate and answer answer. This form is preferred over the older form -A predicate(answer), which is still supported, because it doesn't use shell special characters. (Shell special characters are the enemy, fr fr.) 🐚
• -A -predicate=answer — cancels an assertion with the predicate predicate and answer answer. Undoing the thing you did. periodt. ✂️
• -C — do not discard comments. All comments get passed through to the output file, except for comments in processed directives — those get deleted along with the directive.

Be prepared for side effects when you use -C, bestie, because it makes the preprocessor treat comments as tokens in their own right. For example: comments appearing at the start of what would be a directive line have the effect of turning that line into an ordinary source line, since the first token on the line is no longer a #. 💅

• -CC — do not discard comments, including during macro expansion. This is like -C, except that comments contained within macros are also passed through to the output file wherever the macro gets expanded. 😌

On top of all the side effects from -C, the -CC option also causes all C++-style comments inside a macro to be converted into C-style comments. Why? To prevent a later use of that macro from inadvertently commenting out the rest of the source line. Thank you for your service. 🫡

The -CC option is generally used to support lint comments.

• -P — inhibits generation of linemarkers in the output from the preprocessor. This might be useful when you're running the preprocessor on something that isn't C code and it's headed to a program that would be confused by linemarkers. Confusion prevention. 🧚
• -traditional
• -traditional-cpp — try to imitate the behavior of pre-standard C preprocessors, as opposed to ISO C preprocessors. See the GNU CPP manual for the details.

Note that GCC does not otherwise attempt to emulate a pre-standard C compiler, and these options are only supported with the -E switch, or when invoking CPP explicitly. Like, only then. No cap. 🦖

• -trigraphs — support ISO C trigraphs. These are three-character sequences, all starting with ??, that ISO C defines to stand for single characters. For example, ??/ stands for \, so '??/n' is a character constant for a newline.

The nine trigraphs and their replacements are:

        Trigraph:       ??(  ??)  ??<  ??>  ??=  ??/  ??'  ??!  ??-
        Replacement:      [    ]    {    }    #    \    ^    |    ~

By default, GCC ignores trigraphs — but in standard-conforming modes it converts them. See the -std and -ansi options. 🔮

• -remap — enables special code to work around file systems which only permit very short file names, such as MS-DOS. A true relic, but iconic in its own way. 📼
• -H — prints the name of each header file used, on top of everything else it's normally doing. Each name is indented to show how deep in the #include stack it is. Precompiled header files also get printed, even if they are found to be invalid — an invalid precompiled header file is printed with ...x and a valid one with ...!. Keep an eye on those little symbols, they're telling you things. 👀
• -dletters — tells GCC to make debugging dumps during compilation as specified by letters. The flags documented here are the ones relevant to the preprocessor. Other letters are interpreted by the compiler proper, or reserved for future versions of GCC, and so are silently ignored. And if you specify letters whose behavior conflicts? The result is undefined. Don't do that. 🙅
  • -dM — instead of the normal output, generate a list of #define directives for all the macros defined during the execution of the preprocessor, including predefined macros. This gives you a way of finding out what's predefined in your version of the preprocessor. Assuming you have no file foo.h, the command:

touch foo.h; cpp -dM foo.h

shows all the predefined macros.

If you use -dM without the -E option, -dM is interpreted as a synonym for -fdump-rtl-mach. (Plot twist.) 🎭

• -dD — like -dM except that it outputs both the #define directives and the result of preprocessing. Both kinds of output go to the standard output file.
• -dN — like -dD, but emit only the macro names, not their expansions. Names only, mystery preserved. ✨
• -dI — output #include directives in addition to the result of preprocessing.
• -dU — like -dD except that only macros that are expanded, or whose definedness is tested in preprocessor directives, are output; the output is delayed until the use or test of the macro; and #undef directives are also output for macros tested but undefined at the time. Tracking every little move. 🕵️

---

• -fdebug-cpp — okay so this one is literally only for debugging GCC itself, bestie 💅 when you use it from CPP or alongside -E, it dumps debugging info about location maps. Like, every single token in the output gets preceded by the dump of the map its location belongs to. That's it, that's the whole vibe.

And when you use it from GCC without -E? Absolutely nothing happens. No effect. Ghosted. 👻

• -Wp,option — the chaotic shortcut: you can use -Wp,option to straight-up bypass the compiler driver and shove option directly through to the preprocessor. If option has commas in it, it gets split into multiple options at those commas, easy.

BUT. And this is a big BUT 🚩 — so many options get modified, translated or interpreted by the compiler driver before they're passed to the preprocessor, and -Wp forcibly skips that whole phase. The preprocessor's direct interface is undocumented and subject to change, so whenever you possibly can, avoid -Wp and just let the driver handle the options instead. Let the driver cook. 👑

• -Xpreprocessor option — pass option as an option to the preprocessor. Perfect for when you need to supply system-specific preprocessor options that GCC doesn't recognize. it's giving niche but valid ✨

⚠️ If you want to pass an option that takes an argument, you must use -Xpreprocessor twice — once for the option and once for the argument. Two separate flags, no shortcuts.

• -no-integrated-cpp — do the preprocessing as a separate pass before compilation. By default, GCC does preprocessing as an integrated part of input tokenization and parsing (one big slay). But if you provide this option, the appropriate language front end — cc1, cc1plus, or cc1obj for C, C++, and Objective-C, respectively — gets invoked twice instead: once for preprocessing only, and once for actual compilation of the preprocessed input.

This one's useful in conjunction with the -B or -wrapper options to specify an alternate preprocessor or to do extra processing of the program source between normal preprocessing and compilation. very DIY, very custom 🎀

---

🔧 Passing Options to the Assembler

You can pass options to the assembler. Yes queen, the assembler deserves attention too 💜

• -Wa,option — pass option as an option to the assembler. If option contains commas, it gets split into multiple options at the commas. periodt.
• -Xassembler option — pass option as an option to the assembler. Use this to supply system-specific assembler options that GCC does not recognize, iykyk.

⚠️ Same deal as before: if you want to pass an option that takes an argument, you must use -Xassembler twice, once for the option and once for the argument.

---

🔗 Options for Linking

These options come into play when the compiler links object files into an executable output file. They are meaningless if the compiler isn't doing a link step — like, they simply don't apply, no cap.

• object-file-name — a file name that does not end in a special recognized suffix is considered to name an object file or library. (Object files get distinguished from libraries by the linker according to the file contents, btw, not the name.) If linking is done, these object files are used as input to the linker.
• -c, -S, -E — if any of these options is used, then the linker is not run, and object file names should not be used as arguments. simple as that 🌸
• -flinker-output=type — this baby controls code generation of the link-time optimizer. By default the linker output is automatically determined by the linker plugin, but for debugging the compiler — or when you want incremental linking with a non-LTO object file — it can be useful to control the type manually. Let's break down every type 🫶
  • If type is exec, code generation produces a static binary. In this case -fpic and -fpie are both disabled.
  • If type is dyn, code generation produces a shared library. In this case -fpic or -fPIC is preserved, but not enabled automatically. This allows building shared libraries without position-independent code on architectures where that's possible, i.e. on x86.
  • If type is pie, code generation produces an -fpie executable. This results in similar optimizations as exec except that -fpie is not disabled if specified at compilation time.
  • If type is rel, the compiler assumes incremental linking is being done. The sections containing intermediate code for link-time optimization are merged, pre-optimized, and output to the resulting object file. On top of that, if -ffat-lto-objects is specified, binary code is produced for future non-LTO linking. The object file produced by incremental linking is smaller than a static library produced from the same object files. And at link time, the result of incremental linking also loads faster than a static library — assuming that the majority of objects in the library are used. obsessed with the efficiency honestly ⚡
  • Finally, nolto-rel configures the compiler for incremental linking where code generation is forced, a final binary is produced, and the intermediate code for later link-time optimization is stripped. When multiple object files are linked together the resulting code is better optimized than with link-time optimizations disabled (for example, cross-module inlining happens, cute), but most of the benefits of whole program optimizations are lost. 😭 a tradeoff bestie.

During the incremental link (by -r) the linker plugin defaults to rel. With current interfaces to GNU Binutils it is however not possible to incrementally link LTO objects and non-LTO objects into a single mixed object file. If any of the object files in an incremental link cannot be used for link-time optimization, the linker plugin issues a warning and uses nolto-rel. To maintain whole program optimization, it is recommended to link such objects into a static library instead. Alternatively it is possible to use H.J. Lu's binutils with support for mixed objects. the more you know 🌈

• -fuse-ld=bfd — use the bfd linker instead of the default linker.
• -fuse-ld=gold — use the gold linker instead of the default linker.
• -fuse-ld=lld — use the LLVM lld linker instead of the default linker.
• -fuse-ld=mold — use the Modern Linker (mold) instead of the default linker. modern queen 💅
• -llibrary, -l library — search the library named library when linking. (The second alternative, with the library as a separate argument, is only for POSIX compliance and is not recommended. fr fr, just don't.)

The -l option is passed directly to the linker by GCC. Refer to your linker documentation for exact details — the general description below applies to the GNU linker.

The linker searches a standard list of directories for the library. Those directories include several standard system directories plus any that you specify with -L.

Static libraries are archives of object files, and have file names like liblibrary.a. Some targets also support shared libraries, which typically have names like liblibrary.so. If both static and shared libraries are found, the linker gives preference to linking with the shared library unless the -static option is used.

⚠️ It makes a difference where in the command you write this option — the linker searches and processes libraries and object files in the order they are specified. So foo.o -lz bar.o searches library z after file foo.o but before bar.o. If bar.o refers to functions in z, those functions may not be loaded. order matters, bestie, it's giving strict 💀

• -lobjc — you need this special case of the -l option in order to link an Objective-C or Objective-C++ program.
• -nostartfiles — do not use the standard system startup files when linking. The standard system libraries are used normally, unless -nostdlib, -nolibc, or -nodefaultlibs is used.
• -nodefaultlibs — do not use the standard system libraries when linking. Only the libraries you specify are passed to the linker, and options specifying linkage of the system libraries — such as -static-libgcc or -shared-libgcc — are ignored. The standard startup files are used normally, unless -nostartfiles is used.

The compiler may generate calls to memcmp, memset, memcpy and memmove. These entries are usually resolved by entries in libc. These entry points should be supplied through some other mechanism when this option is specified. 👀 your problem now.

• -nolibc — do not use the C library or system libraries tightly coupled with it when linking. It still links with the startup files, libgcc, or toolchain-provided language support libraries such as libgnat, libgfortran or libstdc++ — unless options preventing their inclusion are used as well. This typically removes -lc from the link command line, as well as the system libraries that normally go with it and become meaningless when the absence of a C library is assumed — for example -lpthread or -lm in some configurations. This is intended for bare-board targets when there is indeed no C library available. bare metal era 🦋
• -nostdlib — do not use the standard system startup files or libraries when linking. No startup files, and only the libraries you specify are passed to the linker, and options specifying linkage of the system libraries — such as -static-libgcc or -shared-libgcc — are ignored.

The compiler may generate calls to memcmp, memset, memcpy and memmove. These entries are usually resolved by entries in libc. These entry points should be supplied through some other mechanism when this option is specified.

One of the standard libraries bypassed by -nostdlib and -nodefaultlibs is libgcc.a, a library of internal subroutines which GCC uses to overcome shortcomings of particular machines, or special needs for some languages.

In most cases, you need libgcc.a even when you want to avoid other standard libraries. In other words, when you specify -nostdlib or -nodefaultlibs you should usually specify -lgcc as well. This ensures you have no unresolved references to internal GCC library subroutines. (An example of such an internal subroutine is __main, used to ensure C++ constructors are called.) don't forget -lgcc, bestie, that's the whole point 🙈

• -nostdlib++ — do not implicitly link with standard C++ libraries.
• -e entry, --entry=entry — specify that the program entry point is entry. The argument is interpreted by the linker; the GNU linker accepts either a symbol name or an address. very flexible of you ✨
• -pie — produce a dynamically linked position independent executable on targets that support it. For predictable results, you must also specify the same set of options used for compilation (-fpie, -fPIE, or model suboptions) when you specify this linker option. consistency is key 🔑
• -no-pie — don't produce a dynamically linked position independent executable.
• -static-pie — makes a static position independent executable on targets that support it 🫶. Think of it as a static executable's cooler cousin: same energy, but it can get loaded at literally any address without needing a dynamic linker. 💅 For predictable results, you also have to pass the same set of options you used for compilation (-fpie, -fPIE, or model suboptions) when you use this linker option. No mix-and-match bestie, the vibes have to match.
• -pthread — links against the POSIX threads library 🧵. It's supported on GNU/Linux targets, most other Unix derivatives, and also on x86 Cygwin and MinGW targets. Heads up: on some targets this option also sets flags for the preprocessor, so you gotta use it consistently for both compilation and linking. Consistency is the whole thing, fr fr.
• -r — produce a relocatable object as output. It's also known as partial linking. 👑
• -rdynamic — passes the flag -export-dynamic to the ELF linker, on targets that support it. This tells the linker to add all symbols, not just the used ones, to the dynamic symbol table ✨. You need this for some uses of dlopen, or if you wanna pull backtraces from inside a program. Detective mode 🔍.
• -s — removes all symbol table and relocation information from the executable 💀. Poof, gone, no trace.
• -static — on systems that support dynamic linking, this overrides -pie and stops it from linking with the shared libraries. On other systems, this option does nothing at all. 🙈 Nothing. Zero. Nada.
• -shared — produce a shared object, which can then be linked with other objects to form an executable 🌈. Not all systems support this option. For predictable results, you must also specify the same set of options used for compilation (-fpic, -fPIC, or model suboptions) when you specify this linker option.[1]
• -shared-libgcc and -static-libgcc — on systems that provide libgcc as a shared library, these options force the use of either the shared or static version, respectively 🎀. And if no shared version of libgcc was built when the compiler was configured, these options just have no effect.

There are several situations where an app should use the shared libgcc instead of the static version. The most common one is when the application wants to throw and catch exceptions across different shared libraries. In that case, each of the libraries and the application itself should use the shared libgcc. 🫶

That's why the G++ driver automatically adds -shared-libgcc whenever you build a shared library or a main executable — because C++ programs typically use exceptions, so this is the right thing to do. Iconic behavior, honestly.

If, instead, you use the GCC driver to create shared libraries, you may find they're not always linked with the shared libgcc 😭. If GCC finds, at its configuration time, that you have a non-GNU linker or a GNU linker that doesn't support option --eh-frame-hdr, it links the shared version of libgcc into shared libraries by default. Otherwise, it takes advantage of the linker and optimizes away the linking with the shared version of libgcc, linking with the static version of libgcc by default. This lets exceptions propagate through those shared libraries, without incurring relocation costs at library load time. ✨

However — and this is the important part — if a library or main executable is supposed to throw or catch exceptions, you must link it using the G++ driver, or using the option -shared-libgcc, so that it's linked with the shared libgcc. No shortcuts bestie.

• -static-libasan — when the -fsanitize=address option is used to link a program, the GCC driver automatically links against libasan. If libasan is available as a shared library, and the -static option isn't used, then this links against the shared version of libasan. The -static-libasan option tells the GCC driver to link libasan statically, without necessarily linking other libraries statically. 💅
• -static-libtsan — when the -fsanitize=thread option is used to link a program, the GCC driver automatically links against libtsan. If libtsan is available as a shared library, and the -static option isn't used, then this links against the shared version of libtsan. The -static-libtsan option tells the GCC driver to link libtsan statically, without necessarily linking other libraries statically. 🧵
• -static-liblsan — when the -fsanitize=leak option is used to link a program, the GCC driver automatically links against liblsan. If liblsan is available as a shared library, and the -static option isn't used, then this links against the shared version of liblsan. The -static-liblsan option tells the GCC driver to link liblsan statically, without necessarily linking other libraries statically. 💧
• -static-libubsan — when the -fsanitize=undefined option is used to link a program, the GCC driver automatically links against libubsan. If libubsan is available as a shared library, and the -static option isn't used, then this links against the shared version of libubsan. The -static-libubsan option tells the GCC driver to link libubsan statically, without necessarily linking other libraries statically. 🌸
• -static-libstdc++ — when the g++ program is used to link a C++ program, it normally automatically links against libstdc++. If libstdc++ is available as a shared library, and the -static option isn't used, then this links against the shared version of libstdc++. And that's normally totally fine. 😌 But sometimes it's useful to freeze the version of libstdc++ that the program uses without committing all the way to a fully static link. The -static-libstdc++ option tells the g++ driver to link libstdc++ statically, without necessarily linking other libraries statically. Obsessed with the flexibility tbh.
• -symbolic — bind references to global symbols when building a shared object. It warns about any unresolved references (unless you override that with the link editor option -Xlinker -z -Xlinker defs). Only a few systems support this option. Rare behavior ✨
• -T script — use script as the linker script 📜. This option is supported by most systems using the GNU linker. On some targets — like bare-board targets without an operating system — the -T option may be required when linking, to avoid references to undefined symbols. So like, not optional there, bestie.
• -Xlinker option — pass option as an option to the linker. You can use this to supply system-specific linker options that GCC doesn't recognize. 🎀

If you want to pass an option that takes a separate argument, you have to use -Xlinker twice — once for the option, once for the argument. For example, to pass -assert definitions, you must write:

-Xlinker -assert -Xlinker definitions

It does not work to write -Xlinker "-assert definitions", because that passes the entire string as a single argument, which is not what the linker expects. 🫠 Classic trap.

When using the GNU linker, it's usually more convenient to pass arguments to linker options using the option=value syntax than as separate arguments. For example, you can specify -Xlinker -Map=output.map rather than -Xlinker -Map -Xlinker output.map. Other linkers may not support this syntax for command-line options. iykyk 💅

• -Wl,option — pass option as an option to the linker. If option contains commas, it gets split into multiple options at the commas. You can use this syntax to pass an argument to the option. For example, -Wl,-Map,output.map passes -Map output.map to the linker. When using the GNU linker, you can also get the same effect with -Wl,-Map=output.map. 🌈

NOTE: In Ubuntu 8.10 and later versions, for LDFLAGS, the option -Wl,-z,relro is used. To disable, use -Wl,-z,norelro. 🔥

• -u symbol — pretend the symbol symbol is undefined, to force linking of library modules to define it. You can use -u multiple times with different symbols to force loading of additional library modules. 👑
• -z keyword — -z is passed directly on to the linker along with the keyword keyword. See the section in the documentation of your linker for permitted values and their meanings. 📖

---

✨ Options for Directory Search ✨

These options specify directories to search for header files, for libraries, and for parts of the compiler. Directory hunt era 🦋

• -I dir
• -iquote dir
• -isystem dir
• -idirafter dir

Add the directory dir to the list of directories to be searched for header files during preprocessing 🎀. If dir begins with = or $SYSROOT, then the = or $SYSROOT gets replaced by the sysroot prefix; see --sysroot and -isysroot.

Directories specified with -iquote apply only to the quote form of the directive, #include "file". Directories specified with -I, -isystem, or -idirafter apply to lookup for both the #include "file" and #include <file> directives. 🫶

You can specify any number or combination of these options on the command line to search for header files in several directories. The lookup order is as follows:

1. For the quote form of the include directive, the directory of the current file is searched first.
2. For the quote form of the include directive, the directories specified by -iquote options are searched in left-to-right order, as they appear on the command line.
3. Directories specified with -I options are scanned in left-to-right order.
4. Directories specified with -isystem options are scanned in left-to-right order.
5. Standard system directories are scanned.
6. Directories specified with -idirafter options are scanned in left-to-right order.

You can use -I to override a system header file, substituting your own version, since these directories are searched before the standard system header file directories. However — and listen — you should not use this option to add directories that contain vendor-supplied system header files; use -isystem for that. 🙅

The -isystem and -idirafter options also mark the directory as a system directory, so it gets the same special treatment that's applied to the standard system directories. ✨

If a standard system include directory, or a directory specified with -isystem, is also specified with -I, the -I option is ignored 😭. The directory is still searched, but as a system directory at its normal position in the system include chain. This is to make sure GCC's procedure to fix buggy system headers, and the ordering for the #include_next directive, aren't inadvertently changed. If you really need to change the search order for system directories, use the -nostdinc and/or -isystem options instead. 💜

• -I- — bestie, this one is deprecated 💀 so pls use -iquote instead for the -I directories that come before the -I-, and just... yeet the -I- option entirely, okay? 🫶

Any directories you specified with -I options before -I- get searched only for headers requested with "#include "file"" — they are not searched for "#include <file>". But if you drop additional directories with -I options after the -I-, those get searched for all #include directives. Slay.

Also — and this is the sneaky part — -I- inhibits the use of the directory of the current file directory as the first search directory for "#include "file"". And there is no way to override this effect of -I-. Periodt. No takebacks. 🙈

• -iprefix prefix — set prefix as the prefix for all the -iwithprefix options that come after it. If your prefix is a directory, pls include the final / or it's gonna be giving chaos. 🎀
• -iwithprefix dir, -iwithprefixbefore dir — append dir to the prefix you set earlier with -iprefix, then add the resulting directory to the include search path. ✨ Heads up tho: -iwithprefixbefore puts it in the same place -I would, while -iwithprefix puts it where -idirafter would. Different vibes, same dir.
• -isysroot dir — this is like the --sysroot option but it applies only to header files (except for Darwin targets, where it applies to both header files and libraries, main character energy ✨). See the --sysroot option for more info, iykyk.
• -imultilib dir — use dir as a subdirectory of the directory containing target-specific C++ headers. Niche but iconic. 💅
• -nostdinc — do not search the standard system directories for header files. Only the directories you explicitly specified with -I, -iquote, -isystem, and/or -idirafter options (plus the directory of the current file, if appropriate) are searched. So if you didn't name it, it's not getting found, fr fr. 😭
• -nostdinc++ — do not search for header files in the C++-specific standard directories, but you do still search the other standard directories. Fun fact: this option is used when building the C++ library itself. 👑
• --embed-dir=dir — append the dir directory to the list of searched directories for the "#embed" preprocessing directive or the __has_embed macro. And bestie, note this: there are no default directories for "#embed". You gotta bring your own. 🧚

If dir begins with = or $SYSROOT, then the = or $SYSROOT gets replaced by the sysroot prefix; see --sysroot and -isysroot. 💖

• -iplugindir=dir — set the directory to search for plugins that are passed by -fplugin=name instead of -fplugin=path/name.so. Lowkey this option is not meant to be used by you, the user — it's only passed by the driver. So hands off, it's driver business. 🙈
• -Ldir — add the directory dir to the list of directories to be searched for -l. Classic, does what it says. ✨
• -Bprefix — this option specifies where to find the executables, libraries, include files, and data files of the compiler itself. It's kind of a big deal. 💅

The compiler driver program runs one or more of the subprograms cpp, cc1, as and ld. It tries prefix as a prefix for each program it tries to run, both with and without machine/version/ for the corresponding target machine and compiler version.

For each subprogram to be run, the compiler driver first tries the -B prefix, if any. If that name is not found, or if -B is not specified, the driver tries two standard prefixes, /usr/lib/gcc/ and /usr/local/lib/gcc/. If neither of those results in a file name that is found, the unmodified program name is searched for using the directories specified in your PATH environment variable. 🌈

The compiler checks to see if the path provided by -B refers to a directory, and if necessary it adds a directory separator character at the end of the path. Detail-oriented queen. 🎀

-B prefixes that effectively specify directory names also apply to libraries in the linker, because the compiler translates these options into -L options for the linker. They also apply to include files in the preprocessor, because the compiler translates these options into -isystem options for the preprocessor. In this case, the compiler appends include to the prefix. ✨

The runtime support file libgcc.a can also be searched for using the -B prefix, if needed. If it is not found there, the two standard prefixes above are tried, and that is all. The file is left out of the link if it is not found by those means. 😭

Another way to specify a prefix much like the -B prefix is to use the environment variable GCC_EXEC_PREFIX.

As a special kludge, if the path provided by -B is [dir/]stageN/, where N is a number in the range 0 to 9, then it is replaced by [dir/]include. This is to help with boot-strapping the compiler. Sneaky little shortcut, love that for us. 💜

• -no-canonical-prefixes — do not expand any symbolic links, resolve references to /../ or /./, or make the path absolute when generating a relative prefix. Keep it as-is, no glow-up. 🌸
• --sysroot=dir — use dir as the logical root directory for headers and libraries. For example, if the compiler normally searches for headers in /usr/include and libraries in /usr/lib, it instead searches dir/usr/include and dir/usr/lib. Rerooted, iconic. 👑

If you use both this option and the -isysroot option, then the --sysroot option applies to libraries, but the -isysroot option applies to header files. Division of labor, we stan. 💅

The GNU linker (beginning with version 2.16) has the necessary support for this option. If your linker does not support this option, the header file aspect of --sysroot still works, but the library aspect does not. So like... partial slay. 😭

• --no-sysroot-suffix — for some targets, a suffix is added to the root directory specified with --sysroot, depending on the other options used, so that headers may for example be found in dir/suffix/usr/include instead of dir/usr/include. This option disables the addition of such a suffix. No suffix, no problem. ✨

---

✨ Options for Code Generation Conventions ✨

These machine-independent options control the interface conventions used in code generation. Very architectural, very demure. 🎀

Most of them have both positive and negative forms; the negative form of -ffoo is -fno-foo. In the table below, only one of the forms is listed — the one that is not the default. You can figure out the other form by either removing no- or adding it. Cute, right? 🦋

• -fstack-reuse=reuse-level — this option controls stack space reuse for user declared local/auto variables and compiler generated temporaries. reuse_level can be all, named_vars, or none. all enables stack reuse for all local variables and temporaries, named_vars enables the reuse only for user defined local variables with names, and none disables stack reuse completely. The default value is all. 🔥

The option is needed when the program extends the lifetime of a scoped local variable or a compiler generated temporary beyond the end point defined by the language. When a lifetime of a variable ends, and if the variable lives in memory, the optimizing compiler has the freedom to reuse its stack space with other temporaries or scoped local variables whose live range does not overlap with it. Legacy code extending local lifetime is likely to break with the stack reuse optimization. ⚠️

For example,

int *p;
{
  int local1;

  p = &local1;
  local1 = 10;
  ....
}
{
   int local2;
   local2 = 20;
   ...
}

if (*p == 10)  // out of scope use of local1
  {

  }

Another example:

struct A
{
    A(int k) : i(k), j(k) { }
    int i;
    int j;
};

A *ap;

void foo(const A& ar)
{
   ap = &ar;
}

void bar()
{
   foo(A(10)); // temp object's lifetime ends when foo returns

   {
     A a(20);
     ....
   }
   ap->i+= 10;  // ap references out of scope temp whose space
                // is reused with a. What is the value of ap->i?
}

The lifetime of a compiler generated temporary is well defined by the C++ standard. When a lifetime of a temporary ends, and if the temporary lives in memory, the optimizing compiler has the freedom to reuse its stack space with other temporaries or scoped local variables whose live range does not overlap with it. However some of the legacy code relies on the behavior of older compilers in which temporaries' stack space is not reused, the aggressive stack reuse can lead to runtime errors. This option is used to control the temporary stack reuse optimization. No cap. 💀

• -ftrapv — this option generates traps for signed overflow on addition, subtraction, multiplication operations. The options -ftrapv and -fwrapv override each other, so using -ftrapv -fwrapv on the command-line results in -fwrapv being effective. Note that only active options override, so using -ftrapv -fwrapv -fno-wrapv on the command-line results in -ftrapv being effective. The math is mathing. ✨
• -fwrapv — this option instructs the compiler to assume that signed arithmetic overflow of addition, subtraction and multiplication wraps around using twos-complement representation. This flag enables some optimizations and disables others. The options -ftrapv and -fwrapv override each other, so using -ftrapv -fwrapv on the command-line results in -fwrapv being effective. Note that only active options override, so using -ftrapv -fwrapv -fno-wrapv on the command-line results in -ftrapv being effective. Wrap it up, bestie. 🎁
• -fwrapv-pointer — this option instructs the compiler to assume that pointer arithmetic overflow on addition and subtraction wraps around using twos-complement representation. This flag disables some optimizations which assume pointer overflow is invalid. Pointer glow-up denied, basically. 💅
• -fstrict-overflow — this option implies -fno-wrapv -fno-wrapv-pointer and when negated implies -fwrapv -fwrapv-pointer. The domino effect of flags, obsessed. 🔥
• -fexceptions — bestie, this one turns ON exception handling 🔥. It generates the extra code needed to propagate exceptions. On some targets that means GCC makes frame unwind info for every function, which is giving significant data-size overhead (execution is fine tho, she doesn't touch that). Don't specify it? Then GCC picks for you: it's ON by default for languages like C++ that normally need exception handling, and OFF for languages like C that normally don't. But lowkey you might need to flip it on when you're compiling C code that has to interoperate with exception handlers written in C++ 🫶. And you might wanna turn it off if you're compiling old C++ programs that don't use exception handling at all.
• -fnon-call-exceptions — generates code that lets trapping instructions throw exceptions ✨. Heads up: this needs platform-specific runtime support that doesn't exist everywhere, so it's giving ✨limited edition✨. Also it ONLY lets trapping instructions throw — like memory references or floating-point instructions. It does NOT let exceptions get thrown from arbitrary signal handlers such as "SIGALRM". This one enables -fexceptions.
• -fdelete-dead-exceptions — treats instructions that might throw exceptions but don't otherwise contribute to the program's execution as removable 🗑️. This does not affect calls to functions — except those with the "pure" or "const" attributes. Enabled by default for the Ada and C++ compilers, as permitted by the language specs. And the optimization passes that cause dead exceptions to be removed get enabled independently at different optimization levels.
• -funwind-tables — basically -fexceptions' chill sibling: it just generates whatever static data is needed, but does NOT affect the generated code in any other way 💅. You normally don't need to turn this on yourself; instead a language processor that needs this handling enables it on your behalf.
• -fasynchronous-unwind-tables — generates an unwind table in DWARF format, if the target machine supports it 🌈. The table is exact at each instruction boundary, so it can be used for stack unwinding from asynchronous events (like a debugger or a garbage collector 👀).
• -fno-gnu-unique — on systems with a recent GNU assembler and C library, the C++ compiler uses the "STB_GNU_UNIQUE" binding to make sure that definitions of template static data members and static local variables in inline functions stay unique even when "RTLD_LOCAL" is in the mix. Why do we care? Because it avoids problems where a library used by two different "RTLD_LOCAL" plugins depends on a definition in one of them and therefore ends up disagreeing with the other one about the binding of the symbol 👑. BUT — and this is the tea ☕ — it causes "dlclose" to be ignored for affected DSOs. So if your program relies on reinitialization of a DSO via "dlclose" and "dlopen", you can use -fno-gnu-unique.
• -fpcc-struct-return — returns "short" "struct" and "union" values in memory, like the longer ones, instead of in registers 💾. This convention is less efficient, but it has the advantage of allowing intercallability between GCC-compiled files and files compiled with other compilers — particularly the Portable C Compiler (pcc).

The precise convention for returning structures in memory depends on the target configuration macros.

Short structures and unions are those whose size and alignment match that of some integer type.

Warning: code compiled with the -fpcc-struct-return switch is not binary compatible with code compiled with the -freg-struct-return switch. Use it to conform to a non-default application binary interface.

• -freg-struct-return — returns "struct" and "union" values in registers when possible ✨. This is more efficient for small structures than -fpcc-struct-return.

If you specify neither -fpcc-struct-return nor -freg-struct-return, GCC defaults to whichever convention is standard for the target. If there's no standard convention, GCC defaults to -fpcc-struct-return — except on targets where GCC is the principal compiler. In those cases we get to choose the standard, and we chose the more efficient register return alternative 💅.

Warning: code compiled with the -freg-struct-return switch is not binary compatible with code compiled with the -fpcc-struct-return switch. Use it to conform to a non-default application binary interface.

• -fshort-enums — allocates to an "enum" type only as many bytes as it needs for the declared range of possible values 🎀. Specifically, the "enum" type becomes equivalent to the smallest integer type that has enough room. No effect for an enumeration type with a fixed underlying type, tho.

Warning: the -fshort-enums switch causes GCC to generate code that is not binary compatible with code generated without that switch. Use it to conform to a non-default application binary interface.

• -fshort-wchar — overrides the underlying type for "wchar_t" to be "short unsigned int" instead of the default for the target 🌸. Useful for building programs to run under WINE, iykyk.

Warning: the -fshort-wchar switch causes GCC to generate code that is not binary compatible with code generated without that switch. Use it to conform to a non-default application binary interface.

• -fcommon — in C code, this option controls the placement of global variables defined without an initializer, known as tentative definitions in the C standard 💀. Tentative definitions are distinct from declarations of a variable with the "extern" keyword, which do not allocate storage.

The default is -fno-common, which specifies that the compiler places uninitialized global variables in the BSS section of the object file. This inhibits the merging of tentative definitions by the linker — so you get a multiple-definition error if the same variable is accidentally defined in more than one compilation unit 👀.

The -fcommon places uninitialized global variables in a common block. This lets the linker resolve all tentative definitions of the same variable in different compilation units to the same object, or to a non-tentative definition ✨. This behavior is inconsistent with C++, and on many targets implies a speed and code size penalty on global variable references. It's mainly useful to let legacy code link without errors.

• -fno-ident — ignores the "#ident" directive 🙈.
• -finhibit-size-directive — doesn't output a ".size" assembler directive, or anything else that would cause trouble if the function is split in the middle and the two halves get placed at locations far apart in memory 😭. This option is used when compiling crtstuff.c; you should not need to use it for anything else.
• -fverbose-asm — puts extra commentary info in the generated assembly code to make it more readable 🎀. This option is generally only of use to those who actually need to read the generated assembly code (maybe while debugging the compiler itself).

-fno-verbose-asm, the default, causes the extra information to be omitted and is useful when comparing two assembler files.

The added comments include:

• information on the compiler version and command-line options,
• the source code lines associated with the assembly instructions, in the form FILENAME:LINENUMBER:CONTENT OF LINE,
• hints on which high-level expressions correspond to the various assembly instruction operands.

For example, given this C source file:

int test (int n)
{
  int i;
  int total = 0;

  for (i = 0; i < n; i++)
    total += i * i;

  return total;
}

compiling to (x86_64) assembly via -S and emitting the result direct to stdout via -o -

gcc -S test.c -fverbose-asm -Os -o -

gives output similar to this:

        .file   "test.c"
# GNU C11 (GCC) version 7.0.0 20160809 (experimental) (x86_64-pc-linux-gnu)
  [...snip...]
# options passed:
  [...snip...]

        .text
        .globl  test
        .type   test, @function
test:
.LFB0:
        .cfi_startproc
# test.c:4:   int total = 0;
        xorl    %eax, %eax      # <retval>
# test.c:6:   for (i = 0; i < n; i++)
        xorl    %edx, %edx      # i
.L2:
# test.c:6:   for (i = 0; i < n; i++)
        cmpl    %edi, %edx      # n, i
        jge     .L5     #,
# test.c:7:     total += i * i;
        movl    %edx, %ecx      # i, tmp92
        imull   %edx, %ecx      # i, tmp92
# test.c:6:   for (i = 0; i < n; i++)
        incl    %edx    # i
# test.c:7:     total += i * i;
        addl    %ecx, %eax      # tmp92, <retval>
        jmp     .L2     #
.L5:
# test.c:10: }
        ret
        .cfi_endproc
.LFE0:
        .size   test, .-test
        .ident  "GCC: (GNU) 7.0.0 20160809 (experimental)"
        .section        .note.GNU-stack,"",@progbits

The comments are intended for humans rather than machines, so the precise format of the comments is subject to change 💅.

• -frecord-gcc-switches — this switch causes the command line used to invoke the compiler to be recorded into the object file that's being created 📝. It's only implemented on some targets, and the exact format of the recording is target and binary file format dependent — but it usually takes the form of a section containing ASCII text ✨. This switch is related to the -fverbose-asm switch, but that one only records info in the assembler output file as comments, so it never reaches the object file. See also -grecord-gcc-switches for another way of storing compiler options into the object file 👑.

---

• -fpic — ok bestie, this one generates position-independent code (aka PIC) that's suitable for use in a shared library — if the target machine supports it. This kind of code reaches every constant address through a global offset table (GOT). The dynamic loader fills in the GOT entries when the program starts up (heads up: the dynamic loader is not part of GCC, it belongs to the operating system 🫶). If the GOT for your linked executable balloons past some machine-specific max size, the linker will yell at you saying -fpic doesn't work — and in that case you just recompile with -fPIC instead. Those maximums are 8k on the SPARC, 28k on AArch64, and 32k on the m68k and RS/6000. The x86 has no such limit, slay. ✨

Position-independent code needs special support, so it only works on certain machines. On x86, GCC supports PIC for System V but not for the Sun 386i. And code generated for the IBM RS/6000 is always position-independent, periodt.

When this flag is set, the macros "__pic__" and "__PIC__" get defined to 1.

• -fPIC — if the target machine supports it, this emits position-independent code suitable for dynamic linking and dodges any limit on the size of the global offset table. This option actually makes a difference on AArch64, m68k, PowerPC and SPARC.

Position-independent code needs special support, so it only works on certain machines.

When this flag is set, the macros "__pic__" and "__PIC__" get defined to 2.

• -fpie, -fPIE — these are like -fpic and -fPIC, but the position-independent code they generate can only be linked into executables. Usually you use them to compile code that's going to be linked using the -pie GCC option. 🎀

-fpie and -fPIE both define the macros "__pie__" and "__PIE__". Those macros get the value 1 for -fpie and 2 for -fPIE.

• -fno-plt — do not use the PLT for external function calls in position-independent code. Instead, it loads the callee address at the call sites straight from the GOT and branches to it. That's more efficient code because it eliminates PLT stubs and exposes the GOT loads to optimizations. On architectures like 32-bit x86, where PLT stubs expect the GOT pointer in a specific register, this hands the compiler more register allocation freedom. 💅 Lazy binding requires the PLT — so with -fno-plt, all external symbols get resolved at load time.

Alternatively, the function attribute "noplt" can be used to avoid calls through the PLT for specific external functions.

In position-dependent code, a few targets also convert calls to functions that are marked to not use the PLT into using the GOT instead.

• -fno-jump-tables — do not use jump tables for switch statements, even when that'd be more efficient than other code generation strategies. This one is useful alongside -fpic or -fPIC when you're building code that forms part of a dynamic linker and can't reference the address of a jump table. On some targets, jump tables don't require a GOT, so you don't need this option at all.
• -fno-bit-tests — do not use bit tests for switch statements, even when that'd be more efficient than other code generation strategies.
• -ffixed-reg — treat the register named reg as a fixed register, meaning generated code should never refer to it (except maybe as a stack pointer, frame pointer, or in some other fixed role). 🫡

reg has to be the name of a register. The register names accepted are machine-specific and are defined in the "REGISTER_NAMES" macro inside the machine description macro file.

This flag has no negative form, because what it's actually specifying is a three-way choice.

• -fcall-used-reg — treat the register named reg as an allocable register that gets clobbered by function calls. It may be allocated for temporaries or variables that don't live across a call. Functions compiled this way do not save and restore the register reg.

It's an error to use this flag with the frame pointer or stack pointer. And using this flag for other registers that have fixed pervasive roles in the machine's execution model produces disastrous results 😭 — like, genuinely, do not.

This flag has no negative form, because it specifies a three-way choice.

• -fcall-saved-reg — treat the register named reg as an allocable register saved by functions. It may be allocated even for temporaries or variables that live across a call. Functions compiled this way save and restore the register reg if they use it.

It's an error to use this flag with the frame pointer or stack pointer. And using it for other registers with fixed pervasive roles in the machine's execution model produces disastrous results.

A different sort of disaster shows up if you use this flag on a register where function values may be returned. 💀

This flag has no negative form, because it specifies a three-way choice.

• -fpack-struct[=n] — with no value specified, it packs all structure members together with no holes. When you do give a value (which has to be a small power of two), it packs structure members according to that value, representing the max alignment — meaning objects with default alignment requirements larger than this get output potentially unaligned at the next fitting location.

Warning bestie: the -fpack-struct switch makes GCC generate code that is not binary compatible with code generated without that switch. On top of that, it makes the code suboptimal. Use it to conform to a non-default application binary interface.

• -fleading-underscore — this option and its counterpart, -fno-leading-underscore, forcibly change the way C symbols are represented in the object file. One use is helping you link with legacy assembly code. 🌸

Warning: the -fleading-underscore switch makes GCC generate code that is not binary compatible with code generated without that switch. Use it to conform to a non-default application binary interface. Not all targets provide complete support for this switch.

• -ftls-model=model — change up the thread-local storage model in use. The model argument should be one of global-dynamic, local-dynamic, initial-exec or local-exec. Ngl though, the choice is subject to optimization: the compiler may use a more efficient model for symbols that aren't visible outside of the translation unit, or if -fpic isn't given on the command line.

The default without -fpic is initial-exec; with -fpic the default is global-dynamic.

• -ftrampolines — for targets that normally need trampolines for nested functions, always generate them instead of using descriptors. Otherwise, for targets that don't need them — like, for example, HP-PA or IA-64 — do nothing.

A trampoline is a small piece of code created at run time on the stack when the address of a nested function is taken, and it's used to call the nested function indirectly. So yeah, it requires the stack to be made executable for your program to work properly. 🧚

-fno-trampolines is enabled by default on a language-by-language basis, letting the compiler avoid generating them — if it computes that this is safe — and replace them with descriptors. Descriptors are made up of data only, but the generated code has to be prepared to deal with them. As of this writing, -fno-trampolines is enabled by default only for Ada.

Also important, and I mean important: code compiled with -ftrampolines and code compiled with -fno-trampolines are not binary compatible if nested functions are present. So this option absolutely must be used on a program-wide basis and be manipulated with extreme care. 🙈

For languages other than Ada, the "-ftrampolines" and "-fno-trampolines" options currently have no effect, and trampolines are always generated on platforms that need them for nested functions.

• -ftrampoline-impl=[stack|heap] — by default, trampolines get generated on the stack. However, certain platforms (like the Apple M1) don't permit an executable stack. Compiling with -ftrampoline-impl=heap generates calls to "__gcc_nested_func_ptr_created" and "__gcc_nested_func_ptr_deleted" so that trampoline space gets allocated and deallocated on the executable heap. Those functions are implemented in libgcc, and will only be provided on specific targets: x86_64 Darwin, and x86_64 and aarch64 Linux. PLEASE NOTE: heap trampolines are not guaranteed to be correctly deallocated if you "setjmp", instantiate nested functions, and then "longjmp" back to a state prior to having allocated those nested functions. 😭 Do not play with that.
• -fvisibility=[default|internal|hidden|protected] — sets the default ELF image symbol visibility to whatever option you pick — every symbol gets marked with this unless it's overridden inside the code. Using this feature can very substantially improve linking and load times of shared object libraries, produce more optimized code, provide near-perfect API export, and prevent symbol clashes. It is strongly recommended that you use this in any shared objects you distribute. 👑

Despite the fancy nomenclature, default always means public — i.e., available to be linked against from outside the shared object. protected and internal are pretty useless in real-world usage, so the only other commonly used option is hidden. And if -fvisibility isn't specified, the default is default, i.e., make every symbol public.

A seriously good explanation of the benefits of making sure your ELF symbols have the correct visibility is in "How To Write Shared Libraries" by Ulrich Drepper (you can find it at https://www.akkadia.org/drepper/) — however, a superior solution this option makes possible, versus marking things hidden when the default is public, is to make the default hidden and mark things public. That's the norm with DLLs on Windows, and with -fvisibility=hidden and "__attribute__ ((visibility("default")))" instead of "__declspec(dllexport)" you get almost identical semantics with identical syntax. Total boon for anyone working on cross-platform projects. 💖

so like, bestie, if you're adding visibility support to code that's already out there? you're gonna want #pragma GCC visibility. it works by you wrapping the declarations you wanna set visibility on with stuff like #pragma GCC visibility push(hidden) and #pragma GCC visibility pop. 🫶

and real talk — symbol visibility is part of your API interface contract, okay? so all new code should always specify visibility whenever it's not the default. like, declarations that are only for use within the local DSO should always be marked explicitly as hidden, so you dodge PLT indirection overheads — and being loud about it also makes the code more readable and self-documenting. that's the vibe. ✨

heads up tho: because of ISO C++ specification requirements, operator new and operator delete must always be of default visibility. no exceptions, that's the rule. 💅

also be aware that headers from outside your project — especially system headers and headers from any other library you use — might not be expecting to be compiled with visibility other than the default. so you may need to explicitly say #pragma GCC visibility push(default) before including any of those. 🌸

extern declarations are not affected by -fvisibility, which means a lot of code can be recompiled with -fvisibility=hidden with zero modifications. slay. but — and this is the catch — that also means calls to extern functions with no explicit visibility use the PLT. so it's more effective to use __attribute ((visibility)) and/or #pragma GCC visibility to tell the compiler which extern declarations should be treated as hidden. 🔥

ngl -fvisibility does affect C++ vague linkage entities. so like, an exception class that's thrown between DSOs must be explicitly marked with default visibility so the type_info nodes get unified between the DSOs. periodt.

if you want an overview of these techniques, their benefits, and how to use them, it's at https://gcc.gnu.org/wiki/Visibility. 📚

---

-fstrict-volatile-bitfields 💖

use this one if accesses to volatile bit-fields (or other structure fields, though the compiler usually honors those types anyway) should be a single access of the width of the field's type, aligned to natural alignment if possible.

okay example time: targets with memory-mapped peripheral registers might require all such accesses to be 16 bits wide. with this flag you can declare all peripheral bit-fields as unsigned short (assuming short is 16 bits on those targets) to force GCC to use 16-bit accesses instead of, maybe, a more efficient 32-bit access. 🎀

if this option is disabled, the compiler uses the most efficient instruction. in that previous example, that might be a 32-bit load instruction — even though that touches bytes that don't contain any portion of the bit-field, or memory-mapped registers unrelated to the one being updated. 😭

in some cases, like when the packed attribute is applied to a structure field, it may not be possible to access the field with a single read or write that's correctly aligned for the target machine. in that case GCC falls back to generating multiple accesses instead of code that faults or truncates the result at run time. king behavior honestly.

Note: due to restrictions of the C/C++11 memory model, write accesses are not allowed to touch non bit-field members. so it's recommended to define all bits of the field's type as bit-field members. 🫡

the default value of this option is determined by the application binary interface for the target processor.

-fsync-libcalls

this one controls whether any out-of-line instance of the __sync family of functions may be used to implement the C++11 __atomic family of functions.

the default value is enabled, which means the only useful form of the option is -fno-sync-libcalls. this option is used in the implementation of the libatomic runtime library. 💜

-fzero-init-padding-bits=value

guarantees zero initialization of padding bits in automatic variable initializers. certain languages guarantee zero initialization of padding bits in certain cases — e.g. C23 when using empty initializers ({}), or C++ when using zero-initialization, or C guarantees that fields not specified in an initializer have their padding bits zero initialized. this option lets you change when padding bits in initializers are guaranteed to be zero initialized. 🌈

the default is -fzero-init-padding-bits=standard, which makes no further guarantees than the corresponding standard. e.g.

struct A { char a; unsigned long long b; char c; };
union B { char a; unsigned long long b; };
struct A a = {}; // C23 guarantees padding bits are zero.
struct A b = { 1, 2, 3 }; // No guarantees.
union B c = {}; // C23 guarantees padding bits are zero.
union B d = { 1 }; // No guarantees.

-fzero-init-padding-bits=unions guarantees zero initialization of padding bits in unions on top of what the standards guarantee, if the initializer of a union is empty (then all bits of the union are zero initialized) or if the initialized member of the union is smaller than the size of the union (in that case it guarantees padding bits outside of the initialized member to be zero initialized). this was the GCC behavior before GCC 15, and in the above example it guarantees zero initialization of last sizeof (unsigned long long) - 1 bytes in the union. 👑

-fzero-init-padding-bits=all guarantees additionally zero initialization of padding bits of other aggregates, so the padding in between b.a and b.b (if any) and tail padding in the structure (if any).

NOTE: In Ubuntu 25.10 and later versions, -fzero-init-padding-bits=all is enabled by default for C, C++, ObjC, ObjC++, unless -fzero-init-padding-bits=standard is found. 🌟

---

🛠️ GCC Developer Options 🛠️

this section is for the command-line options that are primarily of interest to GCC developers — like options to support compiler testing and investigation of compiler bugs and compile-time performance problems. we're talking options that produce debug dumps at various points in the compilation, options that print statistics like memory use and execution time, and options that print info about GCC's configuration, like where it searches for libraries. 💾

fr fr, you should rarely need any of these for ordinary compilation and linking tasks. they're for the girlies compiling the compiler. 💅

many developer options that cause GCC to dump output to a file take an optional =filename suffix. you can specify stdout or - to dump to standard output, and stderr for standard error. 📝

if =filename is omitted, a default dump file name is constructed by concatenating the base dump file name, a pass number, a phase letter, and a pass name. the base dump file name is the name of the output file produced by the compiler if explicitly specified and not an executable; otherwise it's the source file name. the pass number is determined by the order passes are registered with the compiler's pass manager — this is generally the same as the order of execution, but passes registered by plugins, target-specific passes, or passes that are otherwise registered late are numbered higher than the pass named final, even if they're executed earlier. the phase letter is one of i (inter-procedural analysis), l (language-specific), r (RTL), or t (tree). the files are created in the directory of the output file. 🗂️

-fcallgraph-info, -fcallgraph-info=MARKERS 🦋

makes the compiler output callgraph information for the program, on a per-object-file basis. the information is generated in the common VCG format. it can be decorated with additional, per-node and/or per-edge information if a list of comma-separated markers is additionally specified. 🔍

when the su marker is specified, the callgraph is decorated with stack usage information; it's equivalent to -fstack-usage. when the da marker is specified, the callgraph is decorated with info about dynamically allocated objects. 💖

when compiling with -flto, no callgraph information is output along with the object file. at LTO link time, -fcallgraph-info may generate multiple callgraph information files next to intermediate LTO output files. (iconic, honestly.)

-dletters, -fdump-rtl-pass, -fdump-rtl-pass=filename ⭐

says to make debugging dumps during compilation at times specified by letters. this is used for debugging the RTL-based passes of the compiler. 🔧

some -dletters switches have different meaning when -E is used for preprocessing.

debug dumps can be enabled with a -fdump-rtl switch or some -d option letters. here are the possible letters for use in pass and letters, and their meanings:

• -fdump-rtl-alignments — dump after branch alignments have been computed.
• -fdump-rtl-asmcons — dump after fixing rtl statements that have unsatisfied in/out constraints.
• -fdump-rtl-auto_inc_dec — dump after auto-inc-dec discovery. this pass is only run on architectures that have auto inc or auto dec instructions.
• -fdump-rtl-barriers — dump after cleaning up the barrier instructions.
• -fdump-rtl-bbpart — dump after partitioning hot and cold basic blocks.
• -fdump-rtl-bbro — dump after block reordering.
• -fdump-rtl-btl1, -fdump-rtl-btl2 — -fdump-rtl-btl1 and -fdump-rtl-btl2 enable dumping after the two branch target load optimization passes.
• -fdump-rtl-bypass — dump after jump bypassing and control flow optimizations.
• -fdump-rtl-combine — dump after the RTL instruction combination pass.
• -fdump-rtl-compgotos — dump after duplicating the computed gotos.
• -fdump-rtl-ce1, -fdump-rtl-ce2, -fdump-rtl-ce3 — -fdump-rtl-ce1, -fdump-rtl-ce2, and -fdump-rtl-ce3 enable dumping after the three if conversion passes.
• -fdump-rtl-cprop_hardreg — dump after hard register copy propagation.
• -fdump-rtl-csa — dump after combining stack adjustments.
• -fdump-rtl-cse1, -fdump-rtl-cse2 — -fdump-rtl-cse1 and -fdump-rtl-cse2 enable dumping after the two common subexpression elimination passes.
• -fdump-rtl-dce — dump after the standalone dead code elimination passes.
• -fdump-rtl-dbr — dump after delayed branch scheduling.
• -fdump-rtl-dce1, -fdump-rtl-dce2 — ngl these two are a lil duo 💖 -fdump-rtl-dce1 and -fdump-rtl-dce2 let you dump after the two dead store elimination passes. two passes, two dumps, they're a set, don't split them up bestie.
• -fdump-rtl-eh — dumps after the finalization of EH handling code ✨
• -fdump-rtl-eh_ranges — dumps after conversion of EH handling range regions 🫶
• -fdump-rtl-expand — dumps after RTL generation. the origin story, literally 🌱
• -fdump-rtl-fwprop1, -fdump-rtl-fwprop2 — -fdump-rtl-fwprop1 and -fdump-rtl-fwprop2 enable dumping after the two forward propagation passes 🔥 another matching pair, we love to see it.
• -fdump-rtl-gcse1, -fdump-rtl-gcse2 — -fdump-rtl-gcse1 and -fdump-rtl-gcse2 enable dumping after global common subexpression elimination 👑
• -fdump-rtl-init-regs — dumps after the initialization of the registers 🌸 fresh registers, clean slate, new era.
• -fdump-rtl-initvals — dumps after the computation of the initial value sets 💜
• -fdump-rtl-into_cfglayout — dumps after converting to cfglayout mode 🦋 entering a whole new mode, main character transition fr.
• -fdump-rtl-ira — dumps after iterated register allocation ✨
• -fdump-rtl-jump — dumps after the second jump optimization 🎀
• -fdump-rtl-loop2 — -fdump-rtl-loop2 enables dumping after the rtl loop optimization passes 🔁
• -fdump-rtl-mach — dumps after performing the machine dependent reorganization pass, if that pass exists. okay so the "if that pass exists" is doing a LOT of heavy lifting there, don't come for me if it's a no-show 🫠
• -fdump-rtl-mode_sw — dumps after removing redundant mode switches 💅
• -fdump-rtl-rnreg — dumps after register renumbering 🔢
• -fdump-rtl-outof_cfglayout — dumps after converting from cfglayout mode 🌈 leaving the mode, the exit arc, we stan a full circle moment.
• -fdump-rtl-peephole2 — dumps after the peephole pass 👀 tiny window, huge impact, iconic behaviour.
• -fdump-rtl-postreload — dumps after post-reload optimizations 💖
• -fdump-rtl-pro_and_epilogue — dumps after generating the function prologues and epilogues 🎬 the intro AND the outro, we love a full arc.
• -fdump-rtl-sched1, -fdump-rtl-sched2 — -fdump-rtl-sched1 and -fdump-rtl-sched2 enable dumping after the basic block scheduling passes ⏰
• -fdump-rtl-ree — dumps after sign/zero extension elimination ✂️ bye to the extra bits babe.
• -fdump-rtl-seqabstr — dumps after common sequence discovery 🔍 found them. found the common sequences.
• -fdump-rtl-shorten — dumps after shortening branches 🌿
• -fdump-rtl-split1, -fdump-rtl-split2, -fdump-rtl-split3, -fdump-rtl-split4, -fdump-rtl-split5 — these options enable dumping after five rounds of instruction splitting 💀 five. rounds. girl that's a whole saga.
• -fdump-rtl-sms — dumps after modulo scheduling. heads up bestie: this pass is only run on some architectures 🙈
• -fdump-rtl-stack — dumps after conversion from GCC's "flat register file" registers to the x87's stack-like registers. and no cap, this pass is only run on x86 variants 💅
• -fdump-rtl-subreg1, -fdump-rtl-subreg2 — -fdump-rtl-subreg1 and -fdump-rtl-subreg2 enable dumping after the two subreg expansion passes 🦋
• -fdump-rtl-vartrack — dumps after variable tracking 📍
• -fdump-rtl-vregs — dumps after converting virtual registers to hard registers ⚡ the glow-up from virtual to hard, obsessed.
• -fdump-rtl-web — dumps after live range splitting 🕸️ (web. get it.)
• -fdump-rtl-regclass, -fdump-rtl-subregs_of_mode_init, -fdump-rtl-subregs_of_mode_finish, -fdump-rtl-dfinit, -fdump-rtl-dfinish — okay so these dumps are defined but always produce empty files 💀 they're giving ghost town. they exist. that's it. that's the vibe.
• -da, -fdump-rtl-all — produce all the dumps listed above 🌈✨ the full collection, the whole gallery, every single one.
• -dA — annotate the assembler output with miscellaneous debugging information 🎀
• -dD — dump all macro definitions, at the end of preprocessing, in addition to normal output 📝
• -dH — produce a core dump whenever an error occurs 💀 yeah that's a whole core dump on every error, so like... be aware bestie.
• -dp — annotate the assembler output with a comment indicating which pattern and alternative is used. the length and cost of each instruction are also printed too 💅 receipts included.
• -dP — dump the RTL in the assembler output as a comment before each instruction. also turns on -dp annotation 🔥
• -dx — just generate RTL for a function instead of compiling it. usually used with -fdump-rtl-expand 🧚
• -fdump-debug — dump debugging information generated during the debug generation phase 🐛
• -fdump-earlydebug — dump debugging information generated during the early debug generation phase 🌅 early bird edition.
• -fdump-noaddr — when doing debugging dumps, suppress address output. this makes it more feasible to use diff on debugging dumps for compiler invocations with different compiler binaries and/or different text / bss / data / heap / stack / dso start locations 🫶 genuinely so useful for comparing, no cap.
• -freport-bug — collect and dump debug information into a temporary file if an internal compiler error (ICE) occurs 🧊 yes, ICE. cute acronym for a not-cute situation.
• -fdump-unnumbered — when doing debugging dumps, suppress instruction numbers and address output. this makes it more feasible to use diff on debugging dumps for compiler invocations with different options, in particular with and without -g ✨
• -fdump-unnumbered-links — when doing debugging dumps (see -d option above), suppress instruction numbers for the links to the previous and next instructions in a sequence 🔗
• -fdump-ipa-switch, -fdump-ipa-switch-options — control the dumping at various stages of inter-procedural analysis language tree to a file. the file name is generated by appending a switch specific suffix to the source file name, and the file is created in the same directory as the output file 📁 here's the lineup of possible dumps:
  • all — enables all inter-procedural analysis dumps 🌈
  • cgraph — dumps information about call-graph optimization, unused function removal, and inlining decisions 👑
  • inline — dump after function inlining 🫶
  • strubm — dump after selecting "strub" modes, and recording the selections as function attributes 🎀
  • strub — dump "strub" transformations: interface changes, function wrapping, and insertion of builtin calls for stack scrubbing and watermarking 💅

additionally, the options -optimized, -missed, -note, and -all can be provided, with the same meaning as for -fopt-info, defaulting to -optimized.

for example, -fdump-ipa-inline-optimized-missed will emit information on callsites that were inlined, along with callsites that were not inlined 👀

by default, the dump will contain messages about successful optimizations (equivalent to -optimized) together with low-level details about the analysis 💖

• -fdump-ipa-clones — create a dump file containing information about creation of call graph node clones and removals of call graph nodes during inter-procedural optimizations and transformations. its main intended use is that tools that create live-patches can determine the set of functions that need to be live-patched to completely replace a particular function (see -flive-patching). the file name is generated by appending suffix "ipa-clones" to the source file name, and the file is created in the same directory as the output file. each entry in the file is on a separate line containing semicolon separated fields 📋

in the case of call graph clone creation, the individual fields are:

1. String "Callgraph clone".
2. Name of the function being cloned as it is presented to the assembler.
3. A number that uniquely represents the function being cloned in the call graph. Note that the number is unique only within a compilation unit or within whole-program analysis but is likely to be different in the two phases.
4. The file name of the source file where the function is defined.
5. The line on which the function definition is located.
6. The column where the function definition is located.
7. Name of the new function clone as it is presented to the assembler.
8. A number that uniquely represents the new function clone in the call graph. Note that the number is unique only within a compilation unit or within whole-program analysis but is likely to be different in the two phases.
9. The file name of the source file where the source code location of the new clone points to.
10. The line to which the source code location of the new clone points to.
11. The column to which the source code location of the new clone points to.
12. A string that determines the reason for cloning.

in the case of call graph clone removal, the individual fields are:

1. String "Callgraph removal".
2. Name of the function being removed as it would be presented to the assembler.
3. A number that uniquely represents the function being cloned in the call graph. Note that the number is unique only within a compilation unit or within whole-program analysis but is likely to be different in the two phases.
4. The file name of the source file where the function is defined.
5. The line on which the function definition is located.
6. The column where the function definition is located.

• -fdump-lang — dump language-specific information. the file name is made by appending .lang to the source file name 🏷️
• -fdump-lang-all, -fdump-lang-switch, -fdump-lang-switch-options, -fdump-lang-switch-options=filename — control the dumping of language-specific information. the options and filename portions behave as described in the -fdump-tree option. the following switch values are accepted 💖
  • all — enable all language-specific dumps 🌈
  • class — dump class hierarchy information. Virtual table information is emitted unless 'slim' is specified. this option is applicable to C++ only 👑
  • module — dump module information. Options lineno (locations), graph (reachability), blocks (clusters), uid (serialization), alias (mergeable), asmname (Elrond), eh (mapper) & vops (macros) may provide additional information. this option is applicable to C++ only 🧚
  • raw — dump the raw internal tree data. this option is applicable to C++ only 🌱
• -fdump-passes — print on stderr the list of optimization passes that are turned on and off by the current command-line options 📢
• -fdump-statistics-option — bestie, this turns on and controls dumping pass statistics into its own little file ✨. The filename gets made by tacking a suffix onto the source file name that ends in .statistics, and the file is created in the same directory as the output file, no cap. If you use the -option form: -stats makes counters get summed over the whole compilation unit, while -details dumps every single event as the passes generate them. With no option at all? Default is summing counters for each function compiled. Cute and simple 💅
• -fdump-tree-all
• -fdump-tree-switch
• -fdump-tree-switch-options
• -fdump-tree-switch-options=filename — these control the dumping at various stages of processing the intermediate language tree, into a file 🎀. If you use the -options form, then options is a list of - separated options that control the details of the dump. Not every option applies to every dump — the ones that aren't meaningful just get ignored, no drama. Here's the full menu of options available:
  • address — prints the address of each node. Usually this isn't meaningful because it changes based on the environment and the source file 🙈. Its main character use is tying up a dump file with a debug environment.
  • asmname — if DECL_ASSEMBLER_NAME has been set for a given decl, use that in the dump instead of DECL_NAME. Its primary vibe is making it easy to work backward from mangled names in the assembly file.
  • slim — when dumping front-end intermediate representations, this inhibits dumping of members of a scope or body of a function just because that scope has been reached. Only dump such items when they're directly reachable by some other path 🦋. When dumping pretty-printed trees, this option inhibits dumping the bodies of control structures. When dumping RTL, print the RTL in slim (condensed) form instead of the default LISP-like representation.
  • raw — print a raw representation of the tree. By default, trees are pretty-printed into a C-like representation.
  • details — enable more detailed dumps (not honored by every dump option 🫶). Also includes information from the optimization passes.
  • stats — enable dumping various statistics about the pass (also not honored by every dump option).
  • blocks — enable showing basic block boundaries (disabled in raw dumps).
  • graph — for each of the other indicated dump files (-fdump-rtl-pass), dump a representation of the control flow graph suitable for viewing with GraphViz to file.passid.pass.dot 🌈. Each function in the file is pretty-printed as a subgraph, so GraphViz can render them all in a single plot. This option currently only works for RTL dumps, and the RTL is always dumped in slim form.
  • vops — enable showing virtual operands for every statement.
  • lineno — enable showing line numbers for statements.
  • uid — enable showing the unique ID (DECL_UID) for each variable.
  • verbose — enable showing the tree dump for each statement.
  • eh — enable showing the EH region number holding each statement.
  • scev — enable showing scalar evolution analysis details.
  • optimized — enable showing optimization information (only available in certain passes).
  • missed — enable showing missed optimization information (only available in certain passes).
  • note — enable other detailed optimization information (only available in certain passes).
  • all — turn on all options, except raw, slim, verbose and lineno 💀.
  • optall — turn on all optimization options, i.e., optimized, missed, and note.

So to figure out what tree dumps are available, or to find the dump for a pass you're obsessed with, follow these steps 🧚:

1. Invoke GCC with -fdump-passes and in the stderr output look for a code that corresponds to the pass you're interested in. For example, the codes tree-evrp, tree-vrp1, and tree-vrp2 correspond to the three Value Range Propagation passes. The number at the end distinguishes distinct invocations of the same pass, iykyk.
2. To enable the creation of the dump file, append the pass code to the -fdump- option prefix and invoke GCC with it. For example, to enable the dump from the Early Value Range Propagation pass, invoke GCC with the -fdump-tree-evrp option. Optionally, you may specify the name of the dump file. If you don't specify one, GCC creates it as described below 👑
3. Find the pass dump in a file whose name is composed of three components separated by a period: the name of the source file GCC was invoked to compile, a numeric suffix indicating the pass number followed by the letter t for tree passes (and the letter r for RTL passes), and finally the pass code. For example, the Early VRP pass dump might be in a file named myfile.c.038t.evrp in the current working directory 🌸. Note that the numeric codes are not stable and may change from one version of GCC to another.

• -fopt-info
• -fopt-info-options
• -fopt-info-options=filename — this controls optimization dumps from various optimization passes ✨. If you use the -options form, options is a list of - separated option keywords to select the dump details and optimizations.

The options can be divided into three groups:

1. options describing what kinds of messages should be emitted,
2. options describing the verbosity of the dump, and
3. options describing which optimizations should be included.

You can freely mix options from each group since they're non-overlapping 💜. However, if there's any conflict, the later options override the earlier options on the command line.

The following options control which kinds of messages should be emitted:

• optimized — print information when an optimization is successfully applied. It's up to a pass to decide which information is relevant. For example, the vectorizer passes print the source location of loops which are successfully vectorized 🔥.
• missed — print information about missed optimizations. Individual passes control which information to include in the output.
• note — print verbose information about optimizations, such as certain transformations, more detailed messages about decisions, etc.
• all — print detailed optimization information. This includes optimized, missed, and note.

The following option controls the dump verbosity:

• internals — by default, only "high-level" messages are emitted. This option enables additional, more detailed, messages, which are likely to only be of interest to GCC developers 👩‍💻.

One or more of the following option keywords can be used to describe a group of optimizations:

• ipa — enable dumps from all interprocedural optimizations.
• loop — enable dumps from all loop optimizations.
• inline — enable dumps from all inlining optimizations.
• omp — enable dumps from all OMP (Offloading and Multi Processing) optimizations.
• vec — enable dumps from all vectorization optimizations.
• optall — enable dumps from all optimizations. This is a superset of the optimization groups listed above 🎀.

If options is omitted, it defaults to optimized-optall, which means to dump messages about successful optimizations from all the passes, omitting messages that are treated as "internals".

If the filename is provided, then the dumps from all the applicable optimizations are concatenated into the filename. Otherwise the dump is output onto stderr. Though multiple -fopt-info options are accepted, only one of them can include a filename. If other filenames are provided then all but the first such option are ignored 😭.

Note that the output filename is overwritten in case of multiple translation units. If a combined output from multiple translation units is desired, stderr should be used instead.

In the following example, the optimization info is output to stderr:

        gcc -O3 -fopt-info

This example:

        gcc -O3 -fopt-info-missed=missed.all

outputs missed optimization report from all the passes into missed.all, and this one:

        gcc -O2 -ftree-vectorize -fopt-info-vec-missed

prints information about missed optimization opportunities from vectorization passes on stderr 🦋. Note that -fopt-info-vec-missed is equivalent to -fopt-info-missed-vec. The order of the optimization group names and message types listed after -fopt-info does not matter, bestie.

As another example,

        gcc -O3 -fopt-info-inline-optimized-missed=inline.txt

outputs information about missed optimizations as well as optimized locations from all the inlining passes into inline.txt ⭐.

Finally, consider:

        gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt

Here the two output filenames vec.miss and loop.opt are in conflict since only one output file is allowed 💀. In this case, only the first option takes effect and the subsequent options are ignored. Thus only vec.miss is produced, which contains dumps from the vectorizer about missed opportunities.

• -fsave-optimization-record — write a SRCFILE.opt-record.json.gz file detailing what optimizations were performed, for those optimizations that support -fopt-info 🌸.

This option is experimental and the format of the data within the compressed JSON file is subject to change ⚠️.

It's roughly equivalent to a machine-readable version of -fopt-info-all, as a collection of messages with source file, line number and column number, plus the following additional data for each message:

• the execution count of the code being optimized, along with metadata about whether this was from actual profile data, or just an estimate, allowing consumers to prioritize messages by code hotness 🔥,
• the function name of the code being optimized, where applicable,
• the "inlining chain" for the code being optimized, so that when a function is inlined into several different places (which might themselves be inlined), the reader can distinguish between the copies,

• objects that ID which parts of the message are about expressions, statements or symbol-table nodes — plus which of those categories they fall into, and when it's available, their source code location 📍
• the GCC pass that emitted the message 🔍
• the location in GCC's own code where the message got emitted 📌

Also lowkey: some messages are logically nested inside other messages, which mirrors the implementation details of the optimization passes. iykyk 😌

---

• -fsched-verbose=n — on targets that actually use instruction scheduling, this bad boy controls how much debugging output the scheduler prints into the dump files 🖨️

For n greater than zero, -fsched-verbose spits out the same info as -fdump-rtl-sched1 and -fdump-rtl-sched2. For n greater than one, it also outputs basic block probabilities, detailed ready list info, and unit/insn info. For n greater than two, it throws in RTL at abort point, control-flow and regions info. And for n over four, -fsched-verbose also includes dependence info. it's giving complete lore ✨

• -fenable-kind-pass
• -fdisable-kind-pass=range-list — this is a whole set of options for explicitly disabling/enabling optimization passes. Heads up bestie: these are meant for debugging GCC itself. If you're just a regular compiler user, use the normal options for enabling/disabling passes instead 🙈
  • -fdisable-ipa-pass — disable IPA pass pass. pass is the pass name. If the same pass gets statically invoked in the compiler multiple times, tack a sequential number onto the pass name, starting from 1.
  • -fdisable-rtl-pass
  • -fdisable-rtl-pass=range-list — disable RTL pass pass. pass is the pass name, and same deal: if it's statically invoked multiple times, append a sequential number starting from 1. The range-list is a comma-separated list of function ranges or assembler names. Each range is a number pair separated by a colon, and the range is inclusive on both ends. If the range is trivial, that number pair can be simplified down to a single number. If the function's call graph node's uid falls within one of the specified ranges, the pass is disabled for that function. That uid shows up in the function header of a dump file, and you can dump the pass names by using option -fdump-passes.
  • -fdisable-tree-pass
  • -fdisable-tree-pass=range-list — disable tree pass pass. See -fdisable-rtl for the description of the option arguments.
  • -fenable-ipa-pass — enable IPA pass pass. pass is the pass name. Same sequential-number rule if the pass is statically invoked multiple times, starting from 1.
  • -fenable-rtl-pass
  • -fenable-rtl-pass=range-list — enable RTL pass pass. See -fdisable-rtl for the option argument description and examples.
  • -fenable-tree-pass
  • -fenable-tree-pass=range-list — enable tree pass pass. See -fdisable-rtl for the description of option arguments.

Ok so here are some examples showing uses of these options, bestie — bookmark this fr fr 🔖

# disable ccp1 for all functions
   -fdisable-tree-ccp1
# disable complete unroll for function whose cgraph node uid is 1
   -fenable-tree-cunroll=1
# disable gcse2 for functions at the following ranges [1,1],
# [300,400], and [400,1000]
# disable gcse2 for functions foo and foo2
   -fdisable-rtl-gcse2=foo,foo2
# disable early inlining
   -fdisable-tree-einline
# disable ipa inlining
   -fdisable-ipa-inline
# enable tree full unroll
   -fenable-tree-unroll

• -fchecking
• -fchecking=n — enable internal consistency checking. The default depends on how the compiler was configured. And -fchecking=2 enables further internal consistency checking that might affect code generation 👀 slay but careful
• -frandom-seed=string — this option hands GCC a seed that it uses in place of random numbers when generating certain symbol names that have to be different in every compiled file. It's also used to place unique stamps in coverage data files and the object files that produce them. You can use -frandom-seed to produce reproducibly identical object files 🎯

The string can either be a number (decimal, octal or hex) or an arbitrary string (and in that case it gets converted to a number by computing CRC32).

The string should be different for every file you compile. no cap 💀

• -save-temps — store the usual "temporary" intermediate files permanently, and name them as auxiliary output files, exactly as specified/described under -dumpbase and -dumpdir 🗂️

When you use it together with the -x command-line option, -save-temps is sensible enough to avoid overwriting an input source file that has the same extension as an intermediate file. You can get that corresponding intermediate file by renaming the source file before using -save-temps.

• -save-temps=cwd — equivalent to -save-temps -dumpdir ./.
• -save-temps=obj — equivalent to -save-temps -dumpdir outdir/, where outdir/ is the directory of the output file specified after the -o option, including any directory separators. And if the -o option isn't used, the -save-temps=obj switch behaves like -save-temps=cwd. cute ✨
• -time[=file] — report the CPU time taken by each subprocess in the compilation sequence. For C source files, that's the compiler proper and the assembler (plus the linker, if linking happens) ⏱️

Without an output file specified, the output looks like this:

# cc1 0.12 0.01
# as 0.00 0.01

The first number on each line is the "user time", aka the time spent executing the program itself. The second number is "system time", aka time spent executing operating system routines on behalf of the program. Both numbers are in seconds.

With an output file specified, the output gets appended to the named file, and it looks like this:

0.12 0.01 cc1 <options>
0.00 0.01 as <options>

The "user time" and the "system time" get moved before the program name, and the options passed to the program are displayed, so that later on you can tell what file was being compiled, and with which options. iconic detective work 🔍

• -fdump-final-insns[=file] — dump the final internal representation (RTL) to file. If the optional argument is omitted (or if file is .), the name of the dump file is determined by appending ".gkd" to the dump base name, see -dumpbase.
• -fcompare-debug[=opts] — if no error happens during compilation, run the compiler a second time, adding opts and -fcompare-debug-second to the arguments passed to that second compilation. Dump the final internal representation in both compilations, and print an error if they differ 🫠

If the equal sign is omitted, the default -gtoggle is used.

The environment variable GCC_COMPARE_DEBUG, if defined, non-empty and nonzero, implicitly enables -fcompare-debug. If GCC_COMPARE_DEBUG is defined to a string starting with a dash, then it's used for opts, otherwise the default -gtoggle is used.

-fcompare-debug=, with the equal sign but without opts, is equivalent to -fno-compare-debug, which disables the dumping of the final representation and the second compilation, preventing even GCC_COMPARE_DEBUG from taking effect.

To verify full coverage during -fcompare-debug testing, set GCC_COMPARE_DEBUG to say -fcompare-debug-not-overridden, which GCC rejects as an invalid option in any actual compilation (rather than preprocessing, assembly or linking). If you just want a warning instead, setting GCC_COMPARE_DEBUG to -w%n-fcompare-debug not overridden will do the trick.

• -fcompare-debug-second — this option is implicitly passed to the compiler for the second compilation requested by -fcompare-debug, along with options to silence warnings, and omitting other options that would cause the compiler to produce output to files or to standard output as a side effect. Dump files and preserved temporary files are renamed so they contain the ".gk" additional extension during the second compilation, so they don't overwrite the ones generated by the first 🎀

When this option is passed to the compiler driver, it causes the first compilation to be skipped, which makes it useful for little other than debugging the compiler proper. that's it, that's the vibe 💅

• -gtoggle — turn off generation of debug info if leaving out this option generates it, or turn it on at level 2 otherwise. The position of this argument in the command line does not matter; it takes effect after all other options are processed, and it does so only once, no matter how many times it's given. This is mainly intended to be used with -fcompare-debug 🌸
• -fvar-tracking-assignments-toggle — toggle -fvar-tracking-assignments, in the same way that -gtoggle toggles -g.
• -Q — when used on the command line prior to --help=, -Q acts as a modifier to the help output.

Otherwise, this option makes the compiler print out each function name as it's compiled, and print some statistics about each pass when it finishes 📊

• -ftime-report — makes the compiler print some statistics to stderr about the time consumed by each pass when it finishes.

If SARIF output of diagnostics was requested via -fdiagnostics-format=sarif-file or -fdiagnostics-format=sarif-stderr then the -ftime-report information is instead emitted in JSON form as part of the SARIF output. The precise format of this JSON data is subject to change, and the values may not exactly match those emitted to stderr since they're written out at a slightly different place within the compiler. so don't come for it if the numbers are a lil off 🙈

• -ftime-report-details — record the time consumed by infrastructure parts separately for each pass.
• -fira-verbose=n — control the verbosity of the dump file for the integrated register allocator. The default value is 5. If the value n is greater or equal to 10, the dump output is sent to stderr using the same format as n minus 10 ✨
• -flto-report — bestie, this prints out a whole report with the internal deets on how the link-time optimizer is doing its thing. Heads up tho: what's actually in that report changes from version to version, so don't get too attached 💀. It's meant to be useful for GCC developers processing object files in LTO mode (that's via -flto).

Disabled by default. (yeah, you gotta opt in 💅)

• -flto-report-wpa — like -flto-report, but it only prints for the WPA phase of link-time optimization. Just the one phase, no extras ✨
• -fmem-report — makes the compiler spill some stats about permanent memory allocation when it finishes. Like a lil memory receipts moment 🧾
• -fmem-report-wpa — makes the compiler print stats about permanent memory allocation, but only for the WPA phase. Exclusive behavior, we love that for it 👑
• -fpre-ipa-mem-report
• -fpost-ipa-mem-report — makes the compiler print stats about permanent memory allocation before or after interprocedural optimization. Before and after pics, but make it memory stats 📸
• -fmultiflags — okay this one is doing the most, let me break it down. This option enables multilib-aware "TFLAGS" to be used to build target libraries with options different from those the compiler is configured to use by default, through the use of specs set up by compiler internals, by the target, or by builders at configure time.

Like "TFLAGS", this allows the target libraries to be built for portable baseline environments, while the compiler defaults to more demanding ones. That's useful because users can easily override the defaults the compiler is configured to use to build their own programs, if the defaults are not ideal for their target environment, whereas rebuilding the runtime libraries is usually not as easy or desirable. Lowkey relatable — nobody wants to rebuild the runtime libraries, that's so much effort 😭

Unlike "TFLAGS", the use of specs enables different flags to be selected for different multilibs. The way to accomplish that is to build with make TFLAGS=-fmultiflags, after configuring --with-specs=%{fmultiflags:...}.

This option is discarded by the driver once it's done processing driver self spec. (bye bestie, your job here is done 🫶)

It is also useful to check that "TFLAGS" are being used to build all target libraries, by configuring a non-bootstrap compiler --with-specs='%{!fmultiflags:%emissing TFLAGS}' and building the compiler and target libraries.

• -fprofile-report — makes the compiler print some statistics about consistency of the (estimated) profile and effect of individual passes. Receipts for every pass 💅
• -fstack-usage — makes the compiler output stack usage information for the program, on a per-function basis. The filename for the dump is made by appending .su to the auxname. auxname is generated from the name of the output file, if explicitly specified and it is not an executable, otherwise it is the basename of the source file. An entry is made up of three fields:
  • The name of the function.
  • A number of bytes.
  • One or more qualifiers: "static", "dynamic", "bounded".

The qualifier "static" means that the function manipulates the stack statically: a fixed number of bytes are allocated for the frame on function entry and released on function exit; no stack adjustments are otherwise made in the function. The second field is this fixed number of bytes.

The qualifier "dynamic" means that the function manipulates the stack dynamically: in addition to the static allocation described above, stack adjustments are made in the body of the function, for example to push/pop arguments around function calls. If the qualifier "bounded" is also present, the amount of these adjustments is bounded at compile time and the second field is an upper bound of the total amount of stack used by the function. If it is not present, the amount of these adjustments is not bounded at compile time and the second field only represents the bounded part.

• -fstats — emit statistics about front-end processing at the end of the compilation. Ngl this one is exclusive — it's only supported by the C++ front end, and the info is generally only useful to the G++ development team. So if you're not on that team, it's giving "for the girls who get it" energy 🙈
• -fdbg-cnt-list — print the name and the counter upper bound for all debug counters. The full roster, no gatekeeping ✨
• *-fdbg-cnt=counter-value-list* — set the internal debug counter lower and upper bound. counter-value-list is a comma-separated list of name:lower_bound1-upper_bound1 [:lower_bound2-upper_bound2...] tuples which sets the name of the counter and list of closed intervals. The lower_bound is optional and is zero initialized if not set. For example, with -fdbg-cnt=dce:2-4:10-11,tail_call:10, dbg_cnt(dce) returns true only for second, third, fourth, tenth and eleventh invocation. For dbg_cnt(tail_call) true is returned for first 10 invocations.
• *-print-file-name=library* — print the full absolute name of the library file library that would be used when linking---and don't do anything else. With this option, GCC does not compile or link anything; it just prints the file name. Just the name bestie, that's all it does, that's the whole job 💅
• -print-multi-directory — print the directory name corresponding to the multilib selected by any other switches present in the command line. This directory is supposed to exist in GCC_EXEC_PREFIX. "Supposed to" is doing a lot of heavy lifting there fr fr
• -print-multi-lib — print the mapping from multilib directory names to compiler switches that enable them. The directory name is separated from the switches by ;, and each switch starts with an @ instead of the -, without spaces between multiple switches. This is supposed to ease shell processing. (it's giving "made for script girlies" 🐚)
• -print-multi-os-directory — print the path to OS libraries for the selected multilib, relative to some lib subdirectory. If OS libraries are present in the lib subdirectory and no multilibs are used, this is usually just ., if OS libraries are present in libsuffix sibling directories this prints e.g. ../lib64, ../lib or ../lib32, or if OS libraries are present in lib/subdir subdirectories it prints e.g. amd64, sparcv9 or ev6.
• -print-multiarch — print the path to OS libraries for the selected multiarch, relative to some lib subdirectory. The sequel to the last one, now with multiarch flavor 🌈
• *-print-prog-name=program* — like -print-file-name, but searches for a program such as cpp. Program hunting mode activated 🔍
• -print-libgcc-file-name — same as -print-file-name=libgcc.a.

This is useful when you use -nostdlib or -nodefaultlibs but you do want to link with libgcc.a. You can do:

gcc -nostdlib <files>... `gcc -print-libgcc-file-name`

• -print-search-dirs — print the name of the configured installation directory and a list of program and library directories gcc searches---and don't do anything else.

This is useful when gcc prints the error message installation problem, cannot exec cpp0: No such file or directory. To resolve this you either need to put cpp0 and the other compiler components where gcc expects to find them, or you can set the environment variable GCC_EXEC_PREFIX to the directory where you installed them. Don't forget the trailing /. Bestie, I am BEGGING you, don't forget the trailing slash 😭

• -print-sysroot — print the target sysroot directory that is used during compilation. This is the target sysroot specified either at configure time or using the --sysroot option, possibly with an extra suffix that depends on compilation options. If no target sysroot is specified, the option prints nothing. Silence. The void. Nothing 💀
• -print-sysroot-headers-suffix — print the suffix added to the target sysroot when searching for headers, or give an error if the compiler is not configured with such a suffix---and don't do anything else. It's giving "suffix or error," no in-between 🙈
• -dumpmachine — print the compiler's target machine (for example, i686-pc-linux-gnu)---and don't do anything else. Machine reveal moment 👑
• -dumpversion — print the compiler version (for example, 3.0, 6.3.0 or 7)---and don't do anything else. This is the compiler version used in filesystem paths and specs. Depending on how the compiler has been configured it can be just a single number (major version), two numbers separated by a dot (major and minor version) or three numbers separated by dots (major, minor and patchlevel version). So yeah, it might lowkey be a surprise how much or how little you get 💅
• -dumpfullversion — print the full compiler version---and don't do anything else. The output is always three numbers separated by dots, major, minor and patchlevel version. No mystery here, full version only, periodt ✨
• -dumpspecs — print the compiler's built-in specs---and don't do anything else. (This is used when GCC itself is being built.) Meta moment, looking at your own specs 🪞

🦋 Machine-Dependent Options 🦋

Each target machine supported by GCC can have its own options---for example, to allow you to compile for a particular processor variant or ABI, or to control optimizations specific to that machine. By convention, the names of machine-specific options start with -m. (the -m is for "machine," iykyk 💅)

Some configurations of the compiler also support additional target-specific options, usually for compatibility with other compilers on the same platform.

💖 AArch64 Options 💖

These options are defined for AArch64 implementations:

• *-mabi=name* — generate code for the specified data model. Permissible values are ilp32 for SysV-like data model where int, long int and pointers are 32 bits, and lp64 for SysV-like data model where int is 32 bits, but long int and pointers are 64 bits.

The default depends on the specific target configuration. Note that the LP64 and ILP32 ABIs are not link-compatible; you must compile your entire program with the same ABI, and link with a compatible set of libraries. Mixing them is giving chaos and not the cute kind 😭

The ilp32 model is deprecated. Just so you know, bestie 💀

• -mbig-endian — generate big-endian code. This is the default when GCC is configured for an aarch64_be-*-* target.
• -mgeneral-regs-only — generate code which uses only the general-purpose registers. This will prevent the compiler from using floating-point and Advanced SIMD registers but will not impose any restrictions on the assembler. GP registers only, we're keeping it minimal ✨
• -mlittle-endian — generate little-endian code. This is the default when GCC is configured for an aarch64-*-* but not an aarch64_be-*-* target.
• -mcmodel=tiny — generate code for the tiny code model. The program and its statically defined symbols must be within 1MB of each other. Programs can be statically or dynamically linked. Tiny but mighty 🧚
• -mcmodel=small — generate code for the small code model. The program and its statically defined symbols must be within 4GB of each other. Programs can be statically or dynamically linked. This is the default code model. (the main character of code models 👑)
• -mcmodel=large — generate code for the large code model. This makes no assumptions about addresses and sizes of sections. Programs can be statically linked only. The -mcmodel=large option is incompatible with -mabi=ilp32, -fpic and -fPIC. So don't even try to pair them, it's not gonna work 💀
• -mtp=name — tells GCC which system register to use as the thread pointer. valid names are tpidr_el0, tpidrro_el0, tpidr_el1, tpidr_el2, tpidr_el3 💫. for backwards compat the aliases el0, el1, el2, el3 also work. default is tpidr_el0. bestie, if you want your code to interoperate you gotta compile everything with the same value of this option, otherwise you'll end up reading a different thread pointer from the wrong exception level and that's a whole mess 😭
• -mstrict-align, -mno-strict-align — avoid or allow generating memory accesses that might not be aligned on a natural object boundary like the architecture spec says. strict = no unaligned shenanigans, no-strict = live your truth 🦋
• -momit-leaf-frame-pointer, -mno-omit-leaf-frame-pointer — omit or keep the frame pointer in leaf functions. the omit behaviour is the default, no cap.
• -mstack-protector-guard=guard, -mstack-protector-guard-reg=reg, -mstack-protector-guard-offset=offset — generate stack protection code with the canary sitting at guard 🐤. supported locations: global for a global canary, or sysreg for a canary in an appropriate system register.

with sysreg, the -mstack-protector-guard-reg=reg and -mstack-protector-guard-offset=offset options also say which system register to use as base register for reading the canary, and what offset from that base register to read from. there is no default register or offset, because this is entirely for use inside the Linux kernel. periodt.

• -mtls-dialect=desc — use TLS descriptors as the thread-local storage mechanism for dynamic accesses of TLS variables. this is the default 💅
• -mtls-dialect=traditional — use traditional TLS as the thread-local storage mechanism for dynamic accesses of TLS variables. the vintage option, iconic in its own way.
• -mtls-size=size — specify the bit size of immediate TLS offsets. valid values are 12, 24, 32, 48. heads up: this option requires binutils 2.26 or newer.
• -mfix-cortex-a53-835769, -mno-fix-cortex-a53-835769 — enable or disable the workaround for ARM Cortex-A53 erratum number 835769. it works by inserting a NOP instruction between memory instructions and 64-bit integer multiply-accumulate instructions 🩹. this flag gets ignored if you specify an architecture or cpu on the command line that doesn't need the workaround. it's giving ghosted.
• -mfix-cortex-a53-843419, -mno-fix-cortex-a53-843419 — enable or disable the workaround for ARM Cortex-A53 erratum number 843419. this one is made at link time, so it just passes the corresponding flag to the linker. same deal: the flag is ignored if an architecture or cpu that doesn't need the workaround is specified on the command line.
• -mlow-precision-recip-sqrt, -mno-low-precision-recip-sqrt — enable or disable the reciprocal square root approximation. ⚠️ this only does anything if -ffast-math or -funsafe-math-optimizations is also used. turning it on drops the precision of reciprocal square root results to about 16 bits for single precision and 32 bits for double precision. precision? we don't know her 💀
• -mlow-precision-sqrt, -mno-low-precision-sqrt — enable or disable the square root approximation. again, only has an effect with -ffast-math or -funsafe-math-optimizations. enabling it reduces precision of square root results to about 16 bits single precision and 32 bits double precision. and if you enable it, it implies -mlow-precision-recip-sqrt. bundled, bestie 🎀
• -mlow-precision-div, -mno-low-precision-div — enable or disable the division approximation. same rule: only matters with -ffast-math or -funsafe-math-optimizations. enabling it drops division results to about 16 bits single precision and 32 bits double precision. speed over accuracy, main character energy 🔥
• -mtrack-speculation, -mno-track-speculation — enable or disable generating extra code to track speculative execution through conditional branches 🕵️. that tracking state can then be used by the compiler when expanding calls to __builtin_speculation_safe_copy, which lets it generate a more efficient code sequence.
• -moutline-atomics, -mno-outline-atomics — enable or disable calls to out-of-line helpers to implement atomic operations. these helpers decide at runtime whether the LSE instructions from ARMv8.1-A can be used; if not, they fall back to the load/store-exclusive instructions that exist in the base ARMv8.0 ISA ✨

this option is only applicable when compiling for the base ARMv8.0 instruction set. if you're on a later revision — say -march=armv8.1-a or -march=armv8-a+lse — the ARMv8.1-Atomics instructions get used directly. same story with -mcpu= when the selected cpu supports the lse feature. this option is on by default. slay.

• -march=name — specify the name of the target architecture, optionally plus one or more feature modifiers. the form is -march=arch{+[no]feature}*.

the table below sums up the permissible values for arch and the features they enable by default:

• armv8-a (Armv8-A) — includes by default: +fp, +simd
• armv8.1-a (Armv8.1-A) — includes by default: armv8-a, +crc, +lse, +rdma
• armv8.2-a (Armv8.2-A) — includes by default: armv8.1-a
• armv8.3-a (Armv8.3-A) — includes by default: armv8.2-a, +pauth, +fcma, +jscvt
• armv8.4-a (Armv8.4-A) — includes by default: armv8.3-a, +flagm, +fp16fml, +dotprod, +rcpc2
• armv8.5-a (Armv8.5-A) — includes by default: armv8.4-a, +sb, +ssbs, +predres, +frintts, +flagm2
• armv8.6-a (Armv8.6-A) — includes by default: armv8.5-a, +bf16, +i8mm
• armv8.7-a (Armv8.7-A) — includes by default: armv8.6-a, +wfxt, +xs
• armv8.8-a (Armv8.8-a) — includes by default: armv8.7-a, +mops
• armv8.9-a (Armv8.9-a) — includes by default: armv8.8-a
• armv9-a (Armv9-A) — includes by default: armv8.5-a, +sve, +sve2
• armv9.1-a (Armv9.1-A) — includes by default: armv9-a, +bf16, +i8mm
• armv9.2-a (Armv9.2-A) — includes by default: armv9.1-a, +wfxt, +xs
• armv9.3-a (Armv9.3-A) — includes by default: armv9.2-a, +mops
• armv9.4-a (Armv9.4-A) — includes by default: armv9.3-a, +sve2p1
• armv9.5-a (Armv9.5-A) — includes by default: armv9.4-a, cpa, +faminmax, +lut
• armv8-r (Armv8-R) — includes by default: armv8-r

the value native exists on native AArch64 GNU/Linux and makes the compiler pick the architecture of the host system 🌸. it does nothing if the compiler can't recognise the host architecture. and when -march=native is given with no other -mcpu or -mtune, GCC will pick the host CPU as the CPU to tune for as well as the source of architecture features — meaning -march=native is treated as -mcpu=native.

the permissible values for feature live in the sub-section on aarch64-feature-modifiers,,-march and -mcpu Feature Modifiers. where conflicting feature modifiers are specified, the right-most feature wins 👑

GCC uses name to decide what kind of instructions it can emit when generating assembly code. if -march is given without -mtune or -mcpu also being given, the code gets tuned to perform well across a range of target processors implementing the target architecture. versatile, lowkey.

• -mtune=name — specify the name of the target processor for which GCC should tune the performance of the code. permissible values: generic, cortex-a35, cortex-a53, cortex-a55, cortex-a57, cortex-a72, cortex-a73, cortex-a75, cortex-a76, cortex-a76ae, cortex-a77, cortex-a65, cortex-a65ae, cortex-a34, cortex-a78, cortex-a78ae, cortex-a78c, ares, exynos-m1, emag, falkor, oryon-1, neoverse-512tvb, neoverse-e1, neoverse-n1, neoverse-n2, neoverse-v1, neoverse-v2, grace, neoverse-v3, neoverse-v3ae, armagicpu, neoverse-n3, olympus, cortex-a725, cortex-x925, qdf24xx, saphira, phecda, xgene1, vulcan, octeontx, octeontx81, octeontx83, octeontx2, octeontx2t98, octeontx2t96 octeontx2t93, octeontx2f95, octeontx2f95n, octeontx2f95mm, a64fx, fujitsu-monaka, thunderx, thunderxt88, thunderxt88p1, thunderxt81, tsv110, hip12, thunderxt83, thunderx2t99, thunderx3t110, zeus, cortex-a57.cortex-a53, cortex-a72.cortex-a53, cortex-a73.cortex-a35, cortex-a73.cortex-a53, cortex-a75.cortex-a55, cortex-a76.cortex-a55, cortex-r82, cortex-r82ae, cortex-x1, cortex-x1c, cortex-x2, cortex-x3, cortex-x4, cortex-a510, cortex-a520, cortex-a520ae, cortex-a710, cortex-a715, cortex-a720, cortex-a720ae, ampere1, ampere1a, ampere1b, cobalt-100, apple-m1, apple-m2, apple-m3 and native.

the values cortex-a57.cortex-a53, cortex-a72.cortex-a53, cortex-a73.cortex-a35, cortex-a73.cortex-a53, cortex-a75.cortex-a55, cortex-a76.cortex-a55, apple-m1, apple-m2, apple-m3, gb10 say that GCC should tune for a big.LITTLE system 🐘🐭

the value neoverse-512tvb says GCC should tune for Neoverse cores that (a) implement SVE and (b) have a total vector bandwidth of 512 bits per cycle. in other words, it tells GCC to tune for Neoverse cores that can execute 4 128-bit Advanced SIMD arithmetic instructions a cycle and can execute an equivalent number of SVE arithmetic instructions per cycle (2 for 256-bit SVE, 4 for 128-bit SVE). this is more general than tuning for a specific core like Neoverse V1 but more specific than the default tuning described below. nerdy but make it ✨

also, on native AArch64 GNU/Linux systems the value native tunes performance to the host system. this option does nothing if the compiler can't recognise the host processor.

where none of -mtune=, -mcpu= or -march= are specified, the code is tuned to perform well across a range of target processors.

⚠️ this option cannot be suffixed by feature modifiers. no +whatever for you, bestie.

• -mcpu=name — specify the name of the target processor, optionally suffixed with one or more feature modifiers. the form is -mcpu=cpu{+[no]feature}*, where the permissible values for cpu are the same as the ones available for -mtune 💖. the permissible values for feature are documented in the sub-section on aarch64-feature-modifiers,,-march and -mcpu Feature Modifiers. where conflicting feature modifiers are specified, the right-most feature wins.

GCC uses name to decide what kind of instructions it can emit when generating assembly code (as if by -march) and to decide the target processor to tune for performance (as if by -mtune). where this option is used together with -march or -mtune, those options take precedence over the appropriate part of this option. okay so -mcpu does the most but steps aside when the others show up 😌

---

-mcpu=neoverse-512tvb is the special one, bestie — it's not about one specific core at all 💅 it's about all the Neoverse cores that (a) implement SVE and (b) have a total vector bandwidth of 512 bits a cycle. Unless something overrides it with -march, -mcpu=neoverse-512tvb spits out code that can run on a Neoverse V1 core, because Neoverse V1 is literally the first Neoverse core that has these properties. And unless -mtune overrides it, -mcpu=neoverse-512tvb tunes code exactly the same way as -mtune=neoverse-512tvb does. ✨

• -moverride=string — overrides the tuning decisions that the back-end made in response to a -mtune= switch. Heads up though: the syntax, semantics, and accepted values for string in this option are not promised to stay consistent across releases. 🙈

This option is only meant to be useful when you're developing GCC itself, iykyk.

• -mverbose-cost-dump — turns on verbose cost model dumping in the debug dump files. This one exists purely for debugging the compiler, no cap. 🔍
• -mpc-relative-literal-loads, -mno-pc-relative-literal-loads — enable or disable PC-relative literal loads. With this on, literal pools are accessed using a single instruction and emitted after each function. That does cap the maximum size of functions at 1MB. It's enabled by default for -mcmodel=tiny. 🎀
• -msign-return-address=scope — picks which function scope gets return address signing applied. The permissible values are none (disables return address signing), non-leaf (enables pointer signing for functions that aren't leaf functions), and all (enables pointer signing for every function). Default value is none. Just so you know — this option has been deprecated by -mbranch-protection. 💀
• -mbranch-protection=none|standard|pac-ret[+leaf+b-key]|bti|gcs — chooses which branch protection features to use. none is the default and turns off all types of branch protection. standard turns on all types of branch protection features — and if a feature has extra tuning options, standard sets it to its standard level. pac-ret[+leaf] turns on return address signing to its standard level: it signs functions that save the return address to memory (and non-leaf functions will practically always do this) using the a-key. The optional argument leaf extends the signing to cover leaf functions too. The optional argument b-key signs the functions with the B-key instead of the A-key. bti turns on the branch target identification mechanism. gcs turns on guarded control stack compatible code generation. 🛡️
• -mharden-sls=opts — enables compiler hardening against straight line speculation (SLS). opts is a comma-separated list of the following options:
  • retbr
  • blr

On top of that, -mharden-sls=all enables all SLS hardening while -mharden-sls=none disables all SLS hardening. ⚔️

• -mearly-ra=scope — decides when to enable an early register allocation pass. This pass runs before instruction scheduling and tries to find a spill-free allocation of floating-point and vector code. It also tries to make use of strided multi-register instructions, like SME2's strided LD1 and ST1. 👑

The possible values of scope are: all, which runs the pass on all functions; strided, which runs it on functions that have access to strided multi-register instructions; and none, which disables the pass entirely.

-mearly-ra=all is the default for -O2 and above, and for -Os. -mearly-ra=none is the default otherwise. ✨

• -mearly-ldp-fusion — enables the copy of the AArch64 load/store pair fusion pass that runs before register allocation. Enabled by default at -O and above. 🫶
• -mlate-ldp-fusion — enables the copy of the AArch64 load/store pair fusion pass that runs after register allocation. Enabled by default at -O and above. 🫶
• -msve-vector-bits=bits — specifies the number of bits in an SVE vector register. Quick note: this option only does anything when SVE is enabled. 🌈

GCC supports two forms of SVE code generation: "vector-length agnostic" output that works with any size of vector register, and "vector-length specific" output that lets GCC make assumptions about the vector length when that's useful for optimization reasons. The possible values of bits are: scalable, 128, 256, 512, 1024 and 2048. Specifying scalable selects vector-length agnostic output. At present, -msve-vector-bits=128 also generates vector-length agnostic output for big-endian targets. Every other value generates vector-length specific code. The behavior of these values may change in future releases, and no value except scalable should be relied on for producing code that's portable across different hardware SVE vector lengths. 😭

The default is -msve-vector-bits=scalable, which produces vector-length agnostic code.

• -Wexperimental-fmv-target — warns about use of experimental Function Multi Versioning. The Arm C Language Extension specification for Function Multi Versioning is beta and subject to change. Any usage of FMV comes with the caveat that future behavior change and incompatibility is likely. ⚠️

---

🧚 -march and -mcpu Feature Modifiers 🧚

Feature modifiers you can use with -march and -mcpu can be any of the following, and also their inverses as nofeature: 💖

• crc — enable the CRC extension. This is on by default for -march=armv8.1-a.
• crypto — enable the Crypto extension. This also enables Advanced SIMD and floating-point instructions.
• fp — enable floating-point instructions. This is on by default for all possible values for the options -march and -mcpu.
• simd — enable Advanced SIMD instructions. This also enables floating-point instructions. This is on by default for all possible values for the options -march and -mcpu.
• sve — enable Scalable Vector Extension instructions. This also enables Advanced SIMD and floating-point instructions.
• lse — enable Large System Extension instructions. This is on by default for -march=armv8.1-a.
• rdma — enable Round Double Multiply Accumulate instructions. This is on by default for -march=armv8.1-a.
• fp16 — enable the FP16 extension. This also enables floating-point instructions.
• fp16fml — enable the FP16 fmla extension. This also enables FP16 extensions and floating-point instructions. This option is enabled by default for -march=armv8.4-a. Use of this option with architectures prior to Armv8.2-A is not supported. 🚫
• rcpc — enable the RCpc extension. This enables the use of the LDAPR instructions for load-acquire atomic semantics, and passes it on to the assembler, enabling inline asm statements to use instructions from the RCpc extension.
• dotprod — enable the Dot Product extension. This also enables Advanced SIMD instructions.
• aes — enable the Armv8-a aes and pmull crypto extension. This also enables Advanced SIMD instructions.
• sha2 — enable the Armv8-a sha2 crypto extension. This also enables Advanced SIMD instructions.
• sha3 — enable the sha512 and sha3 crypto extension. This also enables Advanced SIMD instructions. Use of this option with architectures prior to Armv8.2-A is not supported.
• sm4 — enable the sm3 and sm4 crypto extension. This also enables Advanced SIMD instructions. Use of this option with architectures prior to Armv8.2-A is not supported.
• profile — enable the Statistical Profiling extension. This option only enables the extension at the assembler level and does not affect code generation.
• rng — enable the Armv8.5-a Random Number instructions. This option only enables the extension at the assembler level and does not affect code generation.
• memtag — enable the Armv8.5-a Memory Tagging Extensions. Use of this option with architectures prior to Armv8.5-A is not supported.
• sb — enable the Armv8-a Speculation Barrier instruction. This option only enables the extension at the assembler level and does not affect code generation. This option is enabled by default for -march=armv8.5-a.
• ssbs — enable the Armv8-a Speculative Store Bypass Safe instruction. This option only enables the extension at the assembler level and does not affect code generation. This option is enabled by default for -march=armv8.5-a.
• predres — enable the Armv8-a Execution and Data Prediction Restriction instructions. This option only enables the extension at the assembler level and does not affect code generation. This option is enabled by default for -march=armv8.5-a.
• sve2 — enable the Armv8-a Scalable Vector Extension 2. This also enables SVE instructions.
• sve2-bitperm — enable SVE2 bitperm instructions. This also enables SVE2 instructions.
• sve2-sm4 — enable SVE2 sm4 instructions. This also enables SVE2 instructions.
• sve2-aes — enable SVE2 aes instructions. This also enables SVE2 instructions.
• sve2-sha3 — enable SVE2 sha3 instructions. This also enables SVE2 instructions.
• sve2p1 — enable SVE2.1 instructions. This also enables SVE2 instructions.
• tme — enable the Transactional Memory Extension.
• i8mm — enable 8-bit Integer Matrix Multiply instructions. This also enables Advanced SIMD and floating-point instructions. This option is enabled by default for -march=armv8.6-a. Use of this option with architectures prior to Armv8.2-A is not supported.
• f32mm — enable 32-bit Floating point Matrix Multiply instructions. This also enables SVE instructions. Use of this option with architectures prior to Armv8.2-A is not supported.
• f64mm — enable 64-bit Floating point Matrix Multiply instructions. This also enables SVE instructions. Use of this option with architectures prior to Armv8.2-A is not supported.
• bf16 — enable brain half-precision floating-point instructions. This also enables Advanced SIMD and floating-point instructions. This option is enabled by default for -march=armv8.6-a. Use of this option with architectures prior to Armv8.2-A is not supported.
• ls64 — enable the 64-byte atomic load and store instructions for accelerators.
• mops — enable the instructions that accelerate memory operations like memcpy, memmove, memset. This option is enabled by default for -march=armv8.8-a
• flagm — enable the Flag Manipulation instructions Extension.
• flagm2 — enable the FlagM2 flag conversion instructions.
• pauth — enable the Pointer Authentication Extension. 🔐
• cssc — turns on the Common Short Sequence Compression instructions 💫
• sme — turns on the Scalable Matrix Extension. Heads up bestie: this one only works when SVE2 is also enabled, that's the deal, no exceptions ✨
• sme-i16i64 — turns on the FEAT_SME_I16I64 extension to SME. It also switches on SME instructions while it's at it 🫶
• sme-f64f64 — turns on the FEAT_SME_F64F64 extension to SME. Yep, and this one enables SME instructions too 💅
• sme2 — turns on the Scalable Matrix Extension 2. Also enables SME instructions, because it's generous like that 👑
• sme-b16b16 — turns on the FEAT_SME_B16B16 extension to SME. This also enables SME2 and SVE_B16B16 instructions 🌈
• sme-f16f16 — turns on the FEAT_SME_F16F16 extension to SME. Also enables SME2 instructions, periodt ✨
• sme2p1 — turns on the Scalable Matrix Extension version 2.1. Also enables SME2 instructions 💜
• fcma — turns on the complex number SIMD extensions 🧚
• jscvt — turns on the "fjcvtzs" JavaScript conversion instruction 🔥
• frintts — turns on floating-point round to integral value instructions ✨
• wfxt — turns on "wfet" and "wfit" instructions 🎀
• xs — turns on the XS memory attribute extension 💫
• lse128 — turns on the LSE128 128-bit atomic instructions extension. This also enables LSE instructions, okay bestie? 🫶
• d128 — turns on support for 128-bit system register read/write instructions. This also enables the LSE128 extension 💅
• gcs — turns on support for Armv9.4-a Guarded Control Stack extension 🛡️
• the — turns on support for Armv8.9-a/9.4-a translation hardening extension 💜
• rcpc2 — turns on the RCpc2 extension ✨
• rcpc3 — turns on the RCpc3 (Release Consistency) extension 🌸
• fp8 — turns on the fp8 (8-bit floating point) extension 🍓
• fp8fma — turns on the fp8 (8-bit floating point) multiply accumulate extension 🔥
• ssve-fp8fma — turns on the fp8 (8-bit floating point) multiply accumulate extension in streaming mode 🌊
• fp8dot4 — turns on the fp8 (8-bit floating point) to single-precision 4-way dot product extension 💫
• ssve-fp8dot4 — turns on the fp8 (8-bit floating point) to single-precision 4-way dot product extension in streaming mode 🌊
• fp8dot2 — turns on the fp8 (8-bit floating point) to half-precision 2-way dot product extension ✨
• ssve-fp8dot2 — turns on the fp8 (8-bit floating point) to half-precision 2-way dot product extension in streaming mode 🌊
• faminmax — turns on the Floating Point Absolute Maximum/Minimum extension 📐
• lut — turns on the Lookup Table extension 🔍
• cpa — turns on the Checked Pointer Arithmetic instructions 🧮
• sve-b16b16 — turns on the SVE non-widening brain floating-point ("bf16") extension. Real talk: this only has an effect when "sve2" or "sme2" are also enabled, so don't come for me if nothing happens otherwise 🙈

Feature crypto implies aes, sha2, and simd, which implies fp. Conversely, nofp implies nosimd, which implies nocrypto, noaes and nosha2. 💖 That's the whole dependency chain bestie, follow the crumbs.

---

🎀 Adapteva Epiphany Options 🎀

These -m options are defined for Adapteva Epiphany, and yes we're listing all of them 💅

• -mhalf-reg-file — don't allocate any register in the range "r32"..."r63". Why would you ever want that? So your code can run on hardware variants that simply don't have those registers 🌸
• -mprefer-short-insn-regs — preferentially allocate registers that allow short instruction generation. But wait: this can result in an increased instruction count, so it may either reduce or increase overall code size. Chaos, honestly 😭
• -mbranch-cost=num — set the cost of branches to roughly num "simple" instructions. Not gonna lie: this cost is only a heuristic and is not guaranteed to produce consistent results across releases 🙈
• -mcmove — turn on the generation of conditional moves 👑
• -mnops=num — emit num NOPs before every other generated instruction. Every. Other. One. ✨
• -mno-soft-cmpsf — for single-precision floating-point comparisons, emit an "fsub" instruction and test the flags. This is faster than a software comparison, but it can get incorrect results in the presence of NaNs, or when two different small numbers are compared such that their difference is calculated as zero 💀 The default is -msoft-cmpsf, which uses slower, but IEEE-compliant, software comparisons.
• -mstack-offset=num — set the offset between the top of the stack and the stack pointer. Like, a value of 8 means that the eight bytes in the range "sp+0...sp+7" can be used by leaf functions without stack allocation. Values other than 8 or 16 are untested and unlikely to work, so don't try it 🫣 Also note that this option changes the ABI; compiling a program with a different stack offset than the libraries have been compiled with generally does not work. It can be useful if you want to evaluate whether a different stack offset would give you better code, but to actually use a different stack offset to build working programs, it is recommended to configure the toolchain with the appropriate --with-stack-offset=num option 💖
• -mno-round-nearest — make the scheduler assume that the rounding mode has been set to truncating. The default is -mround-nearest 🌸
• -mlong-calls — if not otherwise specified by an attribute, assume all calls might be beyond the offset range of the "b" / "bl" instructions, and therefore load the function address into a register before performing an (otherwise direct) call. This is the default 👑
• -mshort-calls — if not otherwise specified by an attribute, assume all direct calls are in the range of the "b" / "bl" instructions, so use these instructions for direct calls. The default is -mlong-calls ✨
• -msmall16 — assume addresses can be loaded as 16-bit unsigned values. This does not apply to function addresses for which -mlong-calls semantics are in effect 🫶
• -mfp-mode=mode — set the prevailing mode of the floating-point unit. This determines the floating-point mode that is provided and expected at function call and return time. Making this mode match the mode you predominantly need at function start can make your programs smaller and faster by avoiding unnecessary mode switches 💫

mode can be set to one the following values:

• caller — any mode at function entry is valid, and retained or restored when the function returns, and when it calls other functions. This mode is useful for compiling libraries or other compilation units you might want to incorporate into different programs with different prevailing FPU modes, and the convenience of being able to use a single object file outweighs the size and speed overhead for any extra mode switching that might be needed, compared with what would be needed with a more specific choice of prevailing FPU mode 🌈
• truncate — this is the mode used for floating-point calculations with truncating (i.e. round towards zero) rounding mode. That includes conversion from floating point to integer ✨
• round-nearest — this is the mode used for floating-point calculations with round-to-nearest-or-even rounding mode 🎀
• int — this is the mode used to perform integer calculations in the FPU, e.g. integer multiply, or integer multiply-and-accumulate 🧮

The default is -mfp-mode=caller 💜

• -mno-split-lohi, -mno-postinc, -mno-postmodify — code generation tweaks that disable, respectively, splitting of 32-bit loads, generation of post-increment addresses, and generation of post-modify addresses. The defaults are msplit-lohi, -mpost-inc, and -mpost-modify 💅
• -mnovect-double — change the preferred SIMD mode to SImode. The default is -mvect-double, which uses DImode as preferred SIMD mode 🌸
• -max-vect-align=num — the maximum alignment for SIMD vector mode types. num may be 4 or 8. The default is 8. Note: this is an ABI change, even though many library function interfaces are unaffected if they don't use SIMD vector modes in places that affect size and/or alignment of relevant types 🫣
• -msplit-vecmove-early — split vector moves into single word moves before reload. In theory this can give better register allocation, but so far the reverse seems to be generally the case 😭
• -m1reg-reg — specify a register to hold the constant -1, which makes loading small negative constants and certain bitmasks faster. Allowable values for reg are r43 and r63, which specify use of that register as a fixed register, and none, which means that no register is used for this purpose. The default is -m1reg-none ✨

---

🔥 AMD GCN Options 🔥

These options are defined specifically for the AMD GCN port, and they're giving very much "pick your GPU destiny" energy 👑

• -march=gpu, -mtune=gpu — set architecture type or tuning for gpu. Supported values for gpu are:
  • gfx900 — compile for GCN5 Vega 10 devices (gfx900) 💫
  • gfx902 — compile for GCN5 Vega gfx902 devices. (Experimental) 🧪
  • gfx904 — compile for GCN5 Vega gfx904 devices. (Experimental) 🧪
  • gfx906 — compile for GCN5 Vega 20 devices (gfx906) 🌸
  • gfx908 — compile for CDNA1 Instinct MI100 series devices (gfx908) 👑
  • gfx909 — compile for GCN5 Vega gfx909 devices. (Experimental) 🧪
  • gfx90a — compile for CDNA2 Instinct MI200 series devices (gfx90a) 🔥
  • gfx90c — compile for GCN5 Vega 7 devices (gfx90c) ✨
  • gfx9-generic — compile generic code for Vega devices, executable on the following subset of GFX9 devices: gfx900, gfx902, gfx904, gfx906, gfx909 and gfx90c. (Experimental) 🧪
  • gfx1030 — compile for RDNA2 gfx1030 devices (GFX10 series) 💜
  • gfx1031 — compile for RDNA2 gfx1031 devices (GFX10 series). (Experimental) 🧪
  • gfx1032 — compile for RDNA2 gfx1032 devices (GFX10 series). (Experimental) 🧪
  • gfx1033 — compile for RDNA2 gfx1033 devices (GFX10 series). (Experimental) 🧪
  • gfx1034 — compile for RDNA2 gfx1034 devices (GFX10 series). (Experimental) 🧪
  • gfx1035 — compile for RDNA2 gfx1035 devices (GFX10 series). (Experimental) 🧪
  • gfx1036 — compile for RDNA2 gfx1036 devices (GFX10 series) ✨
  • gfx10-3-generic — compile generic code for GFX10-3 devices, executable on gfx1030, gfx1031, gfx1032, gfx1033, gfx1034, gfx1035, and gfx1036. (Experimental) 🧪
  • gfx1100 — compile for RDNA3 gfx1100 devices (GFX11 series) 👑
  • gfx1101 — compile for RDNA3 gfx1101 devices (GFX11 series). (Experimental) 🧪
  • gfx1102 — compile for RDNA3 gfx1102 devices (GFX11 series). (Experimental) 🧪
  • gfx1103 — compile for RDNA3 gfx1103 devices (GFX11 series) 🌈
• gfx1150 — compile for RDNA3 gfx1150 devices (GFX11 series). (Experimental) 🧪✨
• gfx1151 — compile for RDNA3 gfx1151 devices (GFX11 series). (Experimental) 🧪
• gfx1152 — compile for RDNA3 gfx1152 devices (GFX11 series). (Experimental) 🧪
• gfx1153 — compile for RDNA3 gfx1153 devices (GFX11 series). (Experimental) 🧪
• gfx11-generic — compile generic code for GFX11 devices, and it'll run on gfx1100, gfx1101, gfx1102, gfx1103, gfx1150, gfx1151, gfx1152, and gfx1153. (Experimental) 🧪🌈
• -msram-ecc=on, -msram-ecc=off, -msram-ecc=any — compile binaries for devices where the SRAM-ECC feature is enabled, disabled, or either mode. Bestie listen, on some devices this can be flipped per-process, and your compiled code has to match the device mode — no vibes-based guessing here. The default is any, for devices that support it. 💅
• -mstack-size=bytes — tell it how many bytes of stack space to request for each GPU thread (wave-front). ⚠️ Beware: there may be a lot of threads and limited memory to go around, so be normal about it. The size of the stack allocation can also mess with run-time performance. Default is 32KB when you're using OpenACC or OpenMP, and 1MB otherwise. 😭
• -mxnack=on, -mxnack=off, -mxnack=any — compile binaries for devices with the XNACK feature enabled, disabled, or either mode. Some devices always require XNACK, some let the user configure it. The compiled code must match the device mode. Default is -mxnack=any on devices that support Unified Shared Memory, and -mxnack=no otherwise. 👑

---

🎀 ARC Options 🎀

The following options control the architecture variant for which code is being compiled, iykyk:

• -mbarrel-shifter — generate instructions supported by the barrel shifter. This is the default unless -mcpu=ARC601 or -mcpu=ARCEM is in effect. 🌀
• -mjli-always — force calling a function using the jli_s instruction. Only valid for ARCv2 architecture, no exceptions bestie. 🫶
• -mcpu=cpu — sets the architecture type, register usage, and instruction scheduling parameters for cpu. There are also shortcut alias options available for backward compatibility and convenience ✨. Supported values for cpu are:
  • arc600 — compile for ARC600. Aliases: -mA6, -mARC600.
  • arc601 — compile for ARC601. Alias: -mARC601.
  • arc700 — compile for ARC700. Aliases: -mA7, -mARC700. This is the default when configured with --with-cpu=arc700.
  • arcem — compile for ARC EM.
  • archs — compile for ARC HS.
  • em — compile for ARC EM CPU with no hardware extensions.
  • em4 — compile for ARC EM4 CPU.
  • em4_dmips — compile for ARC EM4 DMIPS CPU.
  • em4_fpus — compile for ARC EM4 DMIPS CPU with the single-precision floating-point extension.
  • em4_fpuda — compile for ARC EM4 DMIPS CPU with single-precision floating-point and double assist instructions.
  • hs — compile for ARC HS CPU with no hardware extensions except the atomic instructions.
  • hs34 — compile for ARC HS34 CPU.
  • hs38 — compile for ARC HS38 CPU.
  • hs38_linux — compile for ARC HS38 CPU with all hardware extensions on.
  • hs4x — compile for ARC HS4x CPU.
  • hs4xd — compile for ARC HS4xD CPU.
  • hs4x_rel31 — compile for ARC HS4x CPU release 3.10a.
  • arc600_norm — compile for ARC 600 CPU with "norm" instructions enabled.
  • arc600_mul32x16 — compile for ARC 600 CPU with "norm" and 32x16-bit multiply instructions enabled.
  • arc600_mul64 — compile for ARC 600 CPU with "norm" and "mul64"-family instructions enabled.
  • arc601_norm — compile for ARC 601 CPU with "norm" instructions enabled.
  • arc601_mul32x16 — compile for ARC 601 CPU with "norm" and 32x16-bit multiply instructions enabled.
  • arc601_mul64 — compile for ARC 601 CPU with "norm" and "mul64"-family instructions enabled.
  • nps400 — compile for ARC 700 on NPS400 chip.
  • em_mini — compile for ARC EM minimalist configuration featuring reduced register set.
• -mdpfp, -mdpfp-compact — generate double-precision FPX instructions, tuned for the compact implementation. 🍓
• -mdpfp-fast — generate double-precision FPX instructions, tuned for the fast implementation. 🔥
• -mno-dpfp-lrsr — disable lr and sr instructions from using FPX extension aux registers. Ngl kind of niche but okay. 🙈
• -mea — generate extended arithmetic instructions. Currently only divaw, adds, subs, and sat16 are supported. Only valid for -mcpu=ARC700. ✨
• -mno-mpy — do not generate mpy-family instructions for ARC700. Heads up: this option is deprecated 💀.
• -mmul32x16 — generate 32x16-bit multiply and multiply-accumulate instructions.
• -mmul64 — generate mul64 and mulu64 instructions. Only valid for -mcpu=ARC600.
• -mnorm — generate "norm" instructions. This is the default if -mcpu=ARC700 is in effect.
• -mspfp, -mspfp-compact — generate single-precision FPX instructions, tuned for the compact implementation.
• -mspfp-fast — generate single-precision FPX instructions, tuned for the fast implementation. 🔥
• -msimd — enable generation of ARC SIMD instructions via target-specific builtins. Only valid for -mcpu=ARC700.
• -msoft-float — girl, this option is ignored; it's provided for compatibility purposes only. Software floating-point code is emitted by default, and that default can be overridden by the FPX options: -mspfp, -mspfp-compact, or -mspfp-fast for single precision, and -mdpfp, -mdpfp-compact, or -mdpfp-fast for double precision. 💅
• -mswap — generate "swap" instructions.
• -matomic — this enables use of the locked load/store conditional extension to implement atomic memory built-in functions. Not available for ARC 6xx or ARC EM cores. Sorry bestie. 😭
• -mdiv-rem — enable div and rem instructions for ARCv2 cores.
• -mcode-density — enable code density instructions for ARC EM. This option is on by default for ARC HS. 🎀
• -mll64 — enable double load/store operations for ARC HS cores.
• -mtp-regno=regno — specify the thread pointer register number.
• -mmpy-option=multo — compile ARCv2 code with a multiplier design option. You can specify it with either a string or a numeric value for multo. wlh1 is the default value. The recognized values are:
  • 0 / none — no multiplier available.
  • 1 / w — 16x16 multiplier, fully pipelined. The following instructions are enabled: mpyw and mpyuw.
  • 2 / wlh1 — 32x32 multiplier, fully pipelined (1 stage). The following instructions are additionally enabled: mpy, mpyu, mpym, mpymu, and mpy_s.
  • 3 / wlh2 — 32x32 multiplier, fully pipelined (2 stages). The following instructions are additionally enabled: mpy, mpyu, mpym, mpymu, and mpy_s.
  • 4 / wlh3 — two 16x16 multipliers, blocking, sequential. The following instructions are additionally enabled: mpy, mpyu, mpym, mpymu, and mpy_s.
  • 5 / wlh4 — one 16x16 multiplier, blocking, sequential. The following instructions are additionally enabled: mpy, mpyu, mpym, mpymu, and mpy_s.
  • 6 / wlh5 — one 32x4 multiplier, blocking, sequential. The following instructions are additionally enabled: mpy, mpyu, mpym, mpymu, and mpy_s.
  • 7 / plus_dmpy — ARC HS SIMD support.
  • 8 / plus_macd — ARC HS SIMD support.
  • 9 / plus_qmacw — ARC HS SIMD support.

This option is only available for ARCv2 cores. 👑

• -mfpu=fpu — enables support for specific floating-point hardware extensions for ARCv2 cores. Supported values for fpu are:
  • fpus — enables support for single-precision floating-point hardware extensions.
  • fpud — enables support for double-precision floating-point hardware extensions. The single-precision floating-point extension is also enabled. Not available for ARC EM.
  • fpuda — enables support for double-precision floating-point hardware extensions using double-precision assist instructions. The single-precision floating-point extension is also enabled. This option is only available for ARC EM.
  • fpuda_div — enables support for double-precision floating-point hardware extensions using double-precision assist instructions. The single-precision floating-point, square-root, and divide extensions are also enabled. This option is only available for ARC EM.
  • fpuda_fma — enables support for double-precision floating-point hardware extensions using double-precision assist instructions. The single-precision floating-point and fused multiply and add hardware extensions are also enabled. This option is only available for ARC EM.
  • fpuda_all — enables support for double-precision floating-point hardware extensions using double-precision assist instructions. All single-precision floating-point hardware extensions are also enabled. This option is only available for ARC EM.
  • fpus_div — enables support for single-precision floating-point, square-root and divide hardware extensions.
  • fpud_div — enables support for double-precision floating-point, square-root and divide hardware extensions. This option includes option fpus_div. Not available for ARC EM.
  • fpus_fma — enables support for single-precision floating-point and fused multiply and add hardware extensions.
  • fpud_fma — enables support for double-precision floating-point and fused multiply and add hardware extensions. This option includes option fpus_fma. Not available for ARC EM.
  • fpus_all — enables support for all single-precision floating-point hardware extensions.
  • fpud_all — enables support for all single- and double-precision floating-point hardware extensions. Not available for ARC EM.
• -mirq-ctrl-saved=register-range, blink, lp_count — ok bestie, this one tells the processor which general-purpose registers it automatically saves/restores on interrupt entry and exit. 💅 register-range is two registers separated by a dash. The range always starts with "r0", and the upper limit is the "fp" register. blink and lp_count are optional. This option is only valid for ARC EM and ARC HS cores. ✨
• -mrgf-banked-regs=number — says how many registers get replicated in the second register bank when a fast interrupt enters. Fast interrupts are interrupts with the highest priority level P0. These interrupts save only PC and STATUS32 registers to avoid memory transactions during interrupt entry and exit sequences. Use this when you're using fast interrupts in an ARC V2 family processor. Permitted values are 4, 8, 16, and 32. 🦋
• -mlpc-width=width — sets the width of the "lp_count" register. Valid values for width are 8, 16, 20, 24, 28, and 32 bits. The default width is fixed to 32 bits. If the width is less than 32, the compiler does not attempt to transform loops in your program to use the zero-delay loop mechanism unless it is known that the "lp_count" register can hold the required loop-counter value. Depending on the width specified, the compiler and run-time library might continue to use the loop mechanism for various needs. This option defines macro "__ARC_LPC_WIDTH__" with the value of width. 😌
• -mrf16 — tells the compiler to generate code for a 16-entry register file. This option defines the "__ARC_RF16__" preprocessor macro. 🎀
• -mbranch-index — enables use of "bi" or "bih" instructions to implement jump tables. 🔥

The following options are passed through to the assembler, and also define preprocessor macro symbols.

• -mdsp-packa — passed down to the assembler to enable the DSP Pack A extensions. Also sets the preprocessor symbol "__Xdsp_packa". This option is deprecated. 🚨
• -mdvbf — passed down to the assembler to enable the dual Viterbi butterfly extension. Also sets the preprocessor symbol "__Xdvbf". This option is deprecated. 🚨
• -mlock — passed down to the assembler to enable the locked load/store conditional extension. Also sets the preprocessor symbol "__Xlock". 🔒
• -mmac-d16 — passed down to the assembler. Also sets the preprocessor symbol "__Xxmac_d16". This option is deprecated. 🚨
• -mmac-24 — passed down to the assembler. Also sets the preprocessor symbol "__Xxmac_24". This option is deprecated. 🚨
• -mrtsc — passed down to the assembler to enable the 64-bit time-stamp counter extension instruction. Also sets the preprocessor symbol "__Xrtsc". This option is deprecated. 🚨
• -mswape — passed down to the assembler to enable the swap byte ordering extension instruction. Also sets the preprocessor symbol "__Xswape". 🔄
• -mtelephony — passed down to the assembler to enable dual- and single-operand instructions for telephony. Also sets the preprocessor symbol "__Xtelephony". This option is deprecated. 🚨
• -mxy — passed down to the assembler to enable the XY memory extension. Also sets the preprocessor symbol "__Xxy". 🌈

The following options control how the assembly code is annotated:

• -misize — annotate assembler instructions with estimated addresses. ✨
• -mannotate-align — does nothing. Preserved for backward compatibility. 💀

The following options are passed through to the linker:

• -marclinux — passed through to the linker, to specify use of the "arclinux" emulation. This option is enabled by default in tool chains built for "arc-linux-uclibc" and "arceb-linux-uclibc" targets when profiling is not requested. 🐧
• -marclinux_prof — passed through to the linker, to specify use of the "arclinux_prof" emulation. This option is enabled by default in tool chains built for "arc-linux-uclibc" and "arceb-linux-uclibc" targets when profiling is requested. 📊

The following options control the semantics of generated code:

• -mlong-calls — generate calls as register indirect calls, thus providing access to the full 32-bit address range. 🔥
• -mmedium-calls — don't use less than 25-bit addressing range for calls, which is the offset available for an unconditional branch-and-link instruction. Conditional execution of function calls is suppressed, to allow use of the 25-bit range, rather than the 21-bit range with conditional branch-and-link. This is the default for tool chains built for "arc-linux-uclibc" and "arceb-linux-uclibc" targets. 💅
• -G num — put definitions of externally-visible data in a small data section if that data is no bigger than num bytes. The default value of num is 4 for any ARC configuration, or 8 when we have double load/store operations. 🍓
• -mno-sdata — do not generate sdata references. This is the default for tool chains built for "arc-linux-uclibc" and "arceb-linux-uclibc" targets. 🚫
• -mvolatile-cache — use ordinarily cached memory accesses for volatile references. This is the default. ✅
• -mno-volatile-cache — enable cache bypass for volatile references. 🧊

The following options fine tune code generation:

• -malign-call — does nothing. Preserved for backward compatibility. 💀
• -mauto-modify-reg — enable the use of pre/post modify with register displacement. 🔧
• -mbbit-peephole — does nothing. Preserved for backward compatibility. 💀
• -mno-brcc — this option disables a target-specific pass in arc_reorg to generate compare-and-branch ("brcc") instructions. It has no effect on generation of these instructions driven by the combiner pass. 🚫
• -mcase-vector-pcrel — use PC-relative switch case tables to enable case table shortening. This is the default for -Os. 📉
• -mcompact-casesi — enable compact "casesi" pattern. This is the default for -Os, and only available for ARCv1 cores. This option is deprecated. 🚨
• -mno-cond-exec — disable the ARCompact-specific pass to generate conditional execution instructions.

Due to delay slot scheduling and interactions between operand numbers, literal sizes, instruction lengths, and the support for conditional execution, the target-independent pass to generate conditional execution is often lacking, so the ARC port has kept a special pass around that tries to find more conditional execution generation opportunities after register allocation, branch shortening, and delay slot scheduling have been done. This pass generally, but not always, improves performance and code size, at the cost of extra compilation time, which is why there is an option to switch it off. If you have a problem with call instructions exceeding their allowable offset range because they are conditionalized, you should consider using -mmedium-calls instead. 😭

• -mearly-cbranchsi — enable pre-reload use of the "cbranchsi" pattern. ⚡
• -mexpand-adddi — expand "adddi3" and "subdi3" at RTL generation time into "add.f", "adc" etc. This option is deprecated. 🚨
• -mindexed-loads — enable the use of indexed loads. This can be problematic because some optimizers then assume that indexed stores exist, which is not the case. ⚠️
• -mlra — does nothing. Preserved for backward compatibility. 💀
• -mlra-priority-none — don't indicate any priority for target registers. 🚫
• -mlra-priority-compact — indicate target register priority for r0..r3 / r12..r15. 🎯
• -mlra-priority-noncompact — reduce target register priority for r0..r3 / r12..r15. 📉
• -mmillicode — when optimizing for size (using -Os), prologues and epilogues that have to save or restore a large number of registers are often shortened by using call to a special function in libgcc; this is referred to as a millicode call. As these calls can pose performance issues, and/or cause linking issues when linking in a nonstandard way, this option is provided to turn on or off millicode call generation. 🧚
• -mcode-density-frame — this option enables the compiler to emit "enter" and "leave" instructions. These instructions are only valid for CPUs with code-density feature. 🏗️
• -mmixed-code — does nothing. Preserved for backward compatibility. 💀
• -mq-class — This option is deprecated. Enable q instruction alternatives. This is the default for -Os. 🚨
• -mRcq — does nothing. Preserved for backward compatibility. 💀
• -mRcw — does nothing. Preserved for backward compatibility. 💀
• -msize-level=level — fine-tune size optimization with regards to instruction lengths and alignment. The recognized values for level are:
  • 0 — No size optimization. This level is deprecated and treated like 1.
  • 1 — Short instructions are used opportunistically.
  • 2 — In addition, alignment of loops and of code after barriers are dropped.
  • 3 — In addition, optional data alignment is dropped, and the option Os is enabled. This defaults to 3 when -Os is in effect. Otherwise, the behavior when this is not set is equivalent to level 1. 📏
• -mtune=cpu — set instruction scheduling parameters for cpu, overriding any implied by -mcpu=. Supported values for cpu are:
  • ARC600 — Tune for ARC600 CPU.
  • ARC601 — Tune for ARC601 CPU.
  • ARC700 — Tune for ARC700 CPU with standard multiplier block.
  • ARC700-xmac — Tune for ARC700 CPU with XMAC block.
  • ARC725D — Tune for ARC725D CPU.
  • ARC750D — Tune for ARC750D CPU.
  • core3 — Tune for ARCv2 core3 type CPU. This option enable usage of "dbnz" instruction.
  • release31a — Tune for ARC4x release 3.10a. 👑
• -mmultcost=num — cost to assume for a multiply instruction, with 4 being equal to a normal instruction. ✖️
• -munalign-prob-threshold=probability — does nothing. Preserved for backward compatibility. 💀

The following options are maintained for backward compatibility, but are now deprecated and will be removed in a future release:

• -margonaut — obsolete FPX. 💀
• -mbig-endian, -EB — compile code for big-endian targets. Use of these options is now deprecated. Big-endian code is supported by configuring GCC to build "arceb-elf32" and "arceb-linux-uclibc" targets, for which big endian is the default. 🌍
• -mlittle-endian, -EL — bestie, this compiles your code for little-endian targets. BUT — and I need you to hear me — these options are deprecated now 💀. Little-endian code is supported by configuring GCC to build arc-elf32 and arc-linux-uclibc targets, and for those, little endian is just the default. So like, it's fine, you don't need to ask. 🫶
• -mbarrel_shifter — replaced by -mbarrel-shifter. That's it, that's the whole thing, the underscore did NOT make it. 🙈
• -mdpfp_compact — replaced by -mdpfp-compact.
• -mdpfp_fast — replaced by -mdpfp-fast.
• -mdsp_packa — replaced by -mdsp-packa.
• -mEA — replaced by -mea. case change, iconic, moving on ✨
• -mmac_24 — replaced by -mmac-24.
• -mmac_d16 — replaced by -mmac-d16.
• -mspfp_compact — replaced by -mspfp-compact.
• -mspfp_fast — replaced by -mspfp-fast.
• *-mtune=cpu* — okay so the values arc600, arc601, arc700 and arc700-xmac for cpu are replaced by ARC600, ARC601, ARC700 and ARC700-xmac respectively. lowercase to uppercase, glow-up complete 💅
• *-multcost=num* — replaced by -mmultcost.

---

🍎 ARM Options 🍎

These -m options are defined for the ARM port, and there are so many, buckle up bestie:

• *-mabi=name* — generates code for the specified ABI. Permissible values are: apcs-gnu, atpcs, aapcs, aapcs-linux and iwmmxt.
• -mapcs-frame — generates a stack frame that's compliant with the ARM Procedure Call Standard for all functions, even when it's not strictly necessary for the code to run correctly. If you specify -fomit-frame-pointer alongside this option, then stack frames won't be generated for leaf functions. The default is -mno-apcs-frame. Also? This option is deprecated. 👀
• -mapcs — this is a synonym for -mapcs-frame, and yes, it's also deprecated.
• -mthumb-interwork — generates code that supports calling between the ARM and Thumb instruction sets. Without this option, on pre-v5 architectures, the two instruction sets can't be reliably used inside one program. The default is -mno-thumb-interwork, because specifying -mthumb-interwork makes slightly larger code. In AAPCS configurations this option is meaningless.
• -mno-sched-prolog — prevents the reordering of instructions in the function prologue, and prevents merging those instructions with the instructions in the function's body. This means all functions start with a recognizable set of instructions (or actually one of a choice from a small set of different function prologues), and that info can be used to locate the start of functions inside an executable piece of code. The default is -msched-prolog.
• *-mfloat-abi=name* — specifies which floating-point ABI to use. Permissible values are: soft, softfp and hard.

Specifying soft makes GCC generate output containing library calls for floating-point operations. softfp allows the generation of code using hardware floating-point instructions, but still uses the soft-float calling conventions. hard allows generation of floating-point instructions and uses FPU-specific calling conventions.

The default depends on the specific target configuration. Note that the hard-float and soft-float ABIs are not link-compatible; you must compile your entire program with the same ABI, and link with a compatible set of libraries. no mixing, bestie, this is not a smoothie 🥤

• -mgeneral-regs-only — generates code that uses only the general-purpose registers. This will stop the compiler from using floating-point and Advanced SIMD registers, but it will not impose any restrictions on the assembler. (the assembler stays unbothered, unbothered, moisturised, in its lane 😌)
• -mlittle-endian — generates code for a processor running in little-endian mode. This is the default for all standard configurations.
• -mbig-endian — generates code for a processor running in big-endian mode; the default is to compile code for a little-endian processor.
• -mbe8
• -mbe32 — when linking a big-endian image, select between BE8 and BE32 formats. This option has no effect for little-endian images and is ignored. The default depends on the selected target architecture. For ARMv6 and later architectures the default is BE8, for older architectures the default is BE32. BE32 format has been deprecated by ARM.
• -march=name[+extension...] — this specifies the name of the target ARM architecture. GCC uses this name to work out what kind of instructions it can emit when generating assembly code. This option can be used together with, or instead of, the -mcpu= option.

Permissible names are: armv4t, armv5t, armv5te, armv6, armv6j, armv6k, armv6kz, armv6t2, armv6z, armv6zk, armv7, armv7-a, armv7ve, armv8-a, armv8.1-a, armv8.2-a, armv8.3-a, armv8.4-a, armv8.5-a, armv8.6-a, armv9-a, armv7-r, armv8-r, armv6-m, armv6s-m, armv7-m, armv7e-m, armv8-m.base, armv8-m.main, armv8.1-m.main, iwmmxt and iwmmxt2.

Also, the following architectures, which lack support for the Thumb execution state, are recognized — but support is deprecated: armv4.

A lot of these architectures support extensions. You add them by appending +extension to the architecture name. Extension options are processed in order and capabilities accumulate. An extension will also enable any necessary base extensions that it depends on — for example, the +crypto extension will always enable the +simd extension. The one exception to this additive construction is for extensions prefixed with +no...: those extensions disable the specified option and any other extensions that may depend on that extension being present.

So for example, -march=armv7-a+simd+nofp+vfpv4 is equivalent to writing -march=armv7-a+vfpv4, since the +simd option is entirely disabled by the +nofp option that follows it.

Most extension names are generically named, but their effect depends on the architecture they're applied to. For example, +simd can be applied to both armv7-a and armv8-a architectures, but it'll enable the original ARMv7-A Advanced SIMD (Neon) extensions for armv7-a, and the ARMv8-A variant for armv8-a.

The table below lists the supported extensions for each architecture. Architectures not mentioned don't support any extensions. 📋✨

armv5te, armv6, armv6j, armv6k, armv6kz, armv6t2, armv6z, armv6zk

• +fp — the VFPv2 floating-point instructions. The extension +vfpv2 can be used as an alias for this extension.
• +nofp — disable the floating-point instructions.

armv7

• The common subset of the ARMv7-A, ARMv7-R and ARMv7-M architectures.
• +fp — the VFPv3 floating-point instructions, with 16 double-precision registers. The extension +vfpv3-d16 can be used as an alias for this extension. Note that floating-point is not supported by the base ARMv7-M architecture, but is compatible with both the ARMv7-A and ARMv7-R architectures.
• +nofp — disable the floating-point instructions.

armv7-a

• +mp — the multiprocessing extension.
• +sec — the security extension.
• +fp — the VFPv3 floating-point instructions, with 16 double-precision registers. The extension +vfpv3-d16 can be used as an alias for this extension.
• +simd — the Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions. The extensions +neon and +neon-vfpv3 can be used as aliases for this extension.
• +vfpv3 — the VFPv3 floating-point instructions, with 32 double-precision registers.
• +vfpv3-d16-fp16 — the VFPv3 floating-point instructions, with 16 double-precision registers and the half-precision floating-point conversion operations.
• +vfpv3-fp16 — the VFPv3 floating-point instructions, with 32 double-precision registers and the half-precision floating-point conversion operations.
• +vfpv4-d16 — the VFPv4 floating-point instructions, with 16 double-precision registers.
• +vfpv4 — the VFPv4 floating-point instructions, with 32 double-precision registers.
• +neon-fp16 — the Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions, with the half-precision floating-point conversion operations.
• +neon-vfpv4 — the Advanced SIMD (Neon) v2 and the VFPv4 floating-point instructions.
• +nosimd — disable the Advanced SIMD instructions (this does not disable floating point).
• +nofp — disable the floating-point and Advanced SIMD instructions.

armv7ve

• The extended version of the ARMv7-A architecture with support for virtualization.
• +fp — the VFPv4 floating-point instructions, with 16 double-precision registers. The extension +vfpv4-d16 can be used as an alias for this extension.
• +simd — the Advanced SIMD (Neon) v2 and the VFPv4 floating-point instructions. The extension +neon-vfpv4 can be used as an alias for this extension.
• +vfpv3-d16 — the VFPv3 floating-point instructions, with 16 double-precision registers.
• +vfpv3 — the VFPv3 floating-point instructions, with 32 double-precision registers.
• +vfpv3-d16-fp16 — the VFPv3 floating-point instructions, with 16 double-precision registers and the half-precision floating-point conversion operations.
• +vfpv3-fp16 — the VFPv3 floating-point instructions, with 32 double-precision registers and the half-precision floating-point conversion operations.
• +vfpv4-d16 — the VFPv4 floating-point instructions, with 16 double-precision registers.
• +vfpv4 — the VFPv4 floating-point instructions, with 32 double-precision registers.
• +neon — the Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions. The extension +neon-vfpv3 can be used as an alias for this extension.
• +neon-fp16 — the Advanced SIMD (Neon) v1 and the VFPv3 floating-point instructions, with the half-precision floating-point conversion operations.
• +nosimd — turns off the Advanced SIMD instructions. and no bestie, it does not turn off floating point, those are two separate things 💅
• +nofp — turns off the floating-point and Advanced SIMD instructions. both. at once. iconic.

🦋 armv8-a

• +crc — the Cyclic Redundancy Check (CRC) instructions ✨
• +simd — the ARMv8-A Advanced SIMD and floating-point instructions.
• +crypto — the cryptographic instructions 🔐
• +nocrypto — disables the cryptographic instructions.
• +nofp — disables the floating-point, Advanced SIMD and cryptographic instructions. all three, gone, periodt.
• +sb — Speculation Barrier Instruction.
• +predres — Execution and Data Prediction Restriction Instructions.

🌸 armv8.1-a

• +simd — the ARMv8.1-A Advanced SIMD and floating-point instructions.
• +crypto — the cryptographic instructions. heads up: this also enables the Advanced SIMD and floating-point instructions 🫶
• +nocrypto — disables the cryptographic instructions.
• +nofp — disables the floating-point, Advanced SIMD and cryptographic instructions.
• +sb — Speculation Barrier Instruction.
• +predres — Execution and Data Prediction Restriction Instructions.

🎀 armv8.2-a and armv8.3-a (they share the same extension list, besties fr fr)

• +fp16 — the half-precision floating-point data processing instructions. this also enables the Advanced SIMD and floating-point instructions.
• +fp16fml — the half-precision floating-point fmla extension. this also enables the half-precision floating-point extension and the Advanced SIMD and floating-point instructions.
• +simd — the ARMv8.1-A Advanced SIMD and floating-point instructions.
• +crypto — the cryptographic instructions. also enables the Advanced SIMD and floating-point instructions ✨
• +dotprod — enables the Dot Product extension. this also enables Advanced SIMD instructions.
• +nocrypto — disables the cryptographic extension.
• +nofp — disables the floating-point, Advanced SIMD and cryptographic instructions.
• +sb — Speculation Barrier Instruction.
• +predres — Execution and Data Prediction Restriction Instructions.
• +i8mm — 8-bit Integer Matrix Multiply instructions. also enables Advanced SIMD and floating-point instructions.
• +bf16 — Brain half-precision floating-point instructions. also enables Advanced SIMD and floating-point instructions.

🌈 armv8.4-a

• +fp16 — the half-precision floating-point data processing instructions. this also enables the Advanced SIMD and floating-point instructions as well as the Dot Product extension and the half-precision floating-point fmla extension.
• +simd — the ARMv8.3-A Advanced SIMD and floating-point instructions as well as the Dot Product extension.
• +crypto — the cryptographic instructions. this also enables the Advanced SIMD and floating-point instructions as well as the Dot Product extension.
• +nocrypto — disables the cryptographic extension.
• +nofp — disables the floating-point, Advanced SIMD and cryptographic instructions.
• +sb — Speculation Barrier Instruction.
• +predres — Execution and Data Prediction Restriction Instructions.
• +i8mm — 8-bit Integer Matrix Multiply instructions. also enables Advanced SIMD and floating-point instructions.
• +bf16 — Brain half-precision floating-point instructions. also enables Advanced SIMD and floating-point instructions.

💜 armv8.5-a

• +fp16 — the half-precision floating-point data processing instructions. this also enables the Advanced SIMD and floating-point instructions as well as the Dot Product extension and the half-precision floating-point fmla extension.
• +simd — the ARMv8.3-A Advanced SIMD and floating-point instructions as well as the Dot Product extension.
• +crypto — the cryptographic instructions. this also enables the Advanced SIMD and floating-point instructions as well as the Dot Product extension.
• +nocrypto — disables the cryptographic extension.
• +nofp — disables the floating-point, Advanced SIMD and cryptographic instructions.
• +i8mm — 8-bit Integer Matrix Multiply instructions. also enables Advanced SIMD and floating-point instructions.
• +bf16 — Brain half-precision floating-point instructions. also enables Advanced SIMD and floating-point instructions.

⭐ armv8.6-a

• +fp16 — the half-precision floating-point data processing instructions. this also enables the Advanced SIMD and floating-point instructions as well as the Dot Product extension and the half-precision floating-point fmla extension.
• +simd — the ARMv8.3-A Advanced SIMD and floating-point instructions as well as the Dot Product extension.
• +crypto — the cryptographic instructions. this also enables the Advanced SIMD and floating-point instructions as well as the Dot Product extension.
• +nocrypto — disables the cryptographic extension.
• +nofp — disables the floating-point, Advanced SIMD and cryptographic instructions.
• +i8mm — 8-bit Integer Matrix Multiply instructions. also enables Advanced SIMD and floating-point instructions.
• +bf16 — Brain half-precision floating-point instructions. also enables Advanced SIMD and floating-point instructions.

🍓 armv7-r

• +fp.sp — the single-precision VFPv3 floating-point instructions. fun fact: the extension +vfpv3xd can be used as an alias for this one.
• +fp — the VFPv3 floating-point instructions with 16 double-precision registers. and yeah, +vfpv3-d16 can be used as an alias for this extension too.
• +vfpv3xd-d16-fp16 — the single-precision VFPv3 floating-point instructions with 16 double-precision registers and the half-precision floating-point conversion operations.
• +vfpv3-d16-fp16 — the VFPv3 floating-point instructions with 16 double-precision registers and the half-precision floating-point conversion operations.
• +nofp — disables the floating-point extension.
• +idiv — the ARM-state integer division instructions.
• +noidiv — disables the ARM-state integer division extension.

🧚 armv7e-m

• +fp — the single-precision VFPv4 floating-point instructions.
• +fpv5 — the single-precision FPv5 floating-point instructions.
• +fp.dp — the single- and double-precision FPv5 floating-point instructions.
• +nofp — disables the floating-point extensions.

💖 armv8.1-m.main

• +dsp — the DSP instructions.
• +mve — the M-Profile Vector Extension (MVE) integer instructions.
• +mve.fp — the M-Profile Vector Extension (MVE) integer and single precision floating-point instructions.
• +fp — the single-precision floating-point instructions.
• +fp.dp — the single- and double-precision floating-point instructions.
• +nofp — disables the floating-point extension.
• +cdecp0, +cdecp1, ... , +cdecp7 — enables the Custom Datapath Extension (CDE) on selected coprocessors according to the numbers given in the options, in the range 0 to 7.
• +pacbti — enables the Pointer Authentication and Branch Target Identification Extension.

🌟 armv8-m.main

• +dsp — the DSP instructions.
• +nodsp — disables the DSP extension.
• +fp — the single-precision floating-point instructions.
• +fp.dp — the single- and double-precision floating-point instructions.
• +nofp — disables the floating-point extension.
• +cdecp0, +cdecp1, ... , +cdecp7 — enables the Custom Datapath Extension (CDE) on selected coprocessors according to the numbers given in the options, in the range 0 to 7.

👑 armv8-r

• +crc — the Cyclic Redundancy Check (CRC) instructions.
• +fp.sp — the single-precision FPv5 floating-point instructions.
• +simd — the ARMv8-A Advanced SIMD and floating-point instructions.
• +crypto — the cryptographic instructions.
• +nocrypto — disables the cryptographic instructions.
• +nofp — disables the floating-point, Advanced SIMD and cryptographic instructions.

---

-march=native makes the compiler auto-detect the architecture of the build computer. ngl, right now this feature is only supported on GNU/Linux, and not every architecture is recognized. if the auto-detect doesn't work out, the option just... has no effect. no drama, no warning, nothing.

-march=unset makes the compiler ignore any -march=... options that showed up earlier on the command line, and behave like the option was never passed at all. super useful for dodging warnings about conflicting CPU and architecture options when the two produce different architecture specifications 🫶

• -mtune=name — bestie, this one picks the ARM processor GCC should tune the code's performance for. Some ARM implementations just perform better with this set. The permissible names are: arm7tdmi, arm7tdmi-s, arm710t, arm720t, arm740t, strongarm, strongarm110, strongarm1100, strongarm1110, arm8, arm810, arm9, arm9e, arm920, arm920t, arm922t, arm946e-s, arm966e-s, arm968e-s, arm926ej-s, arm940t, arm9tdmi, arm10tdmi, arm1020t, arm1026ej-s, arm10e, arm1020e, arm1022e, arm1136j-s, arm1136jf-s, mpcore, mpcorenovfp, arm1156t2-s, arm1156t2f-s, arm1176jz-s, arm1176jzf-s, generic-armv7-a, cortex-a5, cortex-a7, cortex-a8, cortex-a9, cortex-a12, cortex-a15, cortex-a17, cortex-a32, cortex-a35, cortex-a53, cortex-a55, cortex-a57, cortex-a72, cortex-a73, cortex-a75, cortex-a76, cortex-a76ae, cortex-a77, cortex-a78, cortex-a78ae, cortex-a78c, cortex-a710, ares, cortex-r4, cortex-r4f, cortex-r5, cortex-r7, cortex-r8, cortex-r52, cortex-r52plus, cortex-m0, cortex-m0plus, cortex-m1, cortex-m3, cortex-m4, cortex-m7, cortex-m23, cortex-m33, cortex-m35p, cortex-m52, cortex-m55, cortex-m85, cortex-x1, cortex-x1c, cortex-m1.small-multiply, cortex-m0.small-multiply, cortex-m0plus.small-multiply, exynos-m1, marvell-pj4, neoverse-n1, neoverse-n2, neoverse-v1, xscale, iwmmxt, iwmmxt2, ep9312, fa526, fa626, fa606te, fa626te, fmp626, fa726te, star-mc1, xgene1.

Alsooo, this option can tell GCC to tune the code for a big.LITTLE system. Permissible names there are: cortex-a15.cortex-a7, cortex-a17.cortex-a7, cortex-a57.cortex-a53, cortex-a72.cortex-a53, cortex-a72.cortex-a35, cortex-a73.cortex-a53, cortex-a75.cortex-a55, cortex-a76.cortex-a55.

And -mtune=generic-arch says GCC should tune the performance for a blend of processors within architecture arch. The goal is code that runs well on the current most popular processors — balancing optimizations that help some CPUs in the range while dodging the performance pitfalls of the others. Heads up: the effects of this option may change in future GCC versions as CPU models come and go. ✨

-mtune permits the same extension options as -mcpu, but the extension options do not affect the tuning of the generated code.

-mtune=native makes the compiler auto-detect the CPU of the build computer. At present, this feature is only supported on GNU/Linux, and not all architectures are recognized. If the auto-detect is unsuccessful the option has no effect. 💅

• -mcpu=name[+extension...] — this specifies the name of the target ARM processor. GCC uses this name to derive the name of the target ARM architecture (as if you'd said -march) and the ARM processor type to tune for performance (as if you'd said -mtune). When this option is used together with -march or -mtune, those options take precedence over the appropriate part of this option. no cap.

Many of the supported CPUs implement optional architectural extensions. Where that's the case, the architectural extensions are normally enabled by default. If implementations that lack the extension exist, then the extension syntax can be used to disable those extensions that have been omitted. For floating-point and Advanced SIMD (Neon) instructions, the settings of the options -mfloat-abi and -mfpu must also be considered: floating-point and Advanced SIMD instructions will only be used if -mfloat-abi is not set to soft; and any setting of -mfpu other than auto will override the available floating-point and SIMD extension instructions.

For example, cortex-a9 can be found in three major configurations: integer only, with just a floating-point unit, or with floating-point and Advanced SIMD. The default is to enable all the instructions, but the extensions +nosimd and +nofp can be used to disable just the SIMD, or both the SIMD and floating-point instructions respectively.

Permissible names for this option are the same as those for -mtune. 👑

The following extension options are common to the listed CPUs:

• +nodsp — Disable the DSP instructions on cortex-m33, cortex-m35p, cortex-m52, cortex-m55 and cortex-m85. Also disable the M-Profile Vector Extension (MVE) integer and single precision floating-point instructions on cortex-m52, cortex-m55 and cortex-m85.
• +nopacbti — Disable the Pointer Authentication and Branch Target Identification Extension on cortex-m52 and cortex-m85.
• +nomve — Disable the M-Profile Vector Extension (MVE) integer and single precision floating-point instructions on cortex-m52, cortex-m55 and cortex-m85.
• +nomve.fp — Disable the M-Profile Vector Extension (MVE) single precision floating-point instructions on cortex-m52, cortex-m55 and cortex-m85.
• +cdecp0, +cdecp1, ... , +cdecp7 — Enable the Custom Datapath Extension (CDE) on selected coprocessors according to the numbers given in the options in the range 0 to 7 on cortex-m52, cortex-m55 and star-mc1.
• +nofp — Disables the floating-point instructions on arm9e, arm946e-s, arm966e-s, arm968e-s, arm10e, arm1020e, arm1022e, arm926ej-s, arm1026ej-s, cortex-r5, cortex-r7, cortex-r8, cortex-m4, cortex-m7, cortex-m33, cortex-m35p, cortex-m52, cortex-m55 and cortex-m85. Disables the floating-point and SIMD instructions on generic-armv7-a, cortex-a5, cortex-a7, cortex-a8, cortex-a9, cortex-a12, cortex-a15, cortex-a17, cortex-a15.cortex-a7, cortex-a17.cortex-a7, cortex-a32, cortex-a35, cortex-a53 and cortex-a55.
• +nofp.dp — Disables the double-precision component of the floating-point instructions on cortex-r5, cortex-r7, cortex-r8, cortex-r52, cortex-r52plus and cortex-m7.
• +nosimd — Disables the SIMD (but not floating-point) instructions on generic-armv7-a, cortex-a5, cortex-a7 and cortex-a9.
• +crypto — Enables the cryptographic instructions on cortex-a32, cortex-a35, cortex-a53, cortex-a55, cortex-a57, cortex-a72, cortex-a73, cortex-a75, exynos-m1, xgene1, cortex-a57.cortex-a53, cortex-a72.cortex-a53, cortex-a73.cortex-a35, cortex-a73.cortex-a53 and cortex-a75.cortex-a55.

Additionally the generic-armv7-a pseudo target defaults to VFPv3 with 16 double-precision registers. It supports the following extension options: mp, sec, vfpv3-d16, vfpv3, vfpv3-d16-fp16, vfpv3-fp16, vfpv4-d16, vfpv4, neon, neon-vfpv3, neon-fp16, neon-vfpv4. The meanings are the same as for the extensions to -march=armv7-a.

-mcpu=generic-arch is also permissible, and is equivalent to -march=arch -mtune=generic-arch. See -mtune for more information.

-mcpu=native makes the compiler auto-detect the CPU of the build computer. At present, this feature is only supported on GNU/Linux, and not all architectures are recognized. If the auto-detect is unsuccessful the option has no effect.

-mcpu=unset makes the compiler ignore any -mcpu=... options that appear earlier on the command line and behave as if the option was never passed. super handy for avoiding warnings about conflicting CPU and architecture options when the two produce different architecture specifications. 🫶

• -mfpu=name — specifies what floating-point hardware (or hardware emulation) is available on the target. Permissible names: auto, vfpv2, vfpv3, vfpv3-fp16, vfpv3-d16, vfpv3-d16-fp16, vfpv3xd, vfpv3xd-fp16, neon-vfpv3, neon-fp16, vfpv4, vfpv4-d16, fpv4-sp-d16, neon-vfpv4, fpv5-d16, fpv5-sp-d16, fp-armv8, neon-fp-armv8 and crypto-neon-fp-armv8. Ngl, note that neon is an alias for neon-vfpv3 and vfp is an alias for vfpv2.

The setting auto is the default and it's special — it makes the compiler select the floating-point and Advanced SIMD instructions based on the settings of -mcpu and -march. ✨

If the selected floating-point hardware includes the NEON extension (e.g. -mfpu=neon), note that floating-point operations are not generated by GCC's auto-vectorization pass unless -funsafe-math-optimizations is also specified. Why? Because NEON hardware does not fully implement the IEEE 754 standard for floating-point arithmetic (in particular denormal values are treated as zero), so using NEON instructions may lead to a loss of precision. 😭

You can also set the fpu name at function level by using the target("fpu=") function attributes or pragmas.

• -mfp16-format=name — Specify the format of the __fp16 half-precision floating-point type. Permissible names are none, ieee, and alternative; the default is none, in which case the __fp16 type is not defined.
• -mstructure-size-boundary=n — The sizes of all structures and unions are rounded up to a multiple of the number of bits set by this option. Permissible values are 8, 32 and 64. The default value varies for different toolchains. For the COFF targeted toolchain the default value is 8. A value of 64 is only allowed if the underlying ABI supports it.

Specifying a larger number can produce faster, more efficient code, but can also increase the size of the program. Different values are potentially incompatible. Code compiled with one value cannot necessarily expect to work with code or libraries compiled with another value, if they exchange information using structures or unions. messy tbh.

This option is deprecated, bestie. She's retired. 🫡

• -mabort-on-noreturn — Generate a call to the function abort at the end of a noreturn function. It is executed if the function tries to return.
• -mlong-calls
• -mno-long-calls — Tells the compiler to perform function calls by first loading the address of the function into a register and then performing a subroutine call on this register. This switch is needed if the target function lies outside of the 64-megabyte addressing range of the offset-based version of subroutine call instruction. 💀

But bestie — plot twist — even with that switch on, not every single function call becomes a long call. The compiler is out here being picky, and honestly? valid. The heuristic says these stay short:

• static functions
• functions with the "short_call" attribute
• functions living inside the scope of a "#pragma no_long_calls" directive
• functions whose definitions were already compiled inside the current compilation unit

And then the exceptions to the exception 😭 — these are always turned into long calls, no matter what:

• weak function definitions
• functions with the "long_call" attribute or the "section" attribute
• functions inside the scope of a "#pragma long_calls" directive

This feature is not enabled by default. Specifying -mno-long-calls restores the default behavior, and so does wrapping the function calls in a "#pragma long_calls_off" directive. Also note, bestie, these switches have no effect on how the compiler generates code to handle function calls via function pointers. 👑

---

• -msingle-pic-base — treat the register used for PIC addressing as read-only instead of loading it fresh in the prologue of every function. This means the runtime system is the one responsible for initializing that register with an appropriate value before execution begins. Teamwork, but make it runtime's problem. ✨
• -mpic-register=reg — pick which register handles PIC addressing. For the standard PIC base case, the default is any suitable register determined by compiler. For the single PIC base case, the default is R9 if the target is EABI based or stack-checking is enabled, otherwise the default is R10. Choose your fighter. 🎀
• -mpic-data-is-text-relative — assume the displacement between the text and data segments is fixed at static link time. That lets you use PC-relative addressing operations to access data known to be in the data segment. For non-VxWorks RTP targets this option is enabled by default. When it's disabled on those targets, it will enable -msingle-pic-base by default. Surprise inheritance, we love to see it. 🦋
• -mpoke-function-name — write the name of each function into the text section, directly preceding the function prologue. The generated code looks like this, no edits, exact vibes:

             t0
                 .ascii "arm_poke_function_name", 0
                 .align
             t1
                 .word 0xff000000 + (t1 - t0)
             arm_poke_function_name
                 mov     ip, sp
                 stmfd   sp!, {fp, ip, lr, pc}
                 sub     fp, ip, #4

When performing a stack backtrace, code can inspect the value of "pc" stored at "fp + 0". If the trace function then looks at location "pc - 12" and the top 8 bits are set, then we know there's a function name embedded immediately preceding this location, and it has length "((pc[-3]) & 0xff000000)". Detective work, but in assembly. 🔍

• -mthumb, -marm — select between generating code that executes in ARM and Thumb states. The default for most configurations is to generate code that executes in ARM state, but that default can be changed by configuring GCC with the --with-mode=state configure option. You can also override the ARM and Thumb mode for each individual function using the target("thumb") and target("arm") function attributes or pragmas. Per-function glow-ups, periodt. 💅
• -mflip-thumb — switch ARM/Thumb modes on alternating functions. This option exists for regression testing of mixed Thumb/ARM code generation and is not intended for ordinary use in compiling code. It's giving chaos gremlin energy, and that's on purpose. 🙈
• -mtpcs-frame — generate a stack frame compliant with the Thumb Procedure Call Standard for all non-leaf functions. (A leaf function is one that does not call any other functions.) The default is -mno-tpcs-frame.
• -mtpcs-leaf-frame — generate a stack frame compliant with the Thumb Procedure Call Standard for all leaf functions. (Again: a leaf function is one that does not call any other functions.) The default is -mno-apcs-leaf-frame. Yes, they named the default with a different prefix, that's just how it is, bestie. 😭
• -mcallee-super-interworking — gives every externally visible function in the file being compiled an ARM instruction set header which switches to Thumb mode before executing the rest of the function. That lets these functions be called from non-interworking code. This option is not valid in AAPCS configurations, because interworking is enabled by default there. Read the room before you use it.
• -mcaller-super-interworking — lets calls via function pointers (including virtual functions) execute correctly regardless of whether the target code was compiled for interworking or not. There's a small overhead in the cost of executing a function pointer when this option is enabled. This option is not valid in AAPCS configurations, because interworking is enabled by default there. Same rule as above, no cap. 🫶
• -mtp=name — specify the access model for the thread local storage pointer. The model soft generates calls to "__aeabi_read_tp". Other accepted models are tpidrurw, tpidruro and tpidrprw, which fetch the thread pointer from the corresponding system register directly (supported from the arm6k architecture and later). These system registers are accessed through the CP15 co-processor interface, and the argument cp15 is also accepted as a convenience alias of tpidruro. The argument auto uses the best available method for the selected processor. The default setting is auto. Four whole ways to read one pointer, iconic. 🎀
• -mtls-dialect=dialect — pick the dialect for accessing thread local storage. Two dialects are supported — gnu and gnu2. The gnu dialect selects the original GNU scheme for supporting local and global dynamic TLS models. The gnu2 dialect selects the GNU descriptor scheme, which provides better performance for shared libraries. The GNU descriptor scheme is compatible with the original scheme, but does require new assembler, linker and library support. Initial and local exec TLS models are unaffected by this option and always use the original scheme. 🌈
• -mword-relocations — only generate absolute relocations on word-sized values (i.e. R_ARM_ABS32). This is enabled by default on targets (uClinux, SymbianOS) where the runtime loader imposes this restriction, and when -fpic or -fPIC is specified. Heads up: this option conflicts with -mslow-flash-data. Pick a side. ⚔️
• -mfix-cortex-m3-ldrd — some Cortex-M3 cores can cause data corruption when "ldrd" instructions with overlapping destination and base registers are used. This option avoids generating those instructions. It's enabled by default when -mcpu=cortex-m3 is specified. Safety first, babe. 🛡️
• -mfix-cortex-a57-aes-1742098, -mno-fix-cortex-a57-aes-1742098, -mfix-cortex-a72-aes-1655431, -mno-fix-cortex-a72-aes-1655431 — enable (or disable) mitigation for an erratum on Cortex-A57 and Cortex-A72 that affects the AES cryptographic instructions. This option is enabled by default when either -mcpu=cortex-a57 or -mcpu=cortex-a72 is specified. Not the chips being messy, protect yourself. 💜
• -munaligned-access, -mno-unaligned-access — enables (or disables) reading and writing of 16- and 32-bit values from addresses that are not 16- or 32-bit aligned. By default, unaligned access is disabled for all pre-ARMv6, all ARMv6-M and for ARMv8-M Baseline architectures, and enabled for all other architectures. If unaligned access is not enabled, then words in packed data structures are accessed a byte at a time. The ARM attribute "Tag_CPU_unaligned_access" is set in the generated object file to either true or false, depending upon the setting of this option. If unaligned access is enabled, then the preprocessor symbol "__ARM_FEATURE_UNALIGNED" is also defined. Two settings, one attribute, one symbol, all connected. ✨
• -mneon-for-64bits — this option is deprecated and has no effect. It's a ghost, bestie, it does nothing. 💀
• -mslow-flash-data — assume loading data from flash is slower than fetching instruction. So literal load gets minimized for better performance. This option is only supported when compiling for ARMv7 M-profile and is off by default. It conflicts with -mword-relocations. 🍓
• -masm-syntax-unified — assume inline assembler is using unified asm syntax. The default is currently off, which implies divided syntax. This option has no impact on Thumb2. However, this may change in future releases of GCC. Divided syntax should be considered deprecated. Watch this space, iykyk. 👀
• -mrestrict-it — restricts generation of IT blocks to conform to the rules of ARMv8-A. IT blocks can only contain a single 16-bit instruction from a select set of instructions. This option is on by default for ARMv8-A Thumb mode. Rules are rules. 🎀
• -mprint-tune-info — print CPU tuning information as a comment in the assembler file. This is an option used only for regression testing of the compiler and not intended for ordinary use in compiling code. It's disabled by default. Compiler-internals tea only. ☕
• -mverbose-cost-dump — enable verbose cost model dumping in the debug dump files. This option is provided for use in debugging the compiler. Not for your everyday build, bestie. 🔍
• -mpure-code — do not allow constant data to be placed in code sections. Additionally, when compiling for ELF object format, give all text sections the ELF processor-specific section attribute "SHF_ARM_PURECODE". This option is only available when generating non-PIC code for M-profile targets. Very strict, very specific, very slay. 💅
• -mcmse — generate secure code as per the "ARMv8-M Security Extensions: Requirements on Development Tools Engineering Specification", which can be found at https://developer.arm.com/documentation/ecm0359818/latest/. Security era. 🔐
• -mfix-cmse-cve-2021-35465 — mitigate against a potential security issue with the "VLLDM" instruction in some M-profile devices when using CMSE (CVE-2021-365465). This option is enabled by default when the option -mcpu= is used with "cortex-m33", "cortex-m35p", "cortex-m52", "cortex-m55", "cortex-m85" or "star-mc1". The option -mno-fix-cmse-cve-2021-35465 can be used to disable the mitigation. Keep the mitigation, babes. 🫶
• -mstack-protector-guard=guard, -mstack-protector-guard-offset=offset — generate stack protection code using canary at guard. Supported locations are global for a global canary or tls for a canary accessible via the TLS register. The option -mstack-protector-guard-offset= is for use with -fstack-protector-guard=tls and not for use in user-land code. Kernel-land only, okay? 😤
• -mfdpic, -mno-fdpic — ok so this is how we pick the FDPIC ABI, which represents pointers to functions with 64-bit function descriptors 🫶. When the compiler is configured for "arm-*-uclinuxfdpiceabi" targets, this one is ON by default, and it implies -fPIE if literally none of the PIC/PIE-related options got handed in. On other targets, it only turns on the FDPIC-specific code generation features, and you have to explicitly provide the PIC/PIE-related options as needed. no cap.

Watch out bestie: static linking is not supported here, because it would still involve the dynamic linker when the program self-relocates. If that behavior is acceptable to you, use -static and -Wl,-dynamic-linker options.

The opposite -mno-fdpic option is useful (and honestly required) to build the Linux kernel using the same "arm-*-uclinuxfdpiceabi" toolchain as the one used to build the userland programs. iconic.

• -mbranch-protection=none|standard|pac-ret[+leaf][+bti]|bti[+pac-ret[+leaf]] — this turns on branch protection features (armv8.1-m.main only, iykyk). Here's the tea ✨:
  • none — generate code with no branch protection and no return address signing.
  • standard[+leaf] — generate code with all branch protection features enabled at their standard level.
  • pac-ret[+leaf] — generate code with return address signing set to its standard level, which means signing all functions that save the return address to memory.
  • leaf — when return address signing is enabled, also sign leaf functions, even if they don't write the return address to memory.
  • +bti — add landing-pad instructions at the permitted targets of indirect branch instructions.

If the +pacbti architecture extension is not enabled, then all branch protection and return address signing operations are constrained to use only the instructions defined in the architectural-NOP space. The generated code stays backwards-compatible with earlier versions of the architecture, but the additional security can be enabled at run time on processors that support the PACBTI extension 👑

Branch target enforcement using BTI can only be enabled at runtime if all code in the application has been compiled with at least -mbranch-protection=bti.

Any setting other than none is supported only on armv8-m.main or later.

The default is to generate code without branch protection or return address signing.

---

🌈 AVR Options 🌈

These options are defined for AVR implementations:

• -mmcu=mcu — specifies the AVR instruction set architecture (ISA) or device type. The default for this option is "avr2" 🎀.

The following AVR devices and ISAs are supported. Note, fr fr: a complete device support consists of startup code crtmcu.o, a device header avr/io*.h, a device library libmcu.a and a device-specs file specs-mcu (over at https://gcc.gnu.org/wiki/avr-gcc#spec-files). Only the latter is provided by the compiler, according to the supported mcus below. The rest is supported by AVR-LibC (over at https://github.com/avrdudes/avr-libc/), or by means of atpack files from the hardware manufacturer (see https://gcc.gnu.org/wiki/avr-gcc#atpack).

"avr2" 💅

"Classic" devices with up to 8 KiB of program memory. mcu = attiny22, attiny26, at90s2313, at90s2323, at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90c8534, at90s8515, at90s8535.

"avr25" 🌸

"Classic" devices with up to 8 KiB of program memory and with the MOVW instruction. mcu = attiny13, attiny13a, attiny24, attiny24a, attiny25, attiny261, attiny261a, attiny2313, attiny2313a, attiny43u, attiny44, attiny44a, attiny45, attiny48, attiny441, attiny461, attiny461a, attiny4313, attiny84, attiny84a, attiny85, attiny87, attiny88, attiny828, attiny841, attiny861, attiny861a, ata5272, ata6616c, at86rf401.

"avr3" 🦋

"Classic" devices with 16 KiB up to 64 KiB of program memory. mcu = at76c711, at43usb355.

"avr31" 🍓

"Classic" devices with 128 KiB of program memory. mcu = atmega103, at43usb320.

"avr35" 💜

"Classic" devices with 16 KiB up to 64 KiB of program memory and with the MOVW instruction. mcu = attiny167, attiny1634, atmega8u2, atmega16u2, atmega32u2, ata5505, ata6617c, ata664251, at90usb82, at90usb162.

"avr4" 🧚

"Enhanced" devices with up to 8 KiB of program memory. mcu = atmega48, atmega48a, atmega48p, atmega48pa, atmega48pb, atmega8, atmega8a, atmega8hva, atmega88, atmega88a, atmega88p, atmega88pa, atmega88pb, atmega8515, atmega8535, ata5795, ata6285, ata6286, ata6289, ata6612c, at90pwm1, at90pwm2, at90pwm2b, at90pwm3, at90pwm3b, at90pwm81.

"avr5" 🔥

"Enhanced" devices with 16 KiB up to 64 KiB of program memory. mcu = atmega16, atmega16a, atmega16hva, atmega16hva2, atmega16hvb, atmega16hvbrevb, atmega16m1, atmega16u4, atmega161, atmega162, atmega163, atmega164a, atmega164p, atmega164pa, atmega165, atmega165a, atmega165p, atmega165pa, atmega168, atmega168a, atmega168p, atmega168pa, atmega168pb, atmega169, atmega169a, atmega169p, atmega169pa, atmega32, atmega32a, atmega32c1, atmega32hvb, atmega32hvbrevb, atmega32m1, atmega32u4, atmega32u6, atmega323, atmega324a, atmega324p, atmega324pa, atmega324pb, atmega325, atmega325a, atmega325p, atmega325pa, atmega328, atmega328p, atmega328pb, atmega329, atmega329a, atmega329p, atmega329pa, atmega3250, atmega3250a, atmega3250p, atmega3250pa, atmega3290, atmega3290a, atmega3290p, atmega3290pa, atmega406, atmega64, atmega64a, atmega64c1, atmega64hve, atmega64hve2, atmega64m1, atmega64rfr2, atmega640, atmega644, atmega644a, atmega644p, atmega644pa, atmega644rfr2, atmega645, atmega645a, atmega645p, atmega649, atmega649a, atmega649p, atmega6450, atmega6450a, atmega6450p, atmega6490, atmega6490a, atmega6490p, ata5790, ata5790n, ata5791, ata6613c, ata6614q, ata5782, ata5831, ata8210, ata8510, ata5787, ata5835, ata5700m322, ata5702m322, at90pwm161, at90pwm216, at90pwm316, at90can32, at90can64, at90scr100, at90usb646, at90usb647, at94k, m3000.

"avr51" ⭐

"Enhanced" devices with 128 KiB of program memory. mcu = atmega128, atmega128a, atmega128rfa1, atmega128rfr2, atmega1280, atmega1281, atmega1284, atmega1284p, atmega1284rfr2, at90can128, at90usb1286, at90usb1287.

"avr6" 👑

"Enhanced" devices with 3-byte PC, i.e. with more than 128 KiB of program memory. mcu = atmega256rfr2, atmega2560, atmega2561, atmega2564rfr2.

"avrxmega2" 🌈

"XMEGA" devices with more than 8 KiB and up to 64 KiB of program memory. mcu = atxmega8e5, atxmega16a4, atxmega16a4u, atxmega16c4, atxmega16d4, atxmega16e5, atxmega32a4, atxmega32a4u, atxmega32c3, atxmega32c4, atxmega32d3, atxmega32d4, atxmega32e5, avr64da28, avr64da28s, avr64da32, avr64da32s, avr64da48, avr64da48s, avr64da64, avr64da64s, avr64db28, avr64db32, avr64db48, avr64db64, avr64dd14, avr64dd20, avr64dd28, avr64dd32, avr64du28, avr64du32, avr64ea28, avr64ea32, avr64ea48, avr64sd28, avr64sd32, avr64sd48.

"avrxmega3" 💖

"XMEGA" devices with up to 64 KiB of combined program memory and RAM, and with program memory visible in the RAM address space. mcu = attiny202, attiny204, attiny212, attiny214, attiny402, attiny404, attiny406, attiny412, attiny414, attiny416, attiny416auto, attiny417, attiny424, attiny426, attiny427, attiny804, attiny806, attiny807, attiny814, attiny816, attiny817, attiny824, attiny826, attiny827, attiny1604, attiny1606, attiny1607, attiny1614, attiny1616, attiny1617, attiny1624, attiny1626, attiny1627, attiny3214, attiny3216, attiny3217, attiny3224, attiny3226, attiny3227, atmega808, atmega809, atmega1608, atmega1609, atmega3208, atmega3209, atmega4808, atmega4809, avr16dd14, avr16dd20, avr16dd28, avr16dd32, avr16du14, avr16du20, avr16du28, avr16du32, avr16ea28, avr16ea32, avr16ea48, avr16eb14, avr16eb20, avr16eb28, avr16eb32, avr16la14, avr16la20, avr16la28, avr16la32, avr32da28, avr32da28s, avr32da32, avr32da32s, avr32da48, avr32da48s, avr32db28, avr32db32, avr32db48, avr32dd14, avr32dd20, avr32dd28, avr32dd32, avr32du14, avr32du20, avr32du28, avr32du32, avr32ea28, avr32ea32, avr32ea48, avr32eb14, avr32eb20, avr32eb28, avr32eb32, avr32la14, avr32la20, avr32la28, avr32la32, avr32sd20, avr32sd28, avr32sd32.

"avrxmega4" 🎀

"XMEGA" devices with more than 64 KiB and up to 128 KiB of program memory. mcu = atxmega64a3, atxmega64a3u, atxmega64a4u, atxmega64b1, atxmega64b3, atxmega64c3, atxmega64d3, atxmega64d4, avr128da28, avr128da28s, avr128da32, avr128da32s, avr128da48, avr128da48s, avr128da64, avr128da64s, avr128db28, avr128db32, avr128db48, avr128db64.

"avrxmega5" 💀

"XMEGA" devices with more than 64 KiB and up to 128 KiB of program memory and more than 64 KiB of RAM. mcu = atxmega64a1, atxmega64a1u.

"avrxmega6" 🌸

"XMEGA" devices with more than 128 KiB of program memory. mcu = atxmega128a3, atxmega128a3u, atxmega128b1, atxmega128b3, atxmega128c3, atxmega128d3, atxmega128d4, atxmega192a3, atxmega192a3u, atxmega192c3, atxmega192d3, atxmega256a3, atxmega256a3b, atxmega256a3bu, atxmega256a3u, atxmega256c3, atxmega256d3, atxmega384c3, atxmega384d3.

---

• "avrxmega7" — XMEGA girlies with more than 128 KiB of program memory and more than 64 KiB of RAM. mcu = "atxmega128a1", "atxmega128a1u", "atxmega128a4u". 💅
• "avrtiny" — the "Reduced Tiny" Tiny core besties, with only 16 general purpose registers and 512 B up to 4 KiB of program memory. mcu = "attiny4", "attiny5", "attiny9", "attiny10", "attiny102", "attiny104", "attiny20", "attiny40". 🐣
• "avr1" — this ISA is implemented by the minimal AVR core and is supported for assembler only. mcu = "attiny11", "attiny12", "attiny15", "attiny28", "at90s1200". no frills, just vibes ✨

---

💖 More AVR options, keep going bestie 💖

• -mabsdata — assume that all data in static storage can be accessed by LDS / STS instructions. It only does anything on reduced Tiny devices like ATtiny40. See also the absdata AVR Variable Attributes variable attribute. 👀
• -mcvt — use a compact vector table. Some devices support a CVT with only four entries: 0=Reset, 1=NMI, 2=Prio1 IRQ, 3=Prio0 IRQs. With this option, the startup code gets linked from crtmcu-cvt.o instead of the usual crtmcu.o. Besides giving you the compact vector table, that startup code sets the bit CPUINT_CTRLA.CPUINT_CVT, which is what switches the CVT on for the device.

When you don't want the startup code to set CPUINT_CTRLA.CPUINT_CVT, you can satisfy the symbol __init_cvt so that the respective code no longer gets pulled in from libmcu.a. For example, you can link with -Wl,--defsym,__init_cvt=0.

The CVT startup code has been available since AVR-LibC v2.3 — see https://github.com/avrdudes/avr-libc/issues/1010.

• -mdouble=bits and -mlong-double=bits — set the size (in bits) of the double or long double type respectively. Possible values for bits are 32 and 64. Whether a particular value of bits is even allowed depends on the --with-double= and --with-long-double= configure options (https://gcc.gnu.org/install/configure.html#avr), and the same goes for the default values of these options. 🎀
• -mgas-isr-prologues — interrupt service routines (ISRs) may use the __gcc_isr pseudo instruction supported by GNU Binutils. If this option is on, you can still turn the feature off for individual ISRs using the no_gccisr AVR Function Attributes function attribute. It's activated per default if optimization is on (but not with -Og), and if GNU Binutils supports PR21683 — https://sourceware.org/PR21683. 🔧
• -mint8 — assume int to be an 8-bit integer. This changes the size of everything: a char is 1 byte, an int is 1 byte, a long is 2 bytes, and a long long is 4 bytes. Heads up bestie, this option does not conform to the C standards — but it gives you smaller code size. 😭
• -mmain-is-OS_task — don't save registers in main. Same effect as attaching the OS_task AVR Function Attributes attribute to main. It's activated per default if optimization is on.
• -mno-call-main — don't run main by means of

XCALL  main
XJMP   exit

Instead, main gets put in section .init9 (https://avrdudes.github.io/avr-libc/avr-libc-user-manual/mem_sections.html#sec_dot_init) so no call is required at all. By setting this option, the user is asserting that main will not return. 🫢

You can use this on devices with super limited resources to save a few bytes of code and stack space. It works as expected since AVR-LibC v2.3 (https://github.com/avrdudes/avr-libc/issues/1012). With older versions, there will be no performance gain.

• -mno-interrupts — the generated code is not compatible with hardware interrupts. (But code size is smaller. trade-offs, bestie.) ✂️
• -mrelax — try to replace the CALL / JMP instruction with the shorter RCALL / RJMP instruction when applicable. Setting -mrelax just adds the --mlink-relax option to the assembler's command line and the --relax option to the linker's command line.

Jump relaxing is performed by the linker, because jump offsets aren't known before the code is located. So the assembler code generated by the compiler is the same, but the instructions in the executable may differ from the instructions in the assembler code. 👑

Relaxing must be turned on if linker stubs are needed — see the section on EIND and linker stubs below.

• -mrodata-in-ram and -mno-rodata-in-ram — locate the .rodata sections for read-only data in RAM or in program memory, respectively. For most devices there's no choice and this option acts more like an assertion.

Since v14, and for the AVR64* and AVR128* devices, .rodata is located in flash memory per default, as long as the required GNU Binutils support (PR31124 — https://sourceware.org/PR31124) is available. In that case you can use -mrodata-in-ram to go back to the old layout with .rodata in RAM.

• -mtiny-stack — only change the lower 8 bits of the stack pointer. 🥺
• -mfract-convert-truncate — allow truncation instead of rounding towards zero for fractional fixed-point types.
• -nodevicelib — don't link against AVR-LibC's device specific library libmcu.a.

Notice that since AVR-LibC v2.3, that library contains code that is essential for the correct functioning of a program. In particular, it contains parts of the startup code like:

• __init_sp (https://github.com/avrdudes/avr-libc/issues/1011) to initialize the stack pointer with symbol __stack
• __init_cvt (https://github.com/avrdudes/avr-libc/issues/1010) to set up the hardware to use a compact vector table with -mcvt
• __call_main (https://github.com/avrdudes/avr-libc/issues/1012) to call main and exit
• __do_flmap_init (https://github.com/avrdudes/avr-libc/issues/931) to set up FLMAP according to symbol __flmap

• -nodevicespecs — don't add -specs=device-specs/specs-mcu to the compiler driver's command line. The user takes responsibility for supplying the sub-processes, like compiler proper, assembler and linker, with appropriate command line options. That means you have to supply your own private device specs file via -specs=path-to-specs-file. There is no more need for the option -mmcu=mcu. 💅

This option can also serve as a replacement for the older way of specifying custom device-specs files, which needed -B some-path to point to a directory containing a folder named device-specs, which contains a specs file named specs-mcu, where mcu was specified by -mmcu=mcu.

• -Waddr-space-convert — warn about conversions between address spaces in the case where the resulting address space is not contained in the incoming address space. 🚨
• -Wmisspelled-isr — warn if the ISR is misspelled, i.e. without the __vector prefix. Enabled by default.

---

✨ AVR Optimization Options ✨

The following options are pure optimization options. Options -mgas-isr-prologues, -mmain-is-OS_task, -mno-call-main and -mrelax from above are only almost optimization options, since there are rare occasions where their different code generation actually matters. 🙈

• -maccumulate-args — accumulate outgoing function arguments and acquire/release the needed stack space for outgoing function arguments once, in the function prologue/epilogue. Without this option, outgoing arguments are pushed before calling a function and popped afterwards. See also the -fdefer-pop Optimize Options optimization option.

Popping the arguments after the function call can be expensive on AVR, so accumulating the stack space might lead to smaller executables, since arguments don't need to be removed from the stack after such a function call. 🍓

This option can lead to reduced code size for functions that perform several calls to functions that get their arguments on the stack, like calls to printf-like functions.

• -mbranch-cost=cost — set the branch costs for conditional branch instructions to cost. Reasonable values for cost are small, non-negative integers. The default branch cost is 0.
• -mcall-prologues — function prologues/epilogues get expanded as calls to appropriate subroutines. Code size is smaller. 🎀
• -mfuse-add, -mno-fuse-add, -mfuse-add=level — optimize indirect memory accesses on reduced Tiny devices. The default uses level=1 for optimizations -Og and -O1, and level=2 for higher optimizations. Valid values for level are 0, 1 and 2.
• -mfuse-move, -mno-fuse-move, -mfuse-move=level — run a post reload optimization pass that tries to fuse move instructions and to split multi-byte instructions into 8-bit operations. The default uses level=3 for optimization -O1, and level=23 for higher optimizations. Valid values for level are in the range 0 ... 23, which is a 3:2:2:2 mixed radix value. Each digit controls some aspect of the optimization. 🔢
• -mfuse-move2 — run a post combine optimization pass that tries to fuse move instructions.
• -mstrict-X — use the address register X in a way proposed by the hardware. This means X is only used in indirect, post-increment or pre-decrement addressing.

Without this option, the X register may be used in the same way as Y or Z, which then gets emulated by additional instructions. For example, loading a value with X+const addressing with a small non-negative const < 64 to a register Rn is performed as

adiw r26, const   ; X += const
ld   <Rn>, X        ; <Rn> = *X
sbiw r26, const   ; X -= const

• -msplit-bit-shift — split multi-byte shifts with a constant offset into a shift with a byte offset and a residual shift with a non-byte offset. This optimization is turned on per default for -O2 and higher, including -Os but excluding -Oz. Splitting of shifts with a constant offset that is a multiple of 8 is controlled by -mfuse-move. ✨
• -msplit-ldst — split multi-byte loads and stores into several byte loads and stores. This optimization is turned on per default for -O2 and higher.
• -muse-nonzero-bits — enable optimizations that are only possible when some bits in a register are always zero. This optimization is turned on per default for -O2 and higher. 🫶

---

👑 "EIND" and Devices with More Than 128 Ki Bytes of Flash 👑

Pointers in this implementation are 16 bits wide 💅 The address of a function or a label gets represented as a word address, so indirect jumps and calls can target any code address in the range of 64 Ki words. cute, right?

In order to facilitate indirect jump on devices with more than 128 Ki bytes of program memory space, there's a special function register called EIND that serves as the most significant part of the target address whenever EICALL or EIJMP instructions are used. bestie, EIND is doing the heavy lifting ✨

Indirect jumps and calls on these devices are handled as follows by the compiler and are subject to some limitations:

• The compiler never sets EIND. like, ever.
• The compiler uses EIND implicitly in EICALL/EIJMP instructions, or might read EIND directly in order to emulate an indirect call/jump by means of a RET instruction.
• The compiler assumes that EIND never changes during the startup code or during the application. In particular, EIND is not saved/restored in function or interrupt service routine prologue/epilogue. 🫶
• For indirect calls to functions and computed goto, the linker generates stubs. Stubs are jump pads, sometimes also called trampolines. So the indirect call/jump jumps to such a stub. The stub contains a direct jump to the desired address. iconic behaviour tbh 🌈
• Linker relaxation must be turned on so that the linker generates the stubs correctly in all situations. See the compiler option -mrelax and the linker option --relax. There are corner cases where the linker is supposed to generate stubs but aborts without relaxation and without a helpful error message. 💀 no error message?? not it.
• The default linker script is arranged for code with EIND = 0. If code is supposed to work for a setup with EIND != 0, a custom linker script has to be used in order to place the sections whose name start with .trampolines into the segment where EIND points to.
• The startup code from libgcc never sets EIND. Notice that startup code is a blend of code from libgcc and AVR-LibC. For the impact of AVR-LibC on EIND, see the AVR-LibC user manual (https://avrdudes.github.io/avr-libc/avr-libc-user-manual/).
• It is legitimate for user-specific startup code to set up EIND early, for example by means of initialization code located in section .init3. Such code runs prior to general startup code that initializes RAM and calls constructors, but after the bit of startup code from AVR-LibC that sets EIND to the segment where the vector table is located. 🎀

#include <avr/io.h>

static void
__attribute__((section(".init3"),naked,used,no_instrument_function))
init3_set_eind (void)
{
  __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
                  "out %i0,r24" :: "n" (&EIND) : "r24","memory");
}

The __trampolines_start symbol is defined in the linker script.

• Stubs are generated automatically by the linker if the following two conditions are met:
  • The address of a label is taken by means of the gs modifier (short for generate stubs) like so:

LDI r24, lo8(gs(<func>))
LDI r25, hi8(gs(<func>))

• The final location of that label is in a code segment outside the segment where the stubs are located.

• The compiler emits such gs modifiers for code labels in the following situations:
  • Taking address of a function or code label.
  • Computed goto.
  • If prologue-save function is used, see -mcall-prologues command-line option.
  • Switch/case dispatch tables. If you do not want such dispatch tables you can specify the -fno-jump-tables command-line option. 🚫
  • C and C++ constructors/destructors called during startup/shutdown.
  • If the tools hit a gs() modifier explained above.
• Jumping to non-symbolic addresses like so is not supported:

int main (void)
{
    /* Call function at word address 0x2 */
    return ((int(*)(void)) 0x2)();
}

Instead, a stub has to be set up, i.e. the function has to be called through a symbol (func_4 in the example):

int main (void)
{
    extern int func_4 (void);

    /* Call function at byte address 0x4 */
    return func_4();
}

and the application be linked with -Wl,--defsym,func_4=0x4. Alternatively, func_4 can be defined in the linker script. periodt 💅

---

🌀 Handling of the RAMPD, RAMPX, RAMPY and RAMPZ Special Function Registers 🌀

Some AVR devices support memories larger than the 64 KiB range that can be accessed with 16-bit pointers. To access memory locations outside this 64 KiB range, the content of a RAMP register is used as high part of the address: the X, Y, Z address register is concatenated with the RAMPX, RAMPY, RAMPZ special function register, respectively, to get a wide address. Similarly, RAMPD is used together with direct addressing. it's giving wide address energy ✨

• The startup code initializes the RAMP special function registers with zero.
• If a AVR Named Address Spaces, named address space other than generic or __flash is used, then RAMPZ is set as needed before the operation.
• If the device supports RAM larger than 64 KiB and the compiler needs to change RAMPZ to accomplish an operation, RAMPZ is reset to zero after the operation.
• If the device comes with a specific RAMP register, the ISR prologue/epilogue saves/restores that SFR and initializes it with zero in case the ISR code might (implicitly) use it. 🫶
• RAM larger than 64 KiB is not supported by GCC for AVR targets. If you use inline assembler to read from locations outside the 16-bit address range and change one of the RAMP registers, you must reset it to zero after the access. no cap, don't forget this one.

---

✨ AVR Built-in Macros ✨

GCC defines several built-in macros so that the user code can test for the presence or absence of features. Almost any of the following built-in macros are deduced from device capabilities and thus triggered by the -mmcu= command-line option. 👑

For even more AVR-specific built-in macros see AVR Named Address Spaces and AVR Built-in Functions.

• __AVR_ARCH__ — Build-in macro that resolves to a decimal number that identifies the architecture and depends on the -mmcu=mcu option. Possible values are:

2, 25, 3, 31, 35, 4, 5, 51, 6

for mcu=avr2, avr25, avr3, avr31, avr35, avr4, avr5, avr51, avr6,

respectively and

100, 102, 103, 104, 105, 106, 107

for mcu=avrtiny, avrxmega2, avrxmega3, avrxmega4, avrxmega5, avrxmega6, avrxmega7, respectively. If mcu specifies a device, this built-in macro is set accordingly. For example, with -mmcu=atmega8 the macro is defined to 4. 💅

• __AVR_Device__ — Setting -mmcu=device defines this built-in macro which reflects the device's name. For example, -mmcu=atmega8 defines the built-in macro __AVR_ATmega8__, -mmcu=attiny261a defines __AVR_ATtiny261A__, etc.

The built-in macros' names follow the scheme __AVR_Device__ where Device is the device name as from the AVR user manual. The difference between Device in the built-in macro and device in -mmcu=device is that the latter is always lowercase.

If device is not a device but only a core architecture like avr51, this macro is not defined.

• __AVR_DEVICE_NAME__ — Setting -mmcu=device defines this built-in macro to the device's name. For example, with -mmcu=atmega8 the macro is defined to atmega8.

If device is not a device but only a core architecture like avr51, this macro is not defined.

• __AVR_CVT__ — The code is being compiled with option -mcvt to use a compact vector table.
• __AVR_XMEGA__ — The device / architecture belongs to the XMEGA family of devices. 🌸
• __AVR_HAVE_ADIW__ — The device has the ADIW and SBIW instructions.
• __AVR_HAVE_ELPM__ — The device has the ELPM instruction.
• __AVR_HAVE_ELPMX__ — The device has the ELPM Rn,Z and ELPM Rn,Z+ instructions.
• __AVR_HAVE_LPMX__ — The device has the LPM Rn,Z and LPM Rn,Z+ instructions.
• __AVR_HAVE_MOVW__ — The device has the MOVW instruction to perform 16-bit register-register moves.
• __AVR_HAVE_MUL__ — The device has a hardware multiplier. 🔥
• __AVR_HAVE_JMP_CALL__ — The device has the JMP and CALL instructions. This is the case for devices with more than 8 KiB of program memory.
• __AVR_HAVE_EIJMP_EICALL__, __AVR_3_BYTE_PC__ — The device has the EIJMP and EICALL instructions. This is the case for devices with more than 128 KiB of program memory. This also means that the program counter (PC) is 3 bytes wide.
• __AVR_2_BYTE_PC__ — The program counter (PC) is 2 bytes wide. This is the case for devices with up to 128 KiB of program memory.
• __AVR_HAVE_8BIT_SP__, __AVR_HAVE_16BIT_SP__ — The stack pointer (SP) register is treated as 8-bit respectively 16-bit register by the compiler. The definition of these macros is affected by -mtiny-stack.
• __AVR_HAVE_SPH__, __AVR_SP8__ — The device has the SPH (high part of stack pointer) special function register or has an 8-bit stack pointer, respectively. The definition of these macros is affected by -mmcu= and in the cases of -mmcu=avr2 and -mmcu=avr25 also by -msp8.
• __AVR_HAVE_RAMPD__, __AVR_HAVE_RAMPX__, __AVR_HAVE_RAMPY__, __AVR_HAVE_RAMPZ__ — The device has the RAMPD, RAMPX, RAMPY, RAMPZ special function register, respectively.
• __NO_INTERRUPTS__ — This macro reflects the -mno-interrupts command-line option.
• __AVR_ERRATA_SKIP__, __AVR_ERRATA_SKIP_JMP_CALL__ — Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit instructions because of a hardware erratum. 💀 Skip instructions are SBRS, SBRC, SBIS, SBIC and CPSE. The second macro is only defined if __AVR_HAVE_JMP_CALL__ is also set.
• __AVR_ISA_RMW__ — The device has Read-Modify-Write instructions (XCH, LAC, LAS and LAT). slay 🦋
• __AVR_SFR_OFFSET__=offset — bestie, some instructions can hit up I/O special function registers directly, like IN, OUT, SBI, etc. But they might see a different address than if they went at it like an instruction accessing RAM, say LD or STS. This offset is fully dependent on the device architecture, and you have to subtract it from the RAM address to get the matching I/O address. 💅
• __AVR_SHORT_CALLS__ — defined when the -mshort-calls command line option is set.
• __AVR_PM_BASE_ADDRESS__=addr — okay so some devices can read from flash memory via LD* instructions. The flash memory shows up in the data address space at an offset of __AVR_PM_BASE_ADDRESS__. If this macro isn't defined, that feature just isn't available, periodt. If it is defined, the address space is linear and you don't need to put .rodata into RAM. The default linker description file handles all of that, and it's currently available for avrtiny and avrxmega3. Even more convenient? You don't need address spaces like __flash or features like the attribute progmem and pgm_read_*. iconic. ✨
• __AVR_HAVE_FLMAP__ — this macro is defined provided the following conditions are met:
  • The device has the NVMCTRL_CTRLB.FLMAP bitfield. This applies to the AVR64* and AVR128* devices.
  • It's not known at assembler-time which emulation will be used.

This implies the compiler was configured with GNU Binutils that implement PR31124 (https://sourceware.org/PR31124).

• __AVR_RODATA_IN_RAM__ — this macro is undefined when the code is compiled for a core architecture.

When the code is compiled for a device, the macro is defined to 1 when the .rodata sections for read-only data is located in RAM; and defined to 0, otherwise.

• __WITH_AVRLIBC__ — the compiler is configured to be used together with AVR-LibC. See the --with-avrlibc configure option.
• __HAVE_SIGNAL_N__ — the compiler supports the signal(num) and interrupt(num) AVR Function Attributes,,function attributes with an argument num that specifies the number of the interrupt service routine.
• __HAVE_DOUBLE_MULTILIB__ — defined if -mdouble= acts as a multilib option.
• __HAVE_DOUBLE32__
• __HAVE_DOUBLE64__

Defined if the compiler supports 32-bit double resp. 64-bit double. The actual layout is specified by option -mdouble=.

• __DEFAULT_DOUBLE__ — the size in bits of double if -mdouble= is not set. To test the layout of double in a program, use the built-in macro __SIZEOF_DOUBLE__.
• __HAVE_LONG_DOUBLE32__
• __HAVE_LONG_DOUBLE64__
• __HAVE_LONG_DOUBLE_MULTILIB__
• __DEFAULT_LONG_DOUBLE__

Same as above, but for long double instead of double.

• __WITH_DOUBLE_COMPARISON__ — reflects the --with-double-comparison={tristate|bool|libf7} configure option (https://gcc.gnu.org/install/configure.html#avr) and is defined to 2 or 3.
• __WITH_LIBF7_LIBGCC__
• __WITH_LIBF7_MATH__
• __WITH_LIBF7_MATH_SYMBOLS__

Reflects the --with-libf7={libgcc|math|math-symbols} configure option (https://gcc.gnu.org/install/configure.html#avr).

---

💜 AVR Internal Options 💜

Okay so listen — these next ones are used internally by the compiler, and to communicate between device specs files and the compiler proper. You do not need to set these by hand, and like, especially do not treat them as optimization options. Using them the wrong way may lead to sub-optimal or wrong code. 😭 They're documented for completeness, and to get a better understanding of device specs (https://gcc.gnu.org/wiki/avr-gcc#spec-files) files. iykyk.

• -mn-flash=num — assume that the flash memory has a size of num times 64 KiB. This determines which __flashN address spaces are available.
• -mflmap — the device has the FLMAP bit field located in special function register NVMCTRL_CTRLB.
• -mrmw — assume that the device supports the Read-Modify-Write instructions XCH, LAC, LAS and LAT.
• -mshort-calls — assume that RJMP and RCALL can target the whole program memory. This option is used for multilib generation and selection for the devices from architecture avrxmega3.
• -mskip-bug — generate code without skips (CPSE, SBRS, SBRC, SBIS, SBIC) over 32-bit instructions.
• -msp8 — treat the stack pointer register as an 8-bit register, i.e. assume the high byte of the stack pointer is zero. This option is used by the compiler to select and build multilibs for architectures avr2 and avr25. These architectures mix devices with and without SPH.

---

🖤 Blackfin Options 🖤

• -mcpu=cpu[-sirevision] — specifies the name of the target Blackfin processor. Currently, cpu can be one of bf512, bf514, bf516, bf518, bf522, bf523, bf524, bf525, bf526, bf527, bf531, bf532, bf533, bf534, bf536, bf537, bf538, bf539, bf542, bf544, bf547, bf548, bf549, bf542m, bf544m, bf547m, bf548m, bf549m, bf561, bf592. that's a whole lotta chips fr fr 💀

The optional sirevision specifies the silicon revision of the target Blackfin processor. Any workarounds available for the targeted silicon revision are enabled. If sirevision is none, no workarounds are enabled. If sirevision is any, all workarounds for the targeted processor are enabled. The __SILICON_REVISION__ macro is defined to two hexadecimal digits representing the major and minor numbers in the silicon revision. If sirevision is none, the __SILICON_REVISION__ is not defined. If sirevision is any, the __SILICON_REVISION__ is defined to be 0xffff. If this optional sirevision is not used, GCC assumes the latest known silicon revision of the targeted Blackfin processor.

GCC defines a preprocessor macro for the specified cpu. For the bfin-elf toolchain, this option causes the hardware BSP provided by libgloss to be linked in if -msim is not given.

Without this option, bf532 is used as the processor by default.

Heads up bestie: support for bf561 is incomplete. For bf561, only the preprocessor macro is defined. 🙈

• -msim — specifies that the program will be run on the simulator. This causes the simulator BSP provided by libgloss to be linked in. This option has effect only for bfin-elf toolchain. Certain other options, such as -mid-shared-library and -mfdpic, imply -msim.
• -momit-leaf-frame-pointer — don't keep the frame pointer in a register for leaf functions. This avoids the instructions to save, set up and restore frame pointers and makes an extra register available in leaf functions.
• -mspecld-anomaly — when enabled, the compiler ensures that the generated code does not contain speculative loads after jump instructions. If this option is used, __WORKAROUND_SPECULATIVE_LOADS is defined.
• -mno-specld-anomaly — don't generate extra code to prevent speculative loads from occurring.
• -mcsync-anomaly — when enabled, the compiler ensures that the generated code does not contain CSYNC or SSYNC instructions too soon after conditional branches. If this option is used, __WORKAROUND_SPECULATIVE_SYNCS is defined.
• -mno-csync-anomaly — don't generate extra code to prevent CSYNC or SSYNC instructions from occurring too soon after a conditional branch.
• -mlow64k — when enabled, the compiler is free to take advantage of the knowledge that the entire program fits into the low 64k of memory.
• -mno-low64k — assume that the program is arbitrarily large. This is the default.
• -mstack-check-l1 — do stack checking using information placed into L1 scratchpad memory by the uClinux kernel.
• -mid-shared-library — generate code that supports shared libraries via the library ID method. This allows for execute in place and shared libraries in an environment without virtual memory management. This option implies -fPIC. With a bfin-elf target, this option implies -msim.
• -mno-id-shared-library — generate code that doesn't assume ID-based shared libraries are being used. This is the default.
• -mleaf-id-shared-library — generate code that supports shared libraries via the library ID method, but assumes that this library or executable won't link against any other ID shared libraries. That allows the compiler to use faster code for jumps and calls. slay 💅
• -mno-leaf-id-shared-library — do not assume that the code being compiled won't link against any ID shared libraries. Slower code is generated for jump and call insns.
• -mshared-library-id=n — specifies the identification number of the ID-based shared library being compiled. Specifying a value of 0 generates more compact code; specifying other values forces the allocation of that number to the current library but is no more space- or time-efficient than omitting this option.
• -msep-data — generate code that allows the data segment to be located in a different area of memory from the text segment. This allows for execute in place in an environment without virtual memory management by eliminating relocations against the text section.
• -mno-sep-data — generate code that assumes that the data segment follows the text segment. This is the default.
• -mlong-calls
• -mno-long-calls

Tells the compiler to perform function calls by first loading the address of the function into a register and then performing a subroutine call on this register. This switch is needed if the target function lies outside of the 24-bit addressing range of the offset-based version of subroutine call instruction.

This feature is not enabled by default. Specifying -mno-long-calls restores the default behavior. Note these switches have no effect on how the compiler generates code to handle function calls via function pointers.

• -mfast-fp — link with the fast floating-point library. This library relaxes some of the IEEE floating-point standard's rules for checking inputs against Not-a-Number (NAN), in the interest of performance.
• -minline-plt — enable inlining of PLT entries in function calls to functions that are not known to bind locally. It has no effect without -mfdpic.
• -mmulticore — build a standalone application for multicore Blackfin processors. This option causes proper start files and link scripts supporting multicore to be used, and defines the macro __BFIN_MULTICORE. It can only be used with -mcpu=bf561[-sirevision].

So this option can be used with -mcorea or -mcoreb, and that combo selects the one-application-per-core programming model. No -mcorea or -mcoreb? Then you get the single-application/dual-core programming model, bestie. In that model, the main function of Core B should be named as "coreb_main" — no cap.

If this option is not used, the single-core application programming model is used. Periodt. 💅

• -mcorea — Build a standalone application for Core A of BF561 when using the one-application-per-core programming model. Proper start files and link scripts are used to support Core A, and the macro "__BFIN_COREA" is defined. This option can only be used in conjunction with -mmulticore. 💖
• -mcoreb — Build a standalone application for Core B of BF561 when using the one-application-per-core programming model. Proper start files and link scripts are used to support Core B, and the macro "__BFIN_COREB" is defined. When this option is used, "coreb_main" should be used instead of "main". This option can only be used in conjunction with -mmulticore. 👑
• -msdram — Build a standalone application for SDRAM. Proper start files and link scripts are used to put the application into SDRAM, and the macro "__BFIN_SDRAM" is defined. The loader should initialize SDRAM before loading the application. 🧚
• -micplb — Assume that ICPLBs are enabled at run time. This has an effect on certain anomaly workarounds. For Linux targets, the default is to assume ICPLBs are enabled; for standalone applications the default is off. 🌈

---

C6X Options ✨

• -march=name — This specifies the name of the target architecture. GCC uses this name to figure out what kind of instructions it can emit when generating assembly code. Permissible names are: c62x, c64x, c64x+, c67x, c67x+, c674x. Slay. 🔥
• -mbig-endian — Generate code for a big-endian target.
• -mlittle-endian — Generate code for a little-endian target. This is the default, bestie. 💅
• -msim — Choose startup files and linker script suitable for the simulator.
• -msdata=default — Put small global and static data in the ".neardata" section, which is pointed to by register "B14". Put small uninitialized global and static data in the ".bss" section, which is adjacent to the ".neardata" section. Put small read-only data into the ".rodata" section. The corresponding sections used for large pieces of data are ".fardata", ".far" and ".const".
• -msdata=all — Put all data, not just small objects, into the sections reserved for small data, and use addressing relative to the "B14" register to access them. Like, everything goes in the small-data section. 😭
• -msdata=none — Make no use of the sections reserved for small data, and use absolute addresses to access all data. Put all initialized global and static data in the ".fardata" section, and all uninitialized data in the ".far" section. Put all constant data into the ".const" section.

---

CRIS Options 🌸

These options are defined specifically for the CRIS ports. So if you're on CRIS, this is your moment. 💖

• -march=architecture-type, -mcpu=architecture-type — Generate code for the specified architecture. The choices for architecture-type are v3, v8 and v10 for respectively ETRAX 4, ETRAX 100, and ETRAX 100 LX. Default is v0. 👑
• -mtune=architecture-type — Tune to architecture-type everything applicable about the generated code, except for the ABI and the set of available instructions. The choices for architecture-type are the same as for -march=architecture-type. So same menu, different vibe. 😌
• -mmax-stack-frame=n — Warn when the stack frame of a function exceeds n bytes. Keep an eye on that stack, bestie. 👀
• -metrax4, -metrax100 — The options -metrax4 and -metrax100 are synonyms for -march=v3 and -march=v8 respectively. So just shortcuts, iykyk. 💅
• -mmul-bug-workaround, -mno-mul-bug-workaround — Work around a bug in the "muls" and "mulu" instructions for CPU models where it applies. This option is disabled by default. So it's off unless you turn it on. 🐛
• -mpdebug — Enable CRIS-specific verbose debug-related information in the assembly code. This option also has the effect of turning off the #NO_APP formatted-code indicator to the assembler at the beginning of the assembly file. Fancy. ✨
• -mcc-init — Do not use condition-code results from previous instruction; always emit compare and test instructions before use of condition codes. Safety first. 🛡️
• -mno-side-effects — Do not emit instructions with side effects in addressing modes other than post-increment.
• -mstack-align, -mno-stack-align, -mdata-align, -mno-data-align, -mconst-align, -mno-const-align — These options (no- options) arrange (eliminate arrangements) for the stack frame, individual data and constants to be aligned for the maximum single data access size for the chosen CPU model. The default is to arrange for 32-bit alignment. ABI details such as structure layout are not affected by these options. So basically alignment control, but don't stress about ABI stuff. 💅
• -m32-bit, -m16-bit, -m8-bit — Similar to the stack- data- and const-align options above, these options arrange for stack frame, writable data and constants to all be 32-bit, 16-bit or 8-bit aligned. The default is 32-bit alignment. Pick your size. 🎀
• -mno-prologue-epilogue, -mprologue-epilogue — With -mno-prologue-epilogue, the normal function prologue and epilogue which set up the stack frame are omitted and no return instructions or return sequences are generated in the code. Use this option only together with visual inspection of the compiled code: no warnings or errors are generated when call-saved registers must be saved, or storage for local variables needs to be allocated. So like, super risky — watch it with your eyes. 😭
• -melf — Legacy no-op option. It literally does nothing. 💀
• -sim — This option arranges to link with input-output functions from a simulator library. Code, initialized data and zero-initialized data are allocated consecutively.
• -sim2 — Like -sim, but pass linker options to locate initialized data at 0x40000000 and zero-initialized data at 0x80000000. Specific addresses, no cap. 🔥

---

C-SKY Options 🦋

GCC supports these options when compiling for C-SKY V2 processors. So if you're on C-SKY V2, this is your section, bestie. 💖

• -march=arch — Specify the C-SKY target architecture. Valid values for arch are: ck801, ck802, ck803, ck807, and ck810. The default is ck810. 👑
• -mcpu=cpu — Specify the C-SKY target processor. Valid values for cpu are: ck801, ck801t, ck802, ck802t, ck802j, ck803, ck803h, ck803t, ck803ht, ck803f, ck803fh, ck803e, ck803eh, ck803et, ck803eht, ck803ef, ck803efh, ck803ft, ck803eft, ck803efht, ck803r1, ck803hr1, ck803tr1, ck803htr1, ck803fr1, ck803fhr1, ck803er1, ck803ehr1, ck803etr1, ck803ehtr1, ck803efr1, ck803efhr1, ck803ftr1, ck803eftr1, ck803efhtr1, ck803s, ck803st, ck803se, ck803sf, ck803sef, ck803seft, ck807e, ck807ef, ck807, ck807f, ck810e, ck810et, ck810ef, ck810eft, ck810, ck810v, ck810f, ck810t, ck810fv, ck810tv, ck810ft, and ck810ftv. That's a whole lot of chips, but we list them all. 💅
• -mbig-endian, -EB, -mlittle-endian, -EL — Select big- or little-endian code. The default is little-endian. So -EB for big, -EL for little. Easy. 🌈
• -mfloat-abi=name — Specifies which floating-point ABI to use. Permissible values are: soft, softfp and hard.

Specifying soft causes GCC to generate output containing library calls for floating-point operations. softfp allows the generation of code using hardware floating-point instructions, but still uses the soft-float calling conventions. hard allows generation of floating-point instructions and uses FPU-specific calling conventions.

The default depends on the specific target configuration. Note that the hard-float and soft-float ABIs are not link-compatible; you must compile your entire program with the same ABI, and link with a compatible set of libraries. So don't mix them, bestie. 😤

• -mhard-float, -msoft-float — Select hardware or software floating-point implementations. The default is soft float. So software by default. 🧚
• -mdouble-float, -mno-double-float — When -mhard-float is in effect, enable generation of double-precision float instructions. This is the default except when compiling for CK803. So CK803 is the exception. 👀
• -mfdivdu, -mno-fdivdu — When -mhard-float is in effect, enable generation of "frecipd", "fsqrtd", and "fdivd" instructions. This is the default except when compiling for CK803. Again, CK803 says no. 💅
• -mfpu=fpu — Select the floating-point processor. This option can only be used with -mhard-float. Values for fpu are fpv2_sf (equivalent to -mno-double-float -mno-fdivdu), fpv2 (-mdouble-float -mno-divdu), and fpv2_divd (-mdouble-float -mdivdu). Choose your fighter. 🎀
• -melrw, -mno-elrw — Enable the extended "lrw" instruction. This option defaults to on for CK801 and off otherwise. So CK801 gets it by default, others don't. 🎀
• -mistack, -mno-istack — Enable interrupt stack instructions; the default is off.

The -mistack option is required to handle the "interrupt" and "isr" function attributes. So if you need those attributes, you need this. 🔥

• -mmp — Enable multiprocessor instructions; the default is off.
• -mcp — Enable coprocessor instructions; the default is off.
• -mcache — Enable coprocessor instructions; the default is off. (Yes, the original says "Enable coprocessor instructions" for -mcache too. It's a bit odd but we keep it exactly.)
• -msecurity — Enable C-SKY security instructions; the default is off.
• -mtrust — Enable C-SKY trust instructions; the default is off.
• -mdsp, -medsp, -mvdsp — Enable C-SKY DSP, Enhanced DSP, or Vector DSP instructions, respectively. All of these options default to off. DSP girlies, this one's for you. 💜
• -mdiv, -mno-div — Generate divide instructions. Default is off.
• -msmart, -mno-smart — Generate code for Smart Mode, using only registers numbered 0-7 to allow use of 16-bit instructions. This option is ignored for CK801 where this is the required behavior, and it defaults to on for CK802. For other targets, the default is off. Smart mode, cute. ✨
• -mhigh-registers, -mno-high-registers — Generate code using the high registers numbered 16-31. This option is not supported on CK801, CK802, or CK803, and is enabled by default for other processors. High registers? High key. 👑
• -manchor, -mno-anchor — Generate code using global anchor symbol addresses.
• -mpushpop, -mno-pushpop — Generate code using "push" and "pop" instructions. This option defaults to on. Push pop, push pop. 🎶
• -mmultiple-stld, -mstm, -mno-multiple-stld, -mno-stm — Generate code using "stm" and "ldm" instructions. This option isn't supported on CK801 but is enabled by default on other processors. So CK801 misses out. 😭
• -mconstpool, -mno-constpool — Create constant pools in the compiler instead of deferring it to the assembler. This option is the default and required for correct code generation on CK801 and CK802, and is optional on other processors. So CK801 and CK802 need it, others can choose. 💅
• -mstack-size, -mno-stack-size — emits .stack_size directives for each function in the assembly output. This one defaults to off, bestie.
• -mccrt, -mno-ccrt — generates code for the C-SKY compiler runtime instead of libgcc. Defaults to off.
• -mbranch-cost=n — sets the branch costs to roughly n instructions. The default is 1.
• -msched-prolog, -mno-sched-prolog — permits scheduling of function prologue and epilogue sequences.

🚨 Using this option can result in code that is not compliant with the C-SKY V2 ABI prologue requirements and that cannot be debugged or backtraced. It is disabled by default.

• -msim — links the library libsemi.a, which is compatible with simulator. Applicable to ELF compiler only.

---

✨ Cygwin and MinGW Options ✨

These extra options are for Microsoft Windows targets, bestie:

• -mconsole — specifies that a console application is to be generated, by instructing the linker to set the PE header subsystem type required for console applications. This option is available for Cygwin and MinGW targets and is enabled by default on those targets.
• -mcrtdll=library — preprocess, compile or link with the specified C RunTime DLL library. This option adjusts predefined macros __CRTDLL__, __MSVCRT__, _UCRT and __MSVCRT_VERSION__ for the specified CRT library, chooses the start file for the CRT library and links with the CRT library. Recognized CRT library names for the preprocessor are: crtdll*, msvcrt10*, msvcrt20*, msvcrt40*, msvcr40*, msvcrtd*, msvcrt-os*, msvcr70*, msvcr71*, msvcr80*, msvcr90*, msvcr100*, msvcr110*, msvcr120* and ucrt*.

If this option isn’t specified, then the default MinGW import library msvcrt is used for linking and no other adjustment for the preprocessor is done. MinGW import library msvcrt is just a symlink to (or a copy of) another MinGW CRT import library chosen during MinGW compilation. MinGW import library msvcrt-os is for the Windows system CRT DLL library msvcrt.dll and in most cases is the default MinGW import library. Generally speaking, changing the CRT DLL requires recompiling the entire MinGW CRT. This option is for experimental and testing purposes only. This option is available for MinGW targets.

• -mdll — available for Cygwin and MinGW targets. It specifies that a DLL—a dynamic link library—is to be generated, enabling the selection of the required runtime startup object and entry point.
• -mnop-fun-dllimport — available for Cygwin and MinGW targets. It specifies that the dllimport attribute should be ignored.
• -mthreads — available for MinGW targets. It specifies that MinGW-specific thread support is to be used.
• -municode — available for MinGW-w64 targets. It causes the UNICODE preprocessor macro to be predefined, and chooses Unicode-capable runtime startup code.
• -mwin32 — available for Cygwin and MinGW targets. It specifies that the typical Microsoft Windows predefined macros are to be set in the pre-processor, but does not influence the choice of runtime library/startup code.
• -mwindows — available for Cygwin and MinGW targets. It specifies that a GUI application is to be generated by instructing the linker to set the PE header subsystem type appropriately.
• -fno-set-stack-executable — available for MinGW targets. It specifies that the executable flag for the stack used by nested functions isn’t set. This is necessary for binaries running in kernel mode of Microsoft Windows, as there the User32 API, which is used to set executable privileges, isn’t available.
• -fwritable-relocated-rdata — available for MinGW and Cygwin targets. It specifies that relocated-data in read-only section is put into the .data section. This is necessary for older runtimes not supporting modification of .rdata sections for pseudo-relocation.
• -mpe-aligned-commons — available for Cygwin and MinGW targets. It specifies that the GNU extension to the PE file format that permits the correct alignment of COMMON variables should be used when generating code. It is enabled by default if GCC detects that the target assembler found during configuration supports the feature.

Also peek under x86 Options for standard options.

---

🍎 Darwin Options 🦋

These options are defined for all architectures running the Darwin operating system.

FSF GCC on Darwin does not create “fat” object files; it creates an object file for the single architecture that GCC was built to target. Apple’s GCC on Darwin does create “fat” files if multiple -arch options are used; it does so by running the compiler or linker multiple times and joining the results together with lipo.

The subtype of the file created (like ppc7400 or ppc970 or i686) is determined by the flags that specify the ISA that GCC is targeting, like -mcpu or -march. The -force_cpusubtype_ALL option can be used to override this.

The Darwin tools vary in their behavior when presented with an ISA mismatch. The assembler, as, only permits instructions to be used that are valid for the subtype of the file it is generating, so you cannot put 64-bit instructions in a ppc750 object file. The linker for shared libraries, /usr/bin/libtool, fails and prints an error if asked to create a shared library with a less restrictive subtype than its input files (for instance, trying to put a ppc970 object file in a ppc7400 library). The linker for executables, ld, quietly gives the executable the most restrictive subtype of any of its input files.

• -Fdir — add the framework directory dir to the head of the list of directories to be searched for header files. These directories are interleaved with those specified by -I options and are scanned in a left-to-right order.

A framework directory is a directory with frameworks in it. A framework is a directory with a Headers and/or PrivateHeaders directory contained directly in it that ends in .framework. The name of a framework is the name of this directory excluding the .framework. Headers associated with the framework are found in one of those two directories, with Headers being searched first. A subframework is a framework directory that is in a framework’s Frameworks directory. Includes of subframework headers can only appear in a header of a framework that contains the subframework, or in a sibling subframework header. Two subframeworks are siblings if they occur in the same framework. A subframework should not have the same name as a framework; a warning is issued if this is violated. Currently a subframework cannot have subframeworks; in the future, the mechanism may be extended to support this. The standard frameworks can be found in /System/Library/Frameworks and /Library/Frameworks. An example include looks like #include <Framework/header.h>, where Framework denotes the name of the framework and header.h is found in the PrivateHeaders or Headers directory.

• -iframeworkdir — like -F except the directory is treated as a system directory. The main difference between this -iframework and -F is that with -iframework the compiler does not warn about constructs contained within header files found via dir. This option is valid only for the C family of languages.
• -gused — emit debugging information for symbols that are used. For stabs debugging format, this enables -feliminate-unused-debug-symbols. This is by default ON.
• -gfull — emit debugging information for all symbols and types.
• -fconstant-cfstrings — this is an alias for -mconstant-cfstrings.
• -mconstant-cfstrings — when the NeXT runtime is being used (the default on these systems), override any -fconstant-string-class setting and cause @"..." literals to be laid out as constant CoreFoundation strings.
• -mmacosx-version-min=version — the earliest version of MacOS X that this executable will run on is version. Typical values supported for version include 12, 10.12, and 10.5.8.

If the compiler was built to use the system’s headers by default, then the default for this option is the system version on which the compiler is running, otherwise the default is to make choices that are compatible with as many systems and code bases as possible.

• -mkernel — enable kernel development mode. The -mkernel option sets -static, -fno-common, -fno-use-cxa-atexit, -fno-exceptions, -fno-non-call-exceptions, -fapple-kext, -fno-weak and -fno-rtti where applicable. This mode also sets -mno-altivec, -msoft-float, -fno-builtin and -mlong-branch for PowerPC targets.
• -mone-byte-bool — override the defaults for bool so that sizeof(bool)==1. By default sizeof(bool) is 4 when compiling for Darwin/PowerPC and 1 when compiling for Darwin/x86, so this option has no effect on x86.

⚠️ Warning: The -mone-byte-bool switch causes GCC to generate code that is not binary compatible with code generated without that switch. Using this switch may require recompiling all other modules in a program, including system libraries. Use this switch to conform to a non-default data model.

• -mfix-and-continue, -ffix-and-continue, -findirect-data — generate code suitable for fast turnaround development, such as to allow GDB to dynamically load .o files into already-running programs. -findirect-data and -ffix-and-continue are provided for backwards compatibility.
• -all_load — loads all members of static archive libraries. See man ld(1) for more information.
• -arch_errors_fatal — cause the errors having to do with files that have the wrong architecture to be fatal.
• -bind_at_load — causes the output file to be marked such that the dynamic linker will bind all undefined references when the file is loaded or launched.
• -bundle — produce a Mach-o bundle format file. See man ld(1) for more information.
• -bundle_loader executable — this option specifies the executable that will load the build output file being linked. See man ld(1) for more information.
• -dynamiclib — when passed this option, GCC produces a dynamic library instead of an executable when linking, using the Darwin libtool command.
• -force_cpusubtype_ALL — this makes GCC's output file carry the ALL subtype, instead of whatever subtype the -mcpu or -march option was controlling. 🎀
• -nodefaultrpaths — don't add the default run paths for the compiler library directories to executables, modules or dynamic libraries. On macOS 10.5 and later the embedded runpath gets added by default unless you add -nodefaultrpaths to the link line. And heads up bestie: run paths are needed — and therefore enforced — to build on macOS version 10.11 or later. 😤
• -allowable_client client_name — passed straight to the Darwin linker.
• -client_name — passed straight to the Darwin linker.
• -compatibility_version — passed straight to the Darwin linker.
• -current_version — passed straight to the Darwin linker.
• -dead_strip — passed straight to the Darwin linker.
• -dependency-file — passed straight to the Darwin linker.
• -dylib_file — passed straight to the Darwin linker.
• -dylinker_install_name — passed straight to the Darwin linker.
• -dynamic — passed straight to the Darwin linker.
• -exported_symbols_list — passed straight to the Darwin linker.
• -filelist — passed straight to the Darwin linker.
• -flat_namespace — passed straight to the Darwin linker.
• -force_flat_namespace — passed straight to the Darwin linker.
• -headerpad_max_install_names — passed straight to the Darwin linker.
• -image_base — passed straight to the Darwin linker.
• -init — passed straight to the Darwin linker.
• -install_name — passed straight to the Darwin linker.
• -keep_private_externs — passed straight to the Darwin linker.
• -multi_module — passed straight to the Darwin linker.
• -multiply_defined — passed straight to the Darwin linker.
• -multiply_defined_unused — passed straight to the Darwin linker.
• -noall_load — passed straight to the Darwin linker.
• -no_dead_strip_inits_and_terms — passed straight to the Darwin linker.
• -nofixprebinding — passed straight to the Darwin linker.
• -nomultidefs — passed straight to the Darwin linker.
• -noprebind — passed straight to the Darwin linker.
• -noseglinkedit — passed straight to the Darwin linker.
• -pagezero_size — passed straight to the Darwin linker.
• -prebind — passed straight to the Darwin linker.
• -prebind_all_twolevel_modules — passed straight to the Darwin linker.
• -private_bundle — passed straight to the Darwin linker.
• -read_only_relocs — passed straight to the Darwin linker.
• -sectalign — passed straight to the Darwin linker.
• -sectobjectsymbols — passed straight to the Darwin linker.
• -whyload — passed straight to the Darwin linker.
• -seg1addr — passed straight to the Darwin linker.
• -sectcreate — passed straight to the Darwin linker.
• -sectobjectsymbols — passed straight to the Darwin linker.
• -sectorder — passed straight to the Darwin linker.
• -segaddr — passed straight to the Darwin linker.
• -segs_read_only_addr — passed straight to the Darwin linker.
• -segs_read_write_addr — passed straight to the Darwin linker.
• -seg_addr_table — passed straight to the Darwin linker.
• -seg_addr_table_filename — passed straight to the Darwin linker.
• -seglinkedit — passed straight to the Darwin linker.
• -segprot — passed straight to the Darwin linker.
• -segs_read_only_addr — passed straight to the Darwin linker.
• -segs_read_write_addr — passed straight to the Darwin linker.
• -single_module — passed straight to the Darwin linker.
• -static — passed straight to the Darwin linker.
• -sub_library — passed straight to the Darwin linker.
• -sub_umbrella — passed straight to the Darwin linker.
• -twolevel_namespace — passed straight to the Darwin linker.
• -umbrella — passed straight to the Darwin linker.
• -undefined — passed straight to the Darwin linker.
• -unexported_symbols_list — passed straight to the Darwin linker.
• -weak_reference_mismatches — passed straight to the Darwin linker.
• -whatsloaded — passed straight to the Darwin linker.

So yeah, that whole lil squad right there? They all get handed off to the Darwin linker. The Darwin linker man page is the one that spills the tea on each of them in detail, so go read that if you need the full lore. 💅

---

🌸 DEC Alpha Options 🌸

These -m options are the ones defined for the DEC Alpha implementations, bestie:

• -mno-soft-float, -msoft-float — use (or don't use 🫶) the hardware floating-point instructions for floating-point operations. When -msoft-float is specified, functions in libgcc.a are used to perform floating-point operations. And here's the catch: unless those functions are replaced by routines that emulate the floating-point operations, or compiled in such a way as to call such emulation routines, those routines themselves issue floating-point operations. So if you're compiling for an Alpha without floating-point operations, you must make sure the library is built so it doesn't call them. 😭

Also note: Alpha implementations without floating-point operations are still required to have floating-point registers. Not optional, it's just the rules. 💀

• -mfp-reg, -mno-fp-regs — generate code that uses (does not use) the floating-point register set. -mno-fp-regs implies -msoft-float. If the floating-point register set isn't used, floating-point operands get passed in integer registers as if they were integers, and floating-point results get passed in $0 instead of $f0. That's a non-standard calling sequence, so any function with a floating-point argument or return value that's called by code compiled with -mno-fp-regs must also be compiled with that option. No cap.

A typical use of this option is building a kernel that doesn't use any floating-point registers, and therefore doesn't need to save and restore them. Main character energy honestly. 👑

• -mieee — the Alpha architecture implements floating-point hardware optimized for maximum performance. It's mostly compliant with the IEEE floating-point standard, but full compliance needs software assistance. This option generates fully IEEE-compliant code — except that the inexact-flag is not maintained (see below, bestie). If this option is turned on, the preprocessor macro _IEEE_FP is defined during compilation. The resulting code is less efficient, but it's able to correctly support denormalized numbers and exceptional IEEE values like not-a-number and plus/minus infinity. Other Alpha compilers call this option -ieee_with_no_inexact.

DEBIAN SPECIFIC: this option is on by default for alpha-linux-gnu, unless -ffinite-math-only (which is part of the -ffast-math set) is specified — because the software functions in the GNU libc math libraries generate denormalized numbers, NaNs, and infs, and all of those will cause a program to SIGFPE when it attempts to use the results without -mieee. So like, don't play with that. 😭

• -mieee-with-inexact — this is like -mieee, except the generated code also maintains the IEEE inexact-flag. Turning this option on makes the generated code implement fully-compliant IEEE math. On top of _IEEE_FP, _IEEE_FP_EXACT is defined as a preprocessor macro. Warning though: on some Alpha implementations the resulting code may execute significantly slower than the code generated by default. And since there's very little code that actually depends on the inexact-flag, you should normally not specify this option. Other Alpha compilers call this option -ieee_with_inexact.
• -mfp-trap-mode=trap-mode — this option controls which floating-point related traps are enabled. Other Alpha compilers call this option -fptm trap-mode. The trap mode can be set to one of four values:
  • n — this is the default (normal) setting. The only traps enabled are the ones that can't be disabled in software, e.g. the division by zero trap.
  • u — in addition to the traps enabled by n, underflow traps are enabled as well.
  • su — like u, but the instructions are marked to be safe for software completion (peep the Alpha architecture manual for the details 🧚).
  • sui — like su, but inexact traps are enabled as well.
• -mfp-rounding-mode=rounding-mode — selects the IEEE rounding mode. Other Alpha compilers call this option -fprm rounding-mode. The rounding-mode can be one of:
  • n — normal IEEE rounding mode. Floating-point numbers are rounded towards the nearest machine number, or towards the even machine number in case of a tie.
  • m — round towards minus infinity.
  • c — chopped rounding mode. Floating-point numbers are rounded towards zero.
  • d — dynamic rounding mode. A field in the floating-point control register (fpcr, see the Alpha architecture reference manual) controls the rounding mode in effect. The C library initializes this register for rounding towards plus infinity, so unless your program modifies the fpcr, d corresponds to round towards plus infinity. ✨
• -mtrap-precision=trap-precision — in the Alpha architecture, floating-point traps are imprecise. That means without software assistance it's impossible to recover from a floating trap, and program execution normally needs to be terminated. 💀 GCC can generate code that can assist operating system trap handlers in determining the exact location that caused a floating-point trap. Depending on what your application needs, you can pick different levels of precision:
  • p — program precision. This option is the default, and it means a trap handler can only identify which program caused a floating-point exception.
  • f — function precision. The trap handler can determine the function that caused a floating-point exception.
  • i — instruction precision. The trap handler can determine the exact instruction that caused a floating-point exception.

Other Alpha compilers provide the equivalent options called -scope_safe and -resumption_safe.

• -mieee-conformant — this option marks the generated code as IEEE conformant. You must not use this option unless you also specify -mtrap-precision=i and either -mfp-trap-mode=su or -mfp-trap-mode=sui. Its only effect is to emit the line .eflag 48 in the function prologue of the generated assembly file. That's it, that's the whole vibe. 🎀
• -mbuild-constants — normally GCC examines a 32- or 64-bit integer constant to see if it can construct it from smaller constants in two or three instructions. If it can't, it outputs the constant as a literal and generates code to load it from the data segment at run time. Use this option to require GCC to construct all integer constants using code, even if it takes more instructions — the maximum is six. You typically use this option to build a shared library dynamic loader: itself a shared library, it must relocate itself in memory before it can find the variables and constants in its own data segment. 🦋
• -mbwx, -mno-bwx — indicate whether GCC should generate code to use the optional BWX instruction set.
• -mcix, -mno-cix — indicate whether GCC should generate code to use the optional CIX instruction set.
• -mfix, -mno-fix — indicate whether GCC should generate code to use the optional FIX instruction set.
• -mmax, -mno-max — indicate whether GCC should generate code to use the optional MAX instruction set.

For that whole lil family: the default is to use the instruction sets supported by the CPU type specified via the -mcpu= option — or that of the CPU on which GCC was built, if none is specified. 🫶

• -msafe-bwa, -mno-safe-bwa — indicate whether, in the absence of the optional BWX instruction set, GCC should generate multi-thread and async-signal safe code for byte and aligned word memory accesses. 🔥
• -msafe-partial, -mno-safe-partial — indicate whether GCC should generate multi-thread and async-signal safe code for partial memory accesses, including piecemeal accesses to unaligned data, as well as block accesses to leading and trailing parts of aggregate types or other objects in memory that don't respectively start and end on an aligned 64-bit data boundary. Yeah, it's a mouthful, but the details matter. 💅
• -mfloat-vax, -mfloat-ieee — generate code that uses (does not use) VAX F and G floating-point arithmetic instead of IEEE single and double precision. Pick your fighter. 🌈
• -mexplicit-relocs, -mno-explicit-relocs — older Alpha assemblers provided no way to generate symbol relocations except via assembler macros, and using those macros doesn't allow optimal instruction scheduling. GNU binutils as of version 2.12 supports a new syntax that lets the compiler explicitly mark which relocations should apply to which instructions. This option is mostly useful for debugging, since GCC detects the capabilities of the assembler when it's built and sets the default accordingly. iykyk. 🙈
• -msmall-data, -mlarge-data — when -mexplicit-relocs is in effect, static data is accessed via gp-relative relocations. When -msmall-data is used, objects 8 bytes long or smaller are placed in a small data area (the .sdata and .sbss sections) and are accessed via 16-bit relocations off of the $gp register. This limits the size of the small data area to 64KB, but allows the variables to be directly accessed via a single instruction. ⭐

---

The default is -mlarge-data 💖 — with that vibe, your data area is limited to just below 2GB. Which is fine until it isn't, and if your program needs more than 2GB of data, it has to call malloc or mmap and put that data in the heap instead of in the program's data segment. No skimping bestie, them's the rules. ✨

And when you're generating code for shared libraries: -fpic implies -msmall-data and -fPIC implies -mlarge-data. Not me connecting the dots for you 🫶

• -msmall-text, -mlarge-text — okay so with -msmall-text the compiler straight up assumes the code of the whole entire program (or shared library) fits in 4MB, so it's reachable with a branch instruction. And when -msmall-data is used, the compiler can assume that all local symbols share the same $gp value, which means the number of instructions needed for a function call drops from 4 to 1. That's the glow up. 💅

The default is -mlarge-text.

• *-mcpu=cpu_type* — set the instruction set and the instruction scheduling parameters for machine type cpu_type. You can give either the EV style name or the corresponding chip number. GCC supports scheduling parameters for the EV4, EV5 and EV6 family of processors, and it picks the default values for the instruction set from the processor you specify. And if you don't specify a processor type at all? GCC just defaults to the processor on which the compiler was built. Main character energy fr fr. 👑

Supported values for cpu_type are:

• ev4
• ev45
• 21064 — schedules as an EV4 and has no instruction set extensions.
• ev5
• 21164 — schedules as an EV5 and has no instruction set extensions.
• ev56
• 21164a — schedules as an EV5 and supports the BWX extension.
• pca56
• 21164pc
• 21164PC — schedules as an EV5 and supports the BWX and MAX extensions.
• ev6
• 21264 — schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
• ev67
• 21264a — schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.

Native toolchains also support the value native, which selects the best architecture option for the host processor. Heads up tho: -mcpu=native has no effect if GCC does not recognize the processor. 🦋

• -mtune=cpu_type — set only the instruction scheduling parameters for machine type cpu_type. The instruction set is **not changed. Just the scheduling, bestie. 🌸

Native toolchains also support the value native, which selects the best architecture option for the host processor. And again, -mtune=native has no effect if GCC does not recognize the processor. 🙈

• *-mmemory-latency=time* — sets the latency the scheduler should assume for typical memory references as seen by the application. Lowkey important: this number is highly dependent on the memory access patterns used by the application and on the size of the external cache on the machine. So don't just vibes-pick it. 😭

Valid options for time are:

• number — a decimal number representing clock cycles.
• L1
• L2
• L3
• main — the compiler contains estimates of the number of clock cycles for "typical" EV4 & EV5 hardware for the Level 1, 2 & 3 caches (also called Dcache, Scache, and Bcache), as well as to main memory. Note that L3 is only valid for EV5. ⭐

---

eBPF Options 💖

• -mframe-limit=bytes — this specifies the **hard limit for frame sizes, in bytes. Currently, the value you can specify should be less than or equal to 32767. It defaults to whatever limit is imposed by the version of the Linux kernel you're targeting. 🎀
• -mbig-endian — generate code for a big-endian target.
• -mlittle-endian — generate code for a little-endian target. This is the default. 🧚
• -mjmpext, -mno-jmpext — enable or disable generation of extra conditional-branch instructions. Enabled for CPU v2 and above.
• -mjmp32, -mno-jmp32 — enable or disable generation of 32-bit jump instructions. Enabled for CPU v3 and above.
• -malu32, -mno-alu32 — enable or disable generation of 32-bit ALU instructions. Enabled for CPU v3 and above.
• -mv3-atomics, -mno-v3-atomics — enable or disable instructions for general atomic operations introduced in CPU v3. Enabled for CPU v3 and above.
• -mbswap, -mno-bswap — enable or disable byte swap instructions. Enabled for CPU v4 and above.
• -msdiv, -mno-sdiv — enable or disable signed division and modulus instructions. Enabled for CPU v4 and above.
• -msmov, -mno-smov — enable or disable sign-extending move and memory load instructions. Enabled for CPU v4 and above. (Yes, that second one is spelled -mno-smov here — I'm just telling you what the page says, not fixing it for you 💀)
• *-mcpu=version* — this specifies which version of the eBPF ISA to target. Newer versions may not be supported by all kernels. The default is v4. 👑

Supported values for version are:

• v1 — the first stable eBPF ISA with no special features or extensions.
• v2 — supports the jump extensions, as in -mjmpext.
• v3 — all features of v2, plus:
  • 32-bit jump operations, as in -mjmp32
  • 32-bit ALU operations, as in -malu32
  • general atomic operations, as in -mv3-atomics
• v4 — all features of v3, plus:
  • Byte swap instructions, as in -mbswap
  • Signed division and modulus instructions, as in -msdiv
  • Sign-extending move and memory load instructions, as in -msmov

• -mco-re — enable BPF Compile Once - Run Everywhere (CO-RE) support. Requires -gbtf, and is implied by -gbtf. 🫶
• -mno-co-re — disable BPF Compile Once - Run Everywhere (CO-RE) support. BPF CO-RE support is enabled by default when generating BTF debug information for the BPF target.
• -mxbpf — generate code for an expanded version of BPF, which relaxes some of the restrictions imposed by the BPF architecture:
  • Save and restore callee-saved registers at function entry and exit, respectively.
• *-masm=dialect* — outputs assembly instructions using the eBPF selected dialect. The default is pseudoc. ✨

Supported values for dialect are:

• normal — outputs normal assembly dialect.
• pseudoc — outputs pseudo-c assembly dialect.

• -minline-memops-threshold=bytes — specifies a size threshold in bytes at or below which memmove, memcpy and memset shall **always be expanded inline. Operations dealing with sizes larger than this threshold would have to be implemented using a library call instead of being expanded inline — but since BPF doesn't allow libcalls, exceeding this threshold results in a compile-time error. The default is 1024 bytes. 😭

---

FR30 Options 🌈

These options are defined specifically for the FR30 port.

• -msmall-model — use the small address space model. This can produce smaller code, but it does assume that all symbolic values and addresses fit into a 20-bit range. Small but spicy. 🔥
• -mno-lsim — assume that runtime support has been provided, and so there is no need to include the simulator library (libsim.a) on the linker command line.

---

FT32 Options 🦋

These options are defined specifically for the FT32 port.

• -msim — specifies that the program will be run on the simulator. This causes an alternate runtime startup and library to be linked.

Okay warning time bestie, for real: you must not use this option when generating programs that will run on real hardware. You must provide your own runtime library for whatever I/O functions are needed. Not a suggestion. 🚨

• -mlra — does nothing. Preserved for backward compatibility. It's giving legacy icon. 💅
• -mnodiv — do not use div and mod instructions.
• -mft32b — enable use of the extended instructions of the FT32B processor.
• -mcompress — compress all code using the Ft32B code compression scheme.
• -mnopm — do not generate code that reads program memory.

---

FRV Options 💜

• -mgpr-32 — only use the first 32 general-purpose registers.
• -mgpr-64 — use all 64 general-purpose registers.
• -mfpr-32 — use only the first 32 floating-point registers.
• -mfpr-64 — use all 64 floating-point registers.
• -mhard-float — use hardware instructions for floating-point operations.
• -msoft-float — use library routines for floating-point operations.
• -malloc-cc — dynamically allocate condition code registers.
• -mfixed-cc — do not try to dynamically allocate condition code registers; only use icc0 and fcc0.
• -mdword — change ABI to use double word insns.
• -mno-dword — do not use double word instructions.
• -mdouble — use floating-point double instructions.
• -mno-double — do not use floating-point double instructions.
• -mmedia — use media instructions.
• -mno-media — do not use media instructions.
• -mmuladd — use multiply and add/subtract instructions.
• -mno-muladd — do not use multiply and add/subtract instructions.
• -mfdpic — select the FDPIC ABI, which uses function descriptors to represent pointers to functions. Without any PIC/PIE-related options, it implies -fPIE. With -fpic or -fpie, it assumes GOT entries and small data are within a 12-bit range from the GOT base address; with -fPIC or -fPIE, GOT offsets are computed with 32 bits. And with a bfin-elf target, this option implies -msim. 🌟
• -minline-plt — enable inlining of PLT entries in function calls to functions that are not known to bind locally. It has no effect without -mfdpic. It's enabled by default if optimizing for speed and compiling for shared libraries (i.e., -fPIC or -fpic), or when an optimization option such as -O3 or above is present in the command line. 👑
• -mTLS — assume a large TLS segment when generating thread-local code.
• -mtls — do not assume a large TLS segment when generating thread-local code.

(Yes, those two differ only by case. Look very carefully. 🙈)

• -mgprel-ro — enable the use of GPREL relocations in the FDPIC ABI for data that is known to be in read-only sections. It's enabled by default, except for -fpic or -fpie: even though it may help make the global offset table smaller, it trades 1 instruction for 4. With -fPIC or -fPIE, it trades 3 instructions for 4, one of which may be shared by multiple symbols, and it avoids the need for a GOT entry for the referenced symbol, so it's more likely to be a win. If it is not, -mno-gprel-ro can be used to disable it. 💅
• -multilib-library-pic — link with the (library, not FD) pic libraries. It's implied by -mlibrary-pic, as well as by -fPIC and -fpic without -mfdpic. You should never have to use it explicitly. 🎀
• -mlinked-fp — bestie, this one follows the EABI requirement where a frame pointer gets created every single time a stack frame is allocated. It's on by default ✨, and you can turn it off with -mno-linked-fp.
• -mlong-calls — uses indirect addressing to call functions that live outside the current compilation unit. This means those functions can be placed literally anywhere within the 32-bit address space. 🦋
• -malign-labels — tries to align labels to an 8-byte boundary by inserting NOPs into the previous packet. Heads up: this option only does anything when VLIW packing is enabled. It doesn't create new packets — it just adds NOPs to the ones that already exist. 💅
• -mlibrary-pic — generates position-independent EABI code. Periodt.
• -macc-4 — use only the first four media accumulator registers. That's it, that's the whole vibe.
• -macc-8 — use all eight media accumulator registers. Go big or go home. 👑
• -mpack — pack VLIW instructions.
• -mno-pack — do not pack VLIW instructions. The anti-packing agenda, iconic.
• -mno-eflags — do not mark ABI switches in e_flags.
• -mcond-move — enables the use of conditional-move instructions (this is the default).

This switch is mainly for debugging the compiler and will likely be removed in a future version. 😭

• -mno-cond-move — disables the use of conditional-move instructions.

This switch is mainly for debugging the compiler and will likely be removed in a future version. 😭

• -mscc — enables the use of conditional set instructions (default).

This switch is mainly for debugging the compiler and will likely be removed in a future version. 😭

• -mno-scc — disables the use of conditional set instructions.

This switch is mainly for debugging the compiler and will likely be removed in a future version. 😭

• -mcond-exec — enables the use of conditional execution (default).

This switch is mainly for debugging the compiler and will likely be removed in a future version. 😭

• -mno-cond-exec — disables the use of conditional execution.

This switch is mainly for debugging the compiler and will likely be removed in a future version. 😭

• -mvliw-branch — runs a pass to pack branches into VLIW instructions (default).

This switch is mainly for debugging the compiler and will likely be removed in a future version. 😭

• -mno-vliw-branch — do not run a pass to pack branches into VLIW instructions.

This switch is mainly for debugging the compiler and will likely be removed in a future version. 😭

• -mmulti-cond-exec — enables optimization of "&&" and "||" in conditional execution (default).

This switch is mainly for debugging the compiler and will likely be removed in a future version. 😭

• -mno-multi-cond-exec — disables optimization of "&&" and "||" in conditional execution.

This switch is mainly for debugging the compiler and will likely be removed in a future version. 😭

• -mnested-cond-exec — enables nested conditional execution optimizations (default).

This switch is mainly for debugging the compiler and will likely be removed in a future version. 😭

• -mno-nested-cond-exec — disables nested conditional execution optimizations.

This switch is mainly for debugging the compiler and will likely be removed in a future version. 😭

• -moptimize-membar — this switch removes redundant "membar" instructions from the compiler-generated code. It's enabled by default. ✨
• -mno-optimize-membar — this switch disables the automatic removal of redundant "membar" instructions from the generated code.
• -mtomcat-stats — causes gas to print out tomcat statistics. 🐱 stats, slay.
• *-mcpu=cpu* — select the processor type you want code generated for. The possible values are frv, fr550, tomcat, fr500, fr450, fr405, fr400, fr300 and simple. Pick your fighter. 💖

🐧 GNU/Linux Options 🐧

These -m options are defined for GNU/Linux targets:

• -mglibc — use the GNU C library. This is the default except on *-*-linux-*uclibc*, *-*-linux-*musl* and *-*-linux-*android* targets.
• -muclibc — use uClibc C library. This is the default on *-*-linux-*uclibc* targets.
• -mmusl — use the musl C library. This is the default on *-*-linux-*musl* targets.
• -mbionic — use Bionic C library. This is the default on *-*-linux-*android* targets.
• -mandroid — compile code compatible with the Android platform. This is the default on *-*-linux-*android* targets.

When compiling, this option enables -mbionic, -fPIC, -fno-exceptions and -fno-rtti by default. When linking, this option makes the GCC driver pass Android-specific options to the linker. And finally, this option causes the preprocessor macro "__ANDROID__" to be defined. 🤖

• -tno-android-cc — disable compilation effects of -mandroid, i.e., do not enable -mbionic, -fPIC, -fno-exceptions and -fno-rtti by default.
• -tno-android-ld — disable linking effects of -mandroid, i.e., pass standard Linux linking options to the linker.

🎀 H8/300 Options 🎀

These -m options are defined for the H8/300 implementations:

• -mrelax — shorten some address references at link time, when possible; uses the linker option -relax. Very demure, very mindful.
• -mh — generate code for the H8/300H.
• -ms — generate code for the H8S.
• -mn — generate code for the H8S and H8/300H in the normal mode. Bestie, this switch must be used either with -mh or -ms. No solo runs. 💅
• -ms2600 — generate code for the H8S/2600. This switch must be used with -ms.
• -mexr — extended registers are stored on stack before execution of function with monitor attribute. Default option is -mexr. This option is valid only for H8S targets.
• -mno-exr — extended registers are not stored on stack before execution of function with monitor attribute. Default option is -mno-exr. This option is valid only for H8S targets.
• -mint32 — make "int" data 32 bits by default.
• -malign-300 — on the H8/300H and H8S, use the same alignment rules as for the H8/300. The default for the H8/300H and H8S is to align longs and floats on 4-byte boundaries. -malign-300 causes them to be aligned on 2-byte boundaries. This option has no effect on the H8/300. 🦋

💜 HPPA Options 💜

These -m options are defined for the HPPA family of computers:

• *-march=architecture-type* — generate code for the specified architecture. The choices for architecture-type are 1.0 for PA 1.0, 1.1 for PA 1.1, and 2.0 for PA 2.0 processors. Refer to /usr/lib/sched.models on an HP-UX system to determine the proper architecture option for your machine. Code compiled for lower numbered architectures runs on higher numbered architectures, but not the other way around. ⚠️
• -mpa-risc-1-0
• -mpa-risc-1-1
• -mpa-risc-2-0 — synonyms for -march=1.0, -march=1.1, and -march=2.0 respectively.
• -matomic-libcalls — generate libcalls for atomic loads and stores when sync libcalls are disabled. This option is enabled by default. It only affects the generation of atomic libcalls by the HPPA backend.

Both the sync and libatomic libcall implementations use locking. As a result, processor stores are not atomic with respect to other atomic operations. Processor loads up to DImode are atomic with respect to other atomic operations provided they are implemented as a single access.

The PA-RISC architecture does not support any atomic operations in hardware except for the "ldcw" instruction. Thus, all atomic support is implemented using sync and atomic libcalls. Sync libcall support is in libgcc.a. Atomic libcall support is in libatomic.

This option generates "__atomic_exchange" calls for atomic stores. It also provides special handling for atomic DImode accesses on 32-bit targets. 🔥

• -mbig-switch — does nothing. Preserved for backward compatibility. The way this option literally does nothing... iconic behavior honestly. 💀
• -mcaller-copies — the caller copies function arguments passed by hidden reference. This option should be used with care as it is not compatible with the default 32-bit runtime. However, only aggregates larger than eight bytes are passed by hidden reference and the option provides better compatibility with OpenMP. ⚠️
• -mcoherent-ldcw — use ldcw/ldcd coherent cache-control hint.
• -mdisable-fpregs — disable floating-point registers. Equivalent to "-msoft-float".
• -mdisable-indexing — prevent the compiler from using indexing address modes. This avoids some rather obscure problems when compiling MIG generated code under MACH.
• -mfast-indirect-calls — generate code that assumes calls never cross space boundaries. This allows GCC to emit code that performs faster indirect calls.

This option does not work in the presence of shared libraries or nested functions. 💀

• *-mfixed-range=register-range* — generate code treating the given register range as fixed registers. A fixed register is one that the register allocator cannot use. This is useful when compiling kernel code. A register range is specified as two registers separated by a dash. Multiple register ranges can be specified separated by a comma.
• -mgas — enable the use of assembler directives only GAS understands.
• -mgnu-ld — use options specific to GNU ld. This passes -shared to ld when building a shared library. It is the default when GCC is configured, explicitly or implicitly, with the GNU linker. This option does not affect which ld is called; it only changes what parameters are passed to that ld. The ld that is called is determined by the --with-ld configure option, GCC's program search path, and finally by the user's PATH. The linker used by GCC can be printed using which gcc -print-prog-name=ld. This option is only available on the 64-bit HP-UX GCC, i.e. configured with hppa*64*-*-hpux*. 🫶
• -mhp-ld — use options specific to HP ld. This passes -b to ld when building a shared library and passes +Accept TypeMismatch to ld on all links. It is the default when GCC is configured, explicitly or implicitly, with the HP linker. This option does not affect which ld is called; it only changes what parameters are passed to that ld. The ld that is called is determined by the --with-ld configure option, GCC's program search path, and finally by the user's PATH. The linker used by GCC can be printed using which gcc -print-prog-name=ld. This option is only available on the 64-bit HP-UX GCC, i.e. configured with hppa*64*-*-hpux*. 🫶
• -mlinker-opt — enable the optimization pass in the HP-UX linker. Note this makes symbolic debugging impossible. ⚠️

---

• -mlong-calls — okay bestie, this one makes GCC emit long call sequences, so a call can always reach linker-generated stubs, no matter what. 💅 By default it's way more chill: long calls only happen when the distance from the call site to the beginning of the function (or the translation unit, depending on the situation) blows past a predefined limit set by the branch type in play. Those limits for normal calls are 7,600,000 and 240,000 bytes, respectively, for the PA 2.0 and PA 1.X architectures. Sibcalls? Always capped at 240,000 bytes. no exceptions, periodt.

Distances get measured from the beginning of functions when you're using -ffunction-sections, or when you're using -mgas and -mno-portable-runtime together under HP-UX with the SOM linker. iykyk.

fr fr though, you normally do not want this option — it degrades performance, and that's not a vibe. 🙈 But it can be useful in large applications, particularly when partial linking is used to build the application.

The types of long calls used depend on the capabilities of the assembler and the linker, and the type of code being generated. On systems that support long absolute calls, and long pic symbol-difference or pc-relative calls, the impact should be relatively small. However, an indirect call is used on 32-bit ELF systems in pic code, and it's quite long. 😭

• -mlong-load-store — generate 3-instruction load and store sequences, as sometimes required by the HP-UX 10 linker. This is equivalent to the +k option to the HP compilers. 🫶
• -mjump-in-delay — this option is ignored and provided for compatibility purposes only. it's giving decorative tbh 💀
• -mno-space-regs — generate code that assumes the target has no space registers. This lets GCC generate faster indirect calls and use unscaled index address modes. ✨ Such code is suitable for level 0 PA systems and kernels.
• -mordered — assume memory references are ordered and barriers are not needed. 🔥
• -mportable-runtime — use the portable calling conventions proposed by HP for ELF systems.
• -mschedule=cpu-type — schedule code according to the constraints for the machine type cpu-type. The choices for cpu-type are 700 7100, 7100LC, 7200, 7300 and 8000. Refer to /usr/lib/sched.models on an HP-UX system to determine the proper scheduling option for your machine. The default scheduling is 8000. 👑
• -msio — generate the predefine, _SIO, for server IO. The default is -mwsio. This generates the predefines, __hp9000s700, __hp9000s700__ and _WSIO, for workstation IO. These options are available under HP-UX and HI-UX.
• -msoft-float — generate output containing library calls for floating point.

⚠️ Warning: the requisite libraries are not available for all HPPA targets. Normally the facilities of the machine's usual C compiler are used, but this cannot be done directly in cross-compilation. You must make your own arrangements to provide suitable library functions for cross-compilation. you have been warned, bestie 🫠

-msoft-float changes the calling convention in the output file; therefore, it is only useful if you compile all of a program with this option. In particular, you need to compile libgcc.a, the library that comes with GCC, with -msoft-float in order for this to work. all or nothing, babe. 💖

• -msoft-mult — use software integer multiplication. This disables the use of the xmpyu instruction.
• -munix=unix-std — generate compiler predefines and select a startfile for the specified UNIX standard. The choices for unix-std are 93, 95 and 98. 93 is supported on all HP-UX versions. 95 is available on HP-UX 10.10 and later. 98 is available on HP-UX 11.11 and later. The default values are 93 for HP-UX 10.00, 95 for HP-UX 10.10 through to 11.00, and 98 for HP-UX 11.11 and later. ✨

-munix=93 provides the same predefines as GCC 3.3 and 3.4. -munix=95 provides additional predefines for XOPEN_UNIX and _XOPEN_SOURCE_EXTENDED, and the startfile unix95.o. -munix=98 provides additional predefines for _XOPEN_UNIX, _XOPEN_SOURCE_EXTENDED, _INCLUDE__STDC_A1_SOURCE and _INCLUDE_XOPEN_SOURCE_500, and the startfile unix98.o.

😳 It is important to note that this option changes the interfaces for various library routines. It also affects the operational behavior of the C library. Thus, extreme care is needed in using this option. like, be so for real careful.

Library code that is intended to operate with more than one UNIX standard must test, set and restore the variable __xpg4_extended_mask as appropriate. Most GNU software doesn't provide this capability. 🙈

• -nolibdld — suppress the generation of link options to search libdld.sl when the -static option is specified on HP-UX 10 and later.
• -static — the HP-UX implementation of setlocale in libc has a dependency on libdld.sl. There isn't an archive version of libdld.sl. Thus, when the -static option is specified, special link options are needed to resolve this dependency.

On HP-UX 10 and later, the GCC driver adds the necessary options to link with libdld.sl when the -static option is specified. This causes the resulting binary to be dynamic. On the 64-bit port, the linkers generate dynamic binaries by default in any case. The -nolibdld option can be used to prevent the GCC driver from adding these link options. 🎀

• -threads — add support for multithreading with the dce thread library under HP-UX. This option sets flags for both the preprocessor and the linker. 🧵✨

---

🌈 IA-64 Options 🌈

These are the -m options defined for the Intel IA-64 architecture. buckle up bestie, there's a lot 💜

• -mbig-endian — generate code for a big-endian target. This is the default for HP-UX.
• -mlittle-endian — generate code for a little-endian target. This is the default for AIX5 and GNU/Linux. 🐧
• -mgnu-as / -mno-gnu-as — generate (or don't) code for the GNU assembler. This is the default.
• -mgnu-ld / -mno-gnu-ld — generate (or don't) code for the GNU linker. This is the default.
• -mno-pic — generate code that does not use a global pointer register. The result is not position independent code, and it violates the IA-64 ABI. 💀 noted.
• -mvolatile-asm-stop / -mno-volatile-asm-stop — generate (or don't) a stop bit immediately before and after volatile asm statements.
• -mregister-names / -mno-register-names — generate (or don't) in, loc, and out register names for the stacked registers. This may make assembler output more readable. 👑
• -mno-sdata / -msdata — disable (or enable) optimizations that use the small data section. This may be useful for working around optimizer bugs. 🐛
• -mconstant-gp — generate code that uses a single constant global pointer value. This is useful when compiling kernel code.
• -mauto-pic — generate code that is self-relocatable. This implies -mconstant-gp. This is useful when compiling firmware code. ✨
• -minline-float-divide-min-latency — generate code for inline divides of floating-point values using the minimum latency algorithm.
• -minline-float-divide-max-throughput — generate code for inline divides of floating-point values using the maximum throughput algorithm. 🔥
• -mno-inline-float-divide — do not generate inline code for divides of floating-point values.
• -minline-int-divide-min-latency — generate code for inline divides of integer values using the minimum latency algorithm.
• -minline-int-divide-max-throughput — generate code for inline divides of integer values using the maximum throughput algorithm.
• -mno-inline-int-divide — do not generate inline code for divides of integer values.
• -minline-sqrt-min-latency — generate code for inline square roots using the minimum latency algorithm.
• -minline-sqrt-max-throughput — generate code for inline square roots using the maximum throughput algorithm.
• -mno-inline-sqrt — do not generate inline code for sqrt.
• -mfused-madd / -mno-fused-madd — do (don't) generate code that uses the fused multiply/add or multiply/subtract instructions. The default is to use these instructions.
• -mno-dwarf2-asm / -mdwarf2-asm — don't (or do) generate assembler code for the DWARF line number debugging info. This may be useful when not using the GNU assembler.
• -mearly-stop-bits / -mno-early-stop-bits — allow stop bits to be placed earlier than immediately preceding the instruction that triggered the stop bit. This can improve instruction scheduling, but does not always do so. 🙃
• -mfixed-range=register-range — generate code treating the given register range as fixed registers. A fixed register is one that the register allocator cannot use. This is useful when compiling kernel code. A register range is specified as two registers separated by a dash. Multiple register ranges can be specified separated by a comma. 💅
• -mtls-size=tls-size — specify bit size of immediate TLS offsets. Valid values are 14, 22, and 64.
• -mtune=cpu-type — tune the instruction scheduling for a particular CPU. Valid values are itanium, itanium1, merced, itanium2, and mckinley.
• -milp32 / -mlp64 — generate code for a 32-bit or 64-bit environment. The 32-bit environment sets int, long and pointer to 32 bits. The 64-bit environment sets int to 32 bits and long and pointer to 64 bits. These are HP-UX specific flags. 🎀
• -mno-sched-br-data-spec / -msched-br-data-spec — (dis/en)able data speculative scheduling before reload. This results in generation of ld.a instructions and the corresponding check instructions (ld.c / chk.a). The default setting is disabled.
• -msched-ar-data-spec / -mno-sched-ar-data-spec — (en/dis)able data speculative scheduling after reload. This results in generation of ld.a instructions and the corresponding check instructions (ld.c / chk.a). The default setting is enabled. ✨
• -mno-sched-control-spec / -msched-control-spec — (dis/en)able control speculative scheduling. This feature is available only during region scheduling (i.e. before reload). This results in generation of the ld.s instructions and the corresponding check instructions chk.s. The default setting is disabled.
• -msched-br-in-data-spec, -mno-sched-br-in-data-spec — turns speculative scheduling of the instructions that depend on data speculative loads before reload on (or off). Only actually does anything when -msched-br-data-spec is enabled, bestie. Default: enabled ✨
• -msched-ar-in-data-spec, -mno-sched-ar-in-data-spec — same deal but for the instructions that depend on data speculative loads after reload. Effective only with -msched-ar-data-spec enabled. Default: enabled 💖
• -msched-in-control-spec, -mno-sched-in-control-spec — (en/dis)ables speculative scheduling of instructions dependent on control speculative loads. Only works when -msched-control-spec is enabled. Default: enabled 🎀
• -mno-sched-prefer-non-data-spec-insns, -msched-prefer-non-data-spec-insns — when this one's on, data-speculative instructions only get picked for the schedule if there's literally nothing else available. Makes data speculation way more conservative, like it's checking its vibe before committing. Default: disabled 🫶
• -mno-sched-prefer-non-control-spec-insns, -msched-prefer-non-control-spec-insns — same energy but for control-speculative instructions — they only get scheduled if there are no other choices at that moment, so control speculation stays conservative. Default: disabled 💅
• -mno-sched-count-spec-in-critical-path, -msched-count-spec-in-critical-path — when enabled, speculative dependencies get counted while the instruction priorities are being computed, which makes the speculation a lil more conservative. Default: disabled 😌
• -msched-spec-ldc — use a simple data speculation check. This one is on by default. 💖
• -msched-control-spec-ldc — use a simple check for control speculation. Also on by default. ✨
• -msched-stop-bits-after-every-cycle — drop a stop bit after every cycle when scheduling, periodt. On by default. 🦋
• -msched-fp-mem-deps-zero-cost — assume floating-point stores and loads probably won't conflict when they land in the same instruction group. Disabled by default. 🌸
• -msel-sched-dont-check-control-spec — generate checks for control speculation in selective scheduling. This flag is disabled by default. 🙈
• *-msched-max-memory-insns=max-insns* — caps how many memory insns can go in one instruction group, and gives lower priority to any subsequent memory insns trying to schedule into that same group. Genuinely handy for dodging cache bank conflicts. Default value is 1. 🍓
• -msched-max-memory-insns-hard-limit — makes the limit from msched-max-memory-insns a hard limit, so nothing over that number gets into an instruction group, no exceptions. Without it, the limit is "soft" — non-memory operations get preference once the limit is hit, but memory operations can still be scheduled. 😭

---

🌸 LM32 Options 🌸

These -m options are the ones defined for the LatticeMico32 architecture, bestie:

• -mbarrel-shift-enabled — enable barrel-shift instructions. 💫
• -mdivide-enabled — enable divide and modulus instructions. ➗
• -mmultiply-enabled — enable multiply instructions. ✖️
• -msign-extend-enabled — enable sign extend instructions. 🫶
• -muser-enabled — enable user-defined instructions. 👑

---

✨ LoongArch Options ✨

These command-line options are the ones defined for LoongArch targets:

• *-march=arch-type* — generate instructions for the machine type arch-type. And listen: -march=arch-type lets GCC generate code that may not run at all on processors other than the one you named. It's giving commitment issues, fr fr.

The choices for arch-type are:

• native — local processor type detected by the native compiler. 💅
• loongarch64 — generic LoongArch 64-bit processor. ✨
• la464 — LoongArch LA464-based processor with LSX, LASX. 🔥
• la664 — LoongArch LA664-based processor with LSX, LASX and all LoongArch v1.1 instructions. 👑
• la64v1.0 — LoongArch64 ISA version 1.0. 🎀
• la64v1.1 — LoongArch64 ISA version 1.1. 🦋

More information about LoongArch ISA versions can be found at <https://github.com/loongson/la-toolchain-conventions>. 🌈

• *-mtune=tune-type* — optimize the generated code for the given processor target.

The choices for tune-type are:

• native — local processor type detected by the native compiler. 💜
• generic — generic LoongArch processor. 🌸
• loongarch64 — generic LoongArch 64-bit processor. ✨
• la464 — LoongArch LA464 core. 💅
• la664 — LoongArch LA664 core. 🔥

• *-mabi=base-abi-type* — generate code for the specified calling convention. base-abi-type can be one of:
  • lp64d — uses 64-bit general purpose registers and 32/64-bit floating-point registers for parameter passing. Data model is LP64, where int is 32 bits, while long int and pointers are 64 bits. 💖
  • lp64f — uses 64-bit general purpose registers and 32-bit floating-point registers for parameter passing. Data model is LP64, where int is 32 bits, while long int and pointers are 64 bits. 🎀
  • lp64s — uses 64-bit general purpose registers and no floating-point registers for parameter passing. Data model is LP64, where int is 32 bits, while long int and pointers are 64 bits. 🫶
• *-mfpu=fpu-type* — generate code for the specified FPU type, which can be one of:
  • 64 — allow the use of hardware floating-point instructions for 32-bit and 64-bit operations. ✨
  • 32 — allow the use of hardware floating-point instructions for 32-bit operations. 🌸
  • none, 0 — prevent the use of hardware floating-point instructions. 🙈
• *-msimd=simd-type* — enable generation of LoongArch SIMD instructions for vectorization and via builtin functions. The value can be one of:
  • lasx — enable generating instructions from the 256-bit LoongArch Advanced SIMD Extension (LASX) and the 128-bit LoongArch SIMD Extension (LSX). 🔥
  • lsx — enable generating instructions from the 128-bit LoongArch SIMD Extension (LSX). 💅
  • none — no LoongArch SIMD instruction may be generated. 💀
• -msoft-float — forces -mfpu=none and prevents the use of floating-point registers for parameter passing. Heads up bestie: this option may change the target ABI. 😭
• -msingle-float — forces -mfpu=32 and allows the use of 32-bit floating-point registers for parameter passing. This option may change the target ABI too. 🦋
• -mdouble-float — forces -mfpu=64 and allows the use of 32/64-bit floating-point registers for parameter passing. Yep, this one may also change the target ABI. ✨
• -mlasx, -mno-lasx, -mlsx, -mno-lsx — incrementally adjust the scope of the SIMD extensions (none / LSX / LASX) the compiler is allowed to use for code generation. Enabling LASX with mlasx automatically enables LSX, and disabling LSX with mno-lsx automatically disables LASX — they're a package deal, iykyk. These driver-only options act upon the final msimd configuration state and make incremental changes in the order they appear on the GCC driver's command line, deriving the final / canonicalized msimd option that gets passed to the compiler proper. 👑
• *-mbranch-cost=n* — set the cost of branches to roughly n instructions. 🧮
• *-maddr-reg-reg-cost=n* — set the cost of ADDRESS_REG_REG to the value calculated by n. 🧚
• -mcheck-zero-division, -mno-check-zero-divison — trap (or don't trap) on integer division by zero. Default is -mcheck-zero-division for -O0 or -Og, and -mno-check-zero-division for the other optimization levels. ⚠️
• -mcond-move-int, -mno-cond-move-int — conditional moves for integral data in general-purpose registers are enabled (disabled). Default is -mcond-move-int. 💅
• -mcond-move-float, -mno-cond-move-float — conditional moves for floating-point registers are enabled (disabled). Default is -mcond-move-float. ✨
• -mmemcpy, -mno-memcpy — force (or don't force) the use of memcpy for non-trivial block moves. Default is -mno-memcpy, which lets GCC inline most constant-sized copies. Setting the optimization level to -Os also forces the use of memcpy, but -mno-memcpy may override this behavior if explicitly specified, regardless of the order these options appear on the command line. 🎀
• -mstrict-align, -mno-strict-align — avoid or allow generating memory accesses that may not be aligned on a natural object boundary as described in the architecture specification. Default is -mno-strict-align. 🦋
• *-G *num — put global and static data smaller than num bytes into a small data section. Default value is 0. 🌸
• *-mmax-inline-memcpy-size=n* — inline all block moves (such as calls to memcpy or structure copies) less than or equal to n bytes. Default value of n is 1024. 💖
• *-mcmodel=code-model* — set the code model to one of:
  • tiny-static (Not implemented yet)
  • tiny (Not implemented yet)
  • normal — the text segment must be within 128MB addressing space. The data segment must be within 2GB addressing space. ✨
  • medium — the text segment and data segment must be within 2GB addressing space. 💅
  • large (Not implemented yet)
  • extreme — this mode does not limit the size of the code segment and data segment. Heads up: the -mcmodel=extreme option is incompatible with -fplt and/or -mexplicit-relocs=none. 😭

The default code model is normal. 👑

• -mexplicit-relocs=style — okay this one decides when GCC uses assembler relocation operators for symbolic addresses. The other option is assembler macros, which can cramp instruction scheduling but let the linker relax stuff. -mexplicit-relocs=none → macros always. -mexplicit-relocs=always → relocation operators always. -mexplicit-relocs=auto → operators where linker relaxation is impossible to improve code quality, macros everywhere else. What's the default? Bestie it's decided at GCC build-time by detecting what your assembler can do, plus whatever -mrelax is set to: -mexplicit-relocs=none if the assembler doesn't support relocation operators at all, -mexplicit-relocs=always if it does support them but -mrelax isn't enabled, and -mexplicit-relocs=auto if it supports them and -mrelax is on. ✨
• -mexplicit-relocs — literally just an alias of -mexplicit-relocs=always, kept around for backward compatibility 🫶
• -mno-explicit-relocs — and this one is an alias of -mexplicit-relocs=none, also for backward compatibility 💅
• -mdirect-extern-access, -mno-direct-extern-access — do not use or use GOT to access external symbols. Default is -mno-direct-extern-access, so GOT gets used for external symbols with default visibility, but not for other external symbols.

With -mdirect-extern-access, GOT is not used and all external symbols are PC-relatively addressed. It's only suitable for environments where no dynamic link is performed — think firmwares, OS kernels, executables linked with -static or -static-pie. And heads up: -mdirect-extern-access is not compatible with -fPIC or -fpic. 🚫

• -mrelax, -mno-relax — take (or do not take) advantage of linker relaxations. If -mpass-mrelax-to-as is enabled, this option is also passed to the assembler. The default is determined during GCC build-time by detecting corresponding assembler support: -mrelax if the assembler supports both the -mrelax option and conditional branch relaxation — and yeah that one's required, otherwise the ".align" directives and conditional branch instructions in the assembly code outputted by GCC may be rejected by the assembler because of a relocation overflow 😭 — and -mno-relax otherwise.
• -mpass-mrelax-to-as, -mno-pass-mrelax-to-as — pass (do not pass) the -mrelax or -mno-relax option to the assembler. Default comes from GCC build-time detection: -mpass-mrelax-to-as if the assembler supports the -mrelax option, -mno-pass-mrelax-to-as otherwise. This one is mostly for debugging, or for interoperating with assemblers that aren't the build-time one. 🔧
• -mrecip — enables use of the reciprocal estimate and reciprocal square root estimate instructions with extra Newton-Raphson steps to boost precision, instead of doing a divide or a square root and divide for floating-point arguments. These instructions are generated only when -funsafe-math-optimizations is enabled together with -ffinite-math-only and -fno-trapping-math. Off by default, periodt. Before you use this you must be sure your target CPU supports frecipe and frsqrte instructions — no cap. And note that while the throughput of the sequence is higher than the throughput of the non-reciprocal instruction, the precision of the sequence can be decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994) 💀
• -mrecip=opt — controls which reciprocal estimate instructions may be used. opt is a comma-separated list of options, and each one may be preceded by a ! to invert it:
  • all — enable all estimate instructions.
  • default — enable the default instructions, equivalent to -mrecip.
  • none — disable all estimate instructions, equivalent to -mno-recip.
  • div — enable the approximation for scalar division.
  • vec-div — enable the approximation for vectorized division.
  • sqrt — enable the approximation for scalar square root.
  • vec-sqrt — enable the approximation for vectorized square root.
  • rsqrt — enable the approximation for scalar reciprocal square root.
  • vec-rsqrt — enable the approximation for vectorized reciprocal square root.

So like, for example, -mrecip=all,!sqrt enables all of the reciprocal approximations, except for scalar square root. Cute, right? 🎀

• -mfrecipe, -mno-frecipe — use (do not use) "frecipe.{s/d}" and "frsqrte.{s/d}" instructions. When build with -march=la664, it is enabled by default. The default is -mno-frecipe.
• -mdiv32, -mno-div32 — use (do not use) "div.w[u]" and "mod.w[u]" instructions with input not sign-extended. When build with -march=la664, it is enabled by default. The default is -mno-div32.
• -mlam-bh, -mno-lam-bh — use (do not use) "am{swap/add}[_db].{b/h}" instructions. When build with -march=la664, it is enabled by default. The default is -mno-lam-bh.
• -mlamcas, -mno-lamcas — use (do not use) "amcas[_db].{b/h/w/d}" instructions. When build with -march=la664, it is enabled by default. The default is -mno-lamcas.
• -mld-seq-sa, -mno-ld-seq-sa — whether a same-address load-load barrier ("dbar 0x700") is needed. When build with -march=la664, it is enabled by default. The default is -mno-ld-seq-sa, meaning the load-load barrier is needed.
• -mtls-dialect=opt — controls which tls dialect may be used for general dynamic and local dynamic TLS models.
  • trad — use traditional TLS. This is the default.
  • desc — use TLS descriptors.
• -mannotate-tablejump, -mno-annotate-tablejump — create an annotation section ".discard.tablejump_annotate" to correlate the "jirl" instruction and the jump table when a jump table is used to optimize the "switch" statement. Some external tools, for example objtool of the Linux kernel building system, need the annotation to analysis the control flow. The default is -mno-annotate-tablejump.
• --param loongarch-vect-unroll-limit=n — the vectorizer will use available tuning information to determine whether it would be beneficial to unroll the main vectorized loop and by how much. This parameter set's the upper bound of how much the vectorizer will unroll the main loop. The default value is six. 🔥

---

🌸 M32C Options 🌸

• -mcpu=name — select the CPU for which code is generated. name may be one of r8c for the R8C/Tiny series, m16c for the M16C (up to /60) series, m32cm for the M16C/80 series, or m32c for the M32C/80 series. Pick your fighter 💅
• -msim — says the program will be run on the simulator. This causes an alternate runtime library to be linked in which supports, for example, file I/O. And bestie listen: you must not use this option when generating programs that will run on real hardware — you have to provide your own runtime library for whatever I/O functions are needed. 🙈
• -memregs=number — specifies the number of memory-based pseudo-registers GCC uses during code generation. These pseudo-registers are used like real registers, so there's a tradeoff between GCC's ability to fit the code into available registers and the performance penalty of using memory instead of registers. Note that all modules in a program must be compiled with the same value for this option. Because of that, you must not use this option with GCC's default runtime libraries. 😭

---

🦋 M32R/D Options 🦋

These -m options are defined for Renesas M32R/D architectures:

• -m32r2 — generate code for the M32R/2.
• -m32rx — generate code for the M32R/X.
• -m32r — generate code for the M32R. This is the default. 👑
• -mmodel=small — assume all objects live in the lower 16MB of memory (so that their addresses can be loaded with the "ld24" instruction), and assume all subroutines are reachable with the "bl" instruction. This is the default.

The addressability of a particular object can be set with the "model" attribute.

• -mmodel=medium — assume objects may be anywhere in the 32-bit address space (the compiler generates "seth/add3" instructions to load their addresses), and assume all subroutines are reachable with the "bl" instruction.
• -mmodel=large — assume objects may be anywhere in the 32-bit address space (the compiler generates "seth/add3" instructions to load their addresses), and assume subroutines may not be reachable with the "bl" instruction — the compiler generates the much slower "seth/add3/jl" instruction sequence. 🐌
• -msdata=none — disable use of the small data area. Variables are put into one of ".data", ".bss", or ".rodata" (unless the "section" attribute has been specified). This is the default.

The small data area consists of sections ".sdata" and ".sbss". Objects may be explicitly put in the small data area with the "section" attribute using one of these sections.

• -msdata=sdata — put small global and static data in the small data area, but do not generate special code to reference them.
• -msdata=use — put small global and static data in the small data area, and generate special instructions to reference them. ✨
• -G num — put global and static objects less than or equal to num bytes into the small data or BSS sections instead of the normal data or BSS sections. The default value of num is 8. The -msdata option must be set to one of sdata or use for this option to have any effect.

All modules should be compiled with the same -G num value. Compiling with different values of num may or may not work; if it doesn't the linker gives an error message---incorrect code is not generated. 🙏

• -mdebug — makes the M32R-specific code in the compiler display some statistics that might help in debugging programs.
• -malign-loops — align all loops to a 32-byte boundary.
• -mno-align-loops — do not enforce a 32-byte alignment for loops. This is the default.
• -missue-rate=number — issue number instructions per cycle. number can only be 1 or 2. ⭐
• -mbranch-cost=number — number can only be 1 or 2. If it is 1 then branches are preferred over conditional code, if it is 2, then the opposite applies.
• -mflush-trap=number — tells GCC which trap number to use for flushing the cache ✨. Default is 12, and the numbers you're allowed to use are <rainbow>0 through 15, inclusive</rainbow> 💅.
• -mno-flush-trap — says "bestie, the cache literally cannot be flushed with a trap" 🙈.
• -mflush-func=name — gives the name of the OS function to call when flushing the cache 🎀. Default is _flush_cache, but heads up — a function call only gets used if there's no trap available 👑.
• -mno-flush-func — declares that there is no OS function for flushing the cache, periodt 💀.

---

🌸 M680x0 Options 🌸

These are the -m options that exist for M680x0 and ColdFire processors 💖. The default settings depend on which architecture was picked when the compiler was configured, and the defaults for the most common choices are listed right below ⬇️.

---

• -march=arch — generates code for one specific M680x0 or ColdFire instruction set architecture 🦋.

The permissible values of arch for M680x0 architectures are: 68000, 68010, 68020, 68030, 68040, 68060 and cpu32.

ColdFire architectures are selected according to Freescale's ISA classification, and the permissible values are: isaa, isaaplus, isab and isac.

GCC defines the macro "__mcfarch__" whenever it's generating code for a ColdFire target ✨. The arch inside that macro is one of the -march arguments listed above.

When you use them together, -march and -mtune select code that runs on a whole family of similar processors but is optimized for one particular microarchitecture 💅.

• -mcpu=cpu — generates code for one specific M680x0 or ColdFire processor 🎀.

The M680x0 cpus are: 68000, 68010, 68020, 68030, 68040, 68060, 68302, 68332 and cpu32.

The ColdFire cpus are given by the table below, which also sorts the CPUs into families 👑:

Family : -mcpu arguments
51 : 51 51ac 51ag 51cn 51em 51je 51jf 51jg 51jm 51mm 51qe 51qm
5206 : 5202 5204 5206
5206e : 5206e
5208 : 5207 5208
5211a : 5210a 5211a
5213 : 5211 5212 5213
5216 : 5214 5216
52235 : 52230 52231 52232 52233 52234 52235
5225 : 5224 5225
52259 : 52252 52254 52255 52256 52258 52259
5235 : 5232 5233 5234 5235 523x
5249 : 5249
5250 : 5250
5271 : 5270 5271
5272 : 5272
5275 : 5274 5275
5282 : 5280 5281 5282 528x
53017 : 53011 53012 53013 53014 53015 53016 53017
5307 : 5307
5329 : 5327 5328 5329 532x
5373 : 5372 5373 537x
5407 : 5407
5475 : 5470 5471 5472 5473 5474 5475 547x 5480 5481 5482 5483 5484
5485

-mcpu=cpu overrides -march=arch if arch is compatible with cpu 🫶. Any other combos of -mcpu and -march get rejected 💀.

GCC defines the macro "__mcf_cpu_cpu" when a ColdFire target cpu is selected ✨. It also defines "__mcf_family_family", where the value of family comes from the table above.

• -mtune=tune — tunes the code for a particular microarchitecture within the constraints set by -march and -mcpu 🌸.

The M680x0 microarchitectures are: 68000, 68010, 68020, 68030, 68040, 68060 and cpu32.

The ColdFire microarchitectures are: cfv1, cfv2, cfv3, cfv4 and cfv4e.

You can also use -mtune=68020-40 for code that needs to run relatively well on 68020, 68030 and 68040 targets 🦋. -mtune=68020-60 is similar but also includes 68060 targets. These two options select the same tuning decisions as -m68020-40 and -m68020-60 respectively.

GCC defines the macros "__mcarch" and "__mcarch__" when tuning for 680x0 architecture arch 💖. It also defines "mcarch" unless either -ansi or a non-GNU -std option is used. If GCC is tuning for a range of architectures — as selected by -mtune=68020-40 or -mtune=68020-60 — it defines the macros for every architecture in the range 👑.

GCC also defines the macro "__muarch__" when tuning for ColdFire microarchitecture uarch, where uarch is one of the arguments given above ✨.

• -m68000
• -mc68000

Generate output for a 68000 🎀. This is the default when the compiler is configured for 68000-based systems. It's equivalent to -march=68000.

Use this option for microcontrollers with a 68000 or EC000 core, including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356 💅.

• -m68010 — generate output for a 68010 🌸. This is the default when the compiler is configured for 68010-based systems. It's equivalent to -march=68010.
• -m68020
• -mc68020

Generate output for a 68020 ✨. This is the default when the compiler is configured for 68020-based systems. It's equivalent to -march=68020.

• -m68030 — generate output for a 68030 🦋. This is the default when the compiler is configured for 68030-based systems. It's equivalent to -march=68030.
• -m68040 — generate output for a 68040 💖. This is the default when the compiler is configured for 68040-based systems. It's equivalent to -march=68040.

This option inhibits the use of 68881/68882 instructions that have to be emulated by software on the 68040 🙈. Use it if your 68040 doesn't have code to emulate those instructions.

• -m68060 — generate output for a 68060 👑. This is the default when the compiler is configured for 68060-based systems. It's equivalent to -march=68060.

This option inhibits the use of 68020 and 68881/68882 instructions that have to be emulated by software on the 68060 🙈. Use it if your 68060 doesn't have code to emulate those instructions.

• -mcpu32 — generate output for a CPU32 🎀. This is the default when the compiler is configured for CPU32-based systems. It's equivalent to -march=cpu32.

Use this option for microcontrollers with a CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334, 68336, 68340, 68341, 68349 and 68360 🌸.

• -m5200 — generate output for a 520X ColdFire CPU ✨. This is the default when the compiler is configured for 520X-based systems. It's equivalent to -mcpu=5206, and it's now deprecated in favor of that option 💀.

Use this option for microcontrollers with a 5200 core, including the MCF5202, MCF5203, MCF5204 and MCF5206.

• -m5206e — generate output for a 5206e ColdFire CPU 🦋. This option is now deprecated in favor of the equivalent -mcpu=5206e ⚠️.
• -m528x — generate output for a member of the ColdFire 528X family 💖. This option is now deprecated in favor of the equivalent -mcpu=528x ⚠️.
• -m5307 — generate output for a ColdFire 5307 CPU 🎀. This option is now deprecated in favor of the equivalent -mcpu=5307 ⚠️.
• -m5407 — generate output for a ColdFire 5407 CPU 🌸. This option is now deprecated in favor of the equivalent -mcpu=5407 ⚠️.
• -mcfv4e — generate output for a ColdFire V4e family CPU (e.g. 547x/548x) 👑. This includes use of hardware floating-point instructions. The option is equivalent to -mcpu=547x, and it's now deprecated in favor of that option ⚠️.
• -m68020-40 — generate output for a 68040, without using any of the new instructions 🦋. This results in code that can run relatively efficiently on either a 68020/68881 or a 68030 or a 68040. The generated code does use the 68881 instructions that are emulated on the 68040 ✨.

The option is equivalent to -march=68020 -mtune=68020-40 💅.

• -m68020-60 — generate output for a 68060, without using any of the new instructions 🌸. This results in code that can run relatively efficiently on either a 68020/68881 or a 68030 or a 68040. The generated code does use the 68881 instructions that are emulated on the 68060 ✨.

The option is equivalent to -march=68020 -mtune=68020-60 💅.

• -mhard-float
• -m68881

Generate floating-point instructions 💖. This is the default for 68020 and above, and for ColdFire devices that have an FPU. It defines the macro "__HAVE_68881__" on M680x0 targets and "__mcffpu__" on ColdFire targets ✨.

• -msoft-float — do not generate floating-point instructions; use library calls instead 🙈. This is the default for 68000, 68010, and 68832 targets. It's also the default for ColdFire devices that have no FPU 💀.
• -mdiv
• -mno-div

Generate (or don't generate) ColdFire hardware divide and remainder instructions 🎀. If -march is used without -mcpu, the default is "on" for ColdFire architectures and "off" for M680x0 architectures. Otherwise, the default comes from the target CPU — either the default CPU, or the one you specified with -mcpu 👑. For example, the default is "off" for -mcpu=5206 and "on" for -mcpu=5206e.

GCC defines the macro "__mcfhwdiv__" when this option is enabled ✨.

• -mshort — treats type "int" as being 16 bits wide, just like "short int" 🦋. On top of that, parameters passed on the stack are also aligned to a 16-bit boundary, even on targets whose API mandates promotion to 32-bit 👀.
• -mno-short — does not treat type "int" as being 16 bits wide 💅. This is the default.
• -mnobitfield
• -mno-bitfield

Do not use the bit-field instructions 🙈. The -m68000, -mcpu32 and -m5200 options imply -mnobitfield.

• -mbitfield — do use the bit-field instructions ✨. The -m68020 option implies -mbitfield. This is the default if you use a configuration designed for a 68020 👑.
• -mrtd — uses a totally different function-calling convention, where functions that take a fixed number of arguments return with the "rtd" instruction, which pops their arguments while returning 🎀. This saves one instruction in the caller, since there's no need to pop the arguments there 💅.

Heads up bestie ⚠️ this calling convention is incompatible with the one normally used on Unix, so you cannot use it if you need to call libraries compiled with the Unix compiler 😭.

Also, you must provide function prototypes for all functions that take variable numbers of arguments (including "printf") — otherwise incorrect code is generated for calls to those functions 💀.

In addition, seriously incorrect code results if you call a function with too many arguments 🙈. (Normally, extra arguments are harmlessly ignored.)

The "rtd" instruction is supported by the 68010, 68020, 68030, 68040, 68060 and CPU32 processors, but not by the 68000 or 5200 ✨.

The default is -mno-rtd.

• -malign-int, -mno-align-int — bestie, this controls whether GCC lines up your int, long, long long, float, double and long double variables on a 32-bit boundary (-malign-int) or a 16-bit boundary (-mno-align-int). Going 32-bit gets you code that runs somewhat faster on processors with 32-bit busses, but it costs you more memory. trade-offs, we love them 💅

Warning: if you use the -malign-int switch, GCC aligns structures containing the above types differently than most published application binary interface specifications for the m68k. so like… heads up fr.

Use the pc-relative addressing mode of the 68000 directly, instead of using a global offset table. At present, this option implies -fpic, allowing at most a 16-bit offset for pc-relative addressing. -fPIC is not presently supported with -mpcrel, though this could be supported for 68020 and higher processors. 👑

• -mno-strict-align, -mstrict-align — do not (do) assume that unaligned memory references are handled by the system.
• -msep-data — generates code that lets the data segment live in a different area of memory from the text segment. this enables execute-in-place in an environment with no virtual memory management. this option implies -fPIC. ✨
• -mno-sep-data — generates code that assumes the data segment follows the text segment. this is the default.
• -mid-shared-library — generates code that supports shared libraries via the library ID method. this allows for execute-in-place and shared libraries in an environment without virtual memory management. implies -fPIC 🦋
• -mno-id-shared-library — generates code that doesn't assume ID-based shared libraries are being used. this is the default.
• -mshared-library-id=n — says which ID-based shared library you're compiling, by its identification number. 0 gets you more compact code; any other value forces that number to be allocated to the current library, but it's no more space- or time-efficient than just omitting this option. 🤷‍♀️
• -mxgot, -mno-xgot — when you're generating position-independent code for ColdFire, this makes code that still works if the GOT has more than 8192 entries. said code is bigger and slower than what you'd get without this option. on M680x0 processors you don't need it — -fPIC suffices.

GCC normally uses a single instruction to load values from the GOT. it's pretty efficient, but it only works if the GOT is smaller than about 64k. anything bigger and your linker starts yelling at you like:

        relocation truncated to fit: R_68K_GOT16O foobar

if that happens, bestie, recompile your code with -mxgot. it should then work with very large GOTs. the catch? code generated with -mxgot is less efficient, since fetching the value of a global symbol takes 4 instructions. 💀

note that some linkers, including newer versions of the GNU linker, can create multiple GOTs and sort GOT entries. if you have one of those, you should only need -mxgot when compiling a single object file that accesses more than 8192 GOT entries. very few do. 🙈

these options have no effect unless GCC is generating position-independent code.

• -mlong-jump-table-offsets — use 32-bit offsets in switch tables. the default is to use 16-bit offsets.

---

🎀 MCore Options 🎀

these are the -m options defined for the Motorola M*Core processors.

• -mhardlit, -mno-hardlit — inline constants into the code stream if it can be done in two instructions or less.
• -mdiv, -mno-div — use the divide instruction. (enabled by default ✅)
• -mrelax-immediate, -mno-relax-immediate — allow arbitrary-sized immediates in bit operations.
• -mwide-bitfields, -mno-wide-bitfields — always treat bit-fields as int-sized.
• -m4byte-functions, -mno-4byte-functions — force all functions to be aligned to a 4-byte boundary.
• -mcallgraph-data, -mno-callgraph-data — emit callgraph information.
• -mslow-bytes, -mno-slow-bytes — prefer word access when reading byte quantities.
• -mlittle-endian, -mbig-endian — generate code for a little-endian target.
• -m210, -m340 — generate code for the 210 processor.
• -mno-lsim — assume that runtime support has been provided, so omit the simulator library (libsim.a) from the linker command line.
• -mstack-increment=size — set the maximum amount for a single stack increment operation. large values can speed up programs that contain functions needing a big ol' amount of stack space, but they can also trigger a segmentation fault if the stack is extended too much. the default value is 0x1000. 😭

---

🌈 MicroBlaze Options 🌈

• -msoft-float — use software emulation for floating point (default).
• -mhard-float — use hardware floating-point instructions.
• -mmemcpy — do not optimize block moves, use memcpy.
• -mno-clearbss — this option is deprecated. use -fno-zero-initialized-in-bss instead. ⚠️
• -mcpu=cpu-type — use features of, and schedule code for, the given CPU. supported values are in the format vX.YY.Z, where X is a major version, YY is the minor version, and Z is compatibility code. example values are v3.00.a, v4.00.b, v5.00.a, v5.00.b, v6.00.a.
• -mxl-soft-mul — use software multiply emulation (default).
• -mxl-soft-div — use software emulation for divides (default).
• -mxl-barrel-shift — use the hardware barrel shifter.
• -mxl-pattern-compare — use pattern compare instructions.
• -msmall-divides — use table lookup optimization for small signed integer divisions.
• -mxl-stack-check — this option is deprecated. use -fstack-check instead. ⚠️
• -mxl-gp-opt — use GP-relative .sdata/.sbss sections.
• -mxl-multiply-high — use multiply high instructions for high part of 32x32 multiply.
• -mxl-float-convert — use hardware floating-point conversion instructions.
• -mxl-float-sqrt — use hardware floating-point square root instruction.
• -mbig-endian — generate code for a big-endian target.
• -mlittle-endian — generate code for a little-endian target.
• -mxl-reorder — use reorder instructions (swap and byte reversed load/store).
• -mxl-mode-app-model — select application model app-model. valid models are:
  • executable — normal executable (default), uses startup code crt0.o.
  • xmdstub — for use with Xilinx Microprocessor Debugger (XMD) based software intrusive debug agent called xmdstub. this uses startup file crt1.o and sets the start address of the program to 0x800.
  • bootstrap — for applications that are loaded using a bootloader. this model uses startup file crt2.o which does not contain a processor reset vector handler. this is suitable for transferring control on a processor reset to the bootloader rather than the application.
  • novectors — for applications that do not require any of the MicroBlaze vectors. this option may be useful for applications running within a monitoring application. this model uses crt3.o as a startup file.

option -xl-mode-app-model is a deprecated alias for -mxl-mode-app-model. 👀

• -mpic-data-is-text-relative — assume that the displacement between the text and data segments is fixed at static link time. this allows data to be referenced by offset from start of text address instead of GOT, since PC-relative addressing is not supported.

---

💜 MIPS Options 💜

• -EB — generate big-endian code.
• -EL — generate little-endian code. this is the default for mips*el-*-* configurations.
• -march=arch — generate code that runs on arch, which can be the name of a generic MIPS ISA, or the name of a particular processor. the ISA names are: mips1, mips2, mips3, mips4, mips32, mips32r2, mips32r3, mips32r5, mips32r6, mips64, mips64r2, mips64r3, mips64r5 and mips64r6. the processor names are: 4kc, 4km, 4kp, 4ksc, 4kec, 4kem, 4kep, 4ksd, 5kc, 5kf, 20kc, 24kc, 24kf2_1, 24kf1_1, 24kec, 24kef2_1, 24kef1_1, 34kc, 34kf2_1, 34kf1_1, 34kn, 74kc, 74kf2_1, 74kf1_1, 74kf3_2, 1004kc, 1004kf2_1, 1004kf1_1, i6400, i6500, interaptiv, loongson2e, loongson2f, loongson3a, gs464, gs464e, gs264e, m4k, m14k, m14kc, m14ke, m14kec, m5100, m5101, octeon, octeon+, octeon2, octeon3, orion, p5600, p6600, r2000, r3000, r3900, r4000, r4400, r4600, r4650, r4700, r5900, r6000, r8000, rm7000, rm9000, r10000, r12000, r14000, r16000, sb1, sr71000, vr4100, vr4111, vr4120, vr4130, vr4300, vr5000, vr5400, vr5500, xlr and xlp. the special value from-abi selects the most compatible architecture for the selected ABI (that is, mips1 for 32-bit ABIs and mips3 for 64-bit ABIs).

the native Linux/GNU toolchain also supports the value native, which selects the best architecture option for the host processor. -march=native has no effect if GCC does not recognize the processor. 🙈

in processor names, a final 000 can be abbreviated as k (for example, -march=r2k). prefixes are optional, and vr may be written r.

names of the form nf2_1 refer to processors with FPUs clocked at half the rate of the core, names of the form nf1_1 refer to processors with FPUs clocked at the same rate as the core, and names of the form nf3_2 refer to processors with FPUs clocked a ratio of 3:2 with respect to the core. for compatibility reasons, nf is accepted as a synonym for nf2_1 while nx and bfx are accepted as synonyms for nf1_1.

GCC defines two macros based on the value of this option. the first is _MIPS_ARCH, which gives the name of target architecture, as a string. the second has the form _MIPS_ARCH_foo, where foo is the capitalized value of _MIPS_ARCH. for example, -march=r2000 sets _MIPS_ARCH to r2000 and defines the macro _MIPS_ARCH_R2000.

note that the _MIPS_ARCH macro uses the processor names given above. in other words, it has the full prefix and does not abbreviate 000 as k. in the case of from-abi, the macro names the resolved architecture (either mips1 or mips3). it names the default architecture when no -march option is given. 💖

• -mtune=arch — optimize for arch. Like, this is the vibe-setter for how instructions get scheduled and how "expensive" arithmetic feels to the compiler. The list of arch values is the same one as -march, no cap. 🎀

If you don't use this option at all, GCC just tunes for whatever processor -march picked. But bestie, using -march and -mtune together is the serve: you get code that runs on a whole family of processors but is optimized for one specific member of that family. Iconic. 💅

-mtune defines the macros _MIPS_TUNE and _MIPS_TUNE_foo, and they work exactly the same way as the -march ones described above. 🦋

• -mips1 — equivalent to -march=mips1.
• -mips2 — equivalent to -march=mips2.
• -mips3 — equivalent to -march=mips3.
• -mips4 — equivalent to -march=mips4.
• -mips32 — equivalent to -march=mips32.
• -mips32r3 — equivalent to -march=mips32r3.
• -mips32r5 — equivalent to -march=mips32r5.
• -mips32r6 — equivalent to -march=mips32r6.
• -mips64 — equivalent to -march=mips64.
• -mips64r2 — equivalent to -march=mips64r2.
• -mips64r3 — equivalent to -march=mips64r3.
• -mips64r5 — equivalent to -march=mips64r5.
• -mips64r6 — equivalent to -march=mips64r6.
• -mips16, -mno-mips16 — generate (or don't generate) MIPS16 code. If GCC is targeting a MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE. ✨

You can also control MIPS16 code generation per-function, using the "mips16" and "nomips16" attributes. One function at a time, we love a granular queen. 👑

• -mmips16e2, -mno-mips16e2 — use (or don't use) the MIPS16e2 ASE. This one tweaks how -mips16 behaves so that it targets the MIPS16e2 ASE instead.
• -mflip-mips16 — generate MIPS16 code on alternating functions. ⚠️ This one exists purely for regression testing of mixed MIPS16/non-MIPS16 code generation, and it is not meant for normal use when you're compiling actual user code. It's giving lab experiment, not production, fr fr.
• -minterlink-compressed, -mno-interlink-compressed — require (or don't require) that code using the standard (uncompressed) MIPS ISA be link-compatible with MIPS16 and microMIPS code, and vice versa. 🔗

Like, for example: code using the standard ISA encoding cannot jump directly to MIPS16 or microMIPS code. It has to use a call or an indirect jump. So -minterlink-compressed disables direct jumps unless GCC knows that the target of the jump isn't compressed. Safety first, bestie. 🫶

• -minterlink-mips16, -mno-interlink-mips16 — these are aliases of -minterlink-compressed and -mno-interlink-compressed. They predate the microMIPS ASE and are retained for backwards compatibility. Vintage, but still valid. 💜
• -mabi=32, -mabi=o64, -mabi=n32, -mabi=64, -mabi=eabi — generate code for the given ABI. Pick your fighter. 🌈

Heads up: the EABI has a 32-bit variant and a 64-bit variant. GCC normally generates 64-bit code when you select a 64-bit architecture, but you can use -mgp32 to get 32-bit code instead.

For deets about the O64 ABI, see <https://gcc.gnu.org/projects/mipso64-abi.html>. 📎

GCC also supports a variant of the o32 ABI where floating-point registers are 64 bits wide instead of 32. You select that combo with -mabi=32 -mfp64. This ABI relies on the "mthc1" and "mfhc1" instructions, so it's only supported for MIPS32R2, MIPS32R3 and MIPS32R5 processors. 🚫 for everything else, sorry.

The register assignments for arguments and return values stay the same, but each scalar value gets passed in a single 64-bit register rather than a pair of 32-bit registers. For example, scalar floating-point values are returned in $f0 only — not a $f0/$f1 pair. The set of call-saved registers also stays the same, in that the even-numbered double-precision registers are saved. Same energy, different register width. ✨

Two extra variants of the o32 ABI exist to help you transition from 32-bit to 64-bit registers:

• FPXX (-mfpxx)
• FP64A (-mfp64 -mno-odd-spreg)

The FPXX extension mandates that all code must execute correctly whether it's run using 32-bit or 64-bit registers. That code can be interlinked with either FP32 or FP64 — but not both. Not both!! Pick one. 💀

The FP64A extension is similar to the FP64 extension but forbids the use of odd-numbered single-precision registers. You can use it together with the "FRE" mode of FPUs in MIPS32R5 processors, and it lets both FP32 and FP64A code interlink and run in the same process without changing FPU modes. ✨ teamwork makes the dream work.

• -mabicalls, -mno-abicalls — generate (or don't generate) code suitable for SVR4-style dynamic objects. -mabicalls is the default for SVR4-based systems.
• -mshared, -mno-shared — generate (or don't generate) code that's fully position-independent and can therefore be linked into shared libraries. This option only affects -mabicalls.

All -mabicalls code has traditionally been position-independent, regardless of options like -fPIC and -fpic. But as an extension, the GNU toolchain lets executables use absolute accesses for locally-binding symbols. It can also use shorter GP initialization sequences and generate direct calls to locally-defined functions. That mode is selected by -mno-shared.

-mno-shared depends on binutils 2.16 or higher, and it generates objects that can only be linked by the GNU linker. The option does not affect the ABI of the final executable though — it only affects the ABI of relocatable objects. Using -mno-shared generally makes executables both smaller and quicker. slay. 🔥

-mshared is the default.

• -mplt, -mno-plt — assume (or don't assume) that the static and dynamic linkers support PLTs and copy relocations. This option only affects -mno-shared -mabicalls. For the n64 ABI, this option has no effect without -msym32.

You can make -mplt the default by configuring GCC with --with-mips-plt. Otherwise, the default is -mno-plt.

• -mxgot, -mno-xgot — lift (or don't lift) the usual restrictions on the size of the global offset table.

GCC normally uses a single instruction to load values from the GOT. While that's relatively efficient, it only works if the GOT is smaller than about 64k. Anything bigger and the linker starts yelling, something like:

relocation truncated to fit: R_MIPS_GOT16 foobar

If that happens to you, recompile your code with -mxgot. 😭 This works with very large GOTs, though the code is also less efficient, since it takes three instructions to fetch the value of a global symbol. Trade-offs, bestie.

Note that some linkers can create multiple GOTs. If you've got one of those, you should only need -mxgot when a single object file accesses more than 64k's worth of GOT entries. Very few do. 🙈

These options have no effect unless GCC is generating position independent code.

• -mgp32 — assume that general-purpose registers are 32 bits wide.
• -mgp64 — assume that general-purpose registers are 64 bits wide.
• -mfp32 — assume that floating-point registers are 32 bits wide.
• -mfp64 — assume that floating-point registers are 64 bits wide.
• -mfpxx — do not assume the width of floating-point registers. mystery era. 🦋
• -mhard-float — use floating-point coprocessor instructions.
• -msoft-float — do not use floating-point coprocessor instructions. Implement floating-point calculations using library calls instead.
• -mno-float — equivalent to -msoft-float, but it additionally asserts that the program being compiled performs no floating-point operations whatsoever. This option is presently supported only by some bare-metal MIPS configurations, where it may select a special set of libraries that lack all floating-point support (including, for example, the floating-point "printf" formats). ⚠️ If code compiled with -mno-float accidentally contains floating-point operations, it's likely to suffer a link-time or run-time failure. Not cute, be careful.
• -msingle-float — assume that the floating-point coprocessor only supports single-precision operations.
• -mdouble-float — assume that the floating-point coprocessor supports double-precision operations. This is the default.
• -modd-spreg, -mno-odd-spreg — enable the use of odd-numbered single-precision floating-point registers for the o32 ABI. This is the default for processors that are known to support these registers. When using the o32 FPXX ABI, -mno-odd-spreg is set by default.
• -mabs=2008, -mabs=legacy — these control how the special not-a-number (NaN) IEEE 754 floating-point data is treated with the "abs.fmt" and "neg.fmt" machine instructions.

By default, or when -mabs=legacy is used, the legacy treatment is selected. In that case these instructions count as arithmetic and get avoided wherever correct operation is required and the input operand might be a NaN. A longer sequence of instructions that manipulate the sign bit of the floating-point datum manually gets used instead — unless the -ffinite-math-only option has also been specified. 💅

The -mabs=2008 option selects the IEEE 754-2008 treatment. In that case these instructions count as non-arithmetic and therefore operate correctly in all cases, including in particular where the input operand is a NaN. So these instructions are always used for the respective operations. periodt. ✨

• -mnan=2008, -mnan=legacy — these control the encoding of the special not-a-number (NaN) IEEE 754 floating-point data.

The -mnan=legacy option selects the legacy encoding. In this case quiet NaNs (qNaNs) are denoted by the first bit of their trailing significand field being 0, whereas signaling NaNs (sNaNs) are denoted by the first bit of their trailing significand field being 1. 🎀

The -mnan=2008 option selects the IEEE 754-2008 encoding. In this case qNaNs are denoted by the first bit of their trailing significand field being 1, whereas sNaNs are denoted by the first bit of their trailing significand field being 0. They literally just flipped it, the audacity. 💀

The default is -mnan=legacy unless GCC has been configured with --with-nan=2008.

• -mllsc, -mno-llsc — use (or don't use) ll, sc, and sync instructions to implement atomic memory built-in functions. When neither option is specified, GCC uses the instructions if the target architecture supports them. 🧚

so like, -mllsc</p> is your girl when the runtime environment can emulate the instructions, and -mno-llsc</p> comes in clutch when you're compiling for nonstandard ISAs. wanna make either one the default? configure GCC with --with-llsc and --without-llsc respectively and boom. --with-llsc is the default for some configurations — peep the installation documentation for the deets. 💅

• -mdsp, -mno-dsp — use (don't use) revision 1 of the MIPS DSP ASE. this one defines the preprocessor macro __mips_dsp, and it also defines __mips_dsp_rev to 1. ✨
• -mdspr2, -mno-dspr2 — use (don't use) revision 2 of the MIPS DSP ASE. defines the preprocessor macros __mips_dsp and __mips_dspr2, and sets __mips_dsp_rev to 2. double the macros, double the slay 💖
• -msmartmips, -mno-smartmips — use (don't use) the MIPS SmartMIPS ASE. it's literally giving smart. ✨
• -mpaired-single, -mno-paired-single — use (don't use) paired-single floating-point instructions. heads up bestie, this option requires hardware floating-point support to be enabled. no hardware FP, no paired-single, periodt. 🫶
• -mdmx, -mno-mdmx — use (don't use) MIPS Digital Media Extension instructions. this one can only be used when generating 64-bit code, and it also needs hardware floating-point support enabled. so many prerequisites omg 😭
• -mips3d, -mno-mips3d — use (don't use) the MIPS-3D ASE. lowkey cute detail: -mips3d implies -mpaired-single. 🎀
• -mmicromips, -mno-micromips — generate (don't generate) microMIPS code.

microMIPS code generation can also be controlled on a per-function basis by means of the "micromips" and "nomicromips" attributes. iykyk 💅

• -mmt, -mno-mt — use (don't use) MT Multithreading instructions. multithreading but make it iconic. 🔥
• -mmcu, -mno-mcu — use (don't use) the MIPS MCU ASE instructions.
• -meva, -mno-eva — use (don't use) the MIPS Enhanced Virtual Addressing instructions.
• -mvirt, -mno-virt — use (don't use) the MIPS Virtualization (VZ) instructions.
• -mxpa, -mno-xpa — use (don't use) the MIPS eXtended Physical Address (XPA) instructions.
• -mcrc, -mno-crc — use (don't use) the MIPS Cyclic Redundancy Check (CRC) instructions.
• -mginv, -mno-ginv — use (don't use) the MIPS Global INValidate (GINV) instructions.
• -mloongson-mmi, -mno-loongson-mmi — use (don't use) the MIPS Loongson MultiMedia extensions Instructions (MMI).
• -mloongson-ext, -mno-loongson-ext — use (don't use) the MIPS Loongson EXTensions (EXT) instructions.
• -mloongson-ext2, -mno-loongson-ext2 — use (don't use) the MIPS Loongson EXTensions r2 (EXT2) instructions.
• -mlong64 — force long types to be 64 bits wide. wanna know about the default and how the pointer size gets determined? see -mlong32 for the whole explanation. 👑
• -mlong32 — force long, int, and pointer types to be 32 bits wide.

kay so the default size of ints, longs and pointers depends on the ABI. all the supported ABIs use 32-bit ints. the n64 ABI uses 64-bit longs, and so does the 64-bit EABI; the others use 32-bit longs. pointers are the same size as longs, or the same size as integer registers — whichever is smaller. okurrr 💅

• -msym32, -mno-sym32 — assume (don't assume) that all symbols have 32-bit values, regardless of the selected ABI. this one's a vibe in combination with -mabi=64 and -mno-abicalls, because it lets GCC generate shorter and faster references to symbolic addresses. 🦋
• -G num — put definitions of externally-visible data in a small data section if that data is no bigger than num bytes. then GCC can generate more efficient accesses to the data; peep -mgpopt for the details.

the default -G option depends on the configuration, bestie. 🌸

• -mlocal-sdata, -mno-local-sdata — extend (don't extend) the -G behavior to local data too, like static variables in C. -mlocal-sdata is the default for all configurations.

if the linker starts complaining that an application is using too much small data, you might wanna try rebuilding the less performance-critical parts with -mno-local-sdata. you might also wanna build large libraries with -mno-local-sdata, so the libraries leave more room for the main program. thoughtful queen behavior 💜

• -mextern-sdata, -mno-extern-sdata — assume (don't assume) that externally-defined data is in a small data section if the size of that data is within the -G limit. -mextern-sdata is the default for all configurations.

ok so here's the tea: if you compile a module Mod with -mextern-sdata -G num -mgpopt, and Mod references a variable Var that's no bigger than num bytes, you must make sure that Var is placed in a small data section. if Var is defined by another module, you must either compile that module with a high-enough -G setting or attach a "section" attribute to Var's definition. and if Var is common, you must link the application with a high-enough -G setting. no skipping steps fr fr 😤

the easiest way to satisfy all these restrictions is to compile and link every module with the same -G option. but maybe you wanna build a library that supports several different small data limits? iconique of you. you can do this by compiling the library with the highest supported -G setting and additionally using -mno-extern-sdata to stop the library from making assumptions about externally-defined data. 💅

• -mgpopt, -mno-gpopt — use (don't use) GP-relative accesses for symbols that are known to be in a small data section; see -G, -mlocal-sdata and -mextern-sdata. -mgpopt is the default for all configurations.

-mno-gpopt is useful for cases where the $gp register might not hold the value of "_gp". like, say the code is part of a library that might get used in a boot monitor — programs that call boot monitor routines pass an unknown value in $gp. (in such situations, the boot monitor itself is usually compiled with -G0.) 💀

also important: -mno-gpopt implies -mno-local-sdata and -mno-extern-sdata. the implications cascade, bestie 🫶

• -membedded-data, -mno-embedded-data — allocate variables to the read-only data section first if possible, then next in the small data section if possible, otherwise in data. this gives slightly slower code than the default, but it reduces the amount of RAM required when executing, so it may be preferred for some embedded systems. trade-off queen 🌈
• -muninit-const-in-rodata, -mno-uninit-const-in-rodata — put uninitialized const variables in the read-only data section. heads up: this option is only meaningful in conjunction with -membedded-data. ⭐
• -mcode-readable=setting — specify whether GCC may generate code that reads from executable sections. three possible settings, bestie:
  • -mcode-readable=yes — instructions may freely access executable sections. this is the default setting. ✨
  • -mcode-readable=pcrel — MIPS16 PC-relative load instructions can access executable sections, but other instructions must not do so. this option is useful on 4KSc and 4KSd processors when the code TLBs have the Read Inhibit bit set. it's also useful on processors that can be configured to have a dual instruction/data SRAM interface and that, like the M4K, automatically redirect PC-relative loads to the instruction RAM. very specific, very slay 🎀
  • -mcode-readable=no — instructions must not access executable sections. this one can be useful on targets configured to have a dual instruction/data SRAM interface but that (unlike the M4K) don't automatically redirect PC-relative loads to the instruction RAM. 😌
• -msplit-addresses, -mno-split-addresses — enable (disable) use of the %hi() and %lo() assembler relocation operators. ngl this option has been superseded by -mexplicit-relocs but is retained for backwards compatibility. legacy icon 💅
• -mexplicit-relocs=none, -mexplicit-relocs=base, -mexplicit-relocs=pcrel, -mexplicit-relocs, -mno-explicit-relocs — these options control whether explicit relocs (like %gp_rel) are used. the default value depends on the version of GAS when GCC itself was built.

the "base" explicit-relocs support was introduced into GAS in 2001. the "pcrel" explicit-relocs support landed in GAS in 2014, and it supports %pcrel_hi and %pcrel_lo. history lesson but make it cute 🦋

• -mcheck-zero-division, -mno-check-zero-division — trap (don't trap) on integer division by zero. the default is -mcheck-zero-division. love that for us 👑
• -mdivide-traps, -mdivide-breaks — MIPS systems check for division by zero by generating either a conditional trap or a break instruction. using traps results in smaller code, but it's only supported on MIPS II and later. also, some versions of the Linux kernel have a bug that prevents trap from generating the proper signal ("SIGFPE"). 😭 use -mdivide-traps to allow conditional traps on architectures that support them, and -mdivide-breaks to force the use of breaks.

the default is usually -mdivide-traps, but this can be overridden at configure time using --with-divide=breaks. and divide-by-zero checks can be completely disabled using -mno-check-zero-division. three ways to handle it, pick your fighter 💀

• -mload-store-pairs, -mno-load-store-pairs — enable (disable) an optimization that pairs consecutive load or store instructions to enable load/store bonding. this option is enabled by default but only takes effect when the selected architecture is known to support bonding. conditional queen ✨
• -mstrict-align, -mno-strict-align, -munaligned-access, -mno-unaligned-access — disable (enable) direct unaligned access for MIPS Release 6. MIPSr6 requires load/store unaligned-access support, by hardware or trap&emulate. so -mstrict-align may be needed by the kernel. the options -munaligned-access and -mno-unaligned-access are obsoleted, and exist only for backward-compatibility. old news, still valid, moving on 💅
• -mmemcpy, -mno-memcpy — force (don't force) the use of memcpy for non-trivial block moves. the default is -mno-memcpy, which lets GCC inline most constant-sized copies. inline it bestie 🔥
• -mlong-calls, -mno-long-calls — disable (don't disable) use of the jal instruction. calling functions using jal is more efficient but it requires the caller and callee to be in the same 256 megabyte segment. tiny little requirement, no big deal 🙈

this option has no effect on abicalls code. the default is -mno-long-calls. 📌

• -mmad, -mno-mad — enable (disable) use of the "mad", "madu" and "mul" instructions, as provided by the R4650 ISA. three instructions, one flag, we love efficiency 🎀

---

• -mimadd, -mno-imadd — turns the "madd" and "msub" integer instructions on (or off). Default is -mimadd on any arch that supports "madd" and "msub" 💖 ... except the 74k, because there it was found to generate slower code. Not the vibe, bestie.
• -mfused-madd, -mno-fused-madd — lets GCC use (or not use) the floating-point multiply-accumulate instructions, whenever they're actually available. The default is -mfused-madd ✨.

⚠️ Heads up on the R8000 CPU: when multiply-accumulate instructions get used, that intermediate product is calculated to infinite precision and is not subject to the FCSR Flush to Zero bit. That can be genuinely undesirable in some circumstances 😬. On other processors though, the result is numerically identical to the equivalent computation done with separate multiply, add, subtract and negate instructions.

• -nocpp — tells the MIPS assembler to not run its preprocessor over your hand-written assembler files (the ones with a .s suffix) when it assembles them. No preprocessing, no problem 🫶.
• -mfix-24k, -mno-fix-24k — works around the 24K E48 errata (that's lost data on stores during refill 💀). Important detail: these workarounds are implemented by the assembler, not by GCC itself.
• -mfix-r4000, -mno-fix-r4000 — works around certain R4000 CPU errata:
  • A double-word or a variable shift may give an incorrect result if executed immediately after starting an integer division.
  • A double-word or a variable shift may give an incorrect result if executed while an integer multiplication is in progress.
  • An integer division may give an incorrect result if started in a delay slot of a taken branch or a jump.
• -mfix-r4400, -mno-fix-r4400 — works around certain R4400 CPU errata:
  • A double-word or a variable shift may give an incorrect result if executed immediately after starting an integer division.
• -mfix-r10000, -mno-fix-r10000 — works around certain R10000 errata:
  • "ll"/"sc" sequences may not behave atomically on revisions prior to 3.0. They may deadlock on revisions 2.6 and earlier. 😭

You can only use this option if the target architecture supports branch-likely instructions. -mfix-r10000 is the default when -march=r10000 is used; -mno-fix-r10000 is the default otherwise.

• -mfix-r5900, -mno-fix-r5900 — do not try to schedule the preceding instruction into the delay slot of a branch instruction placed at the end of a short loop of six instructions or fewer, and always schedule a "nop" instruction there instead. Why? Because of the short loop bug — under certain conditions it causes loops to execute only once or twice, thanks to a hardware bug in the R5900 chip 💀. This workaround is implemented by the assembler rather than by GCC.
• -mfix-rm7000, -mno-fix-rm7000 — works around the RM7000 "dmult"/"dmultu" errata. And yep, you guessed it, these workarounds are implemented by the assembler rather than by GCC.
• -mfix-vr4120, -mno-fix-vr4120 — works around certain VR4120 errata:
  • "dmultu" does not always produce the correct result.
  • "div" and "ddiv" do not always produce the correct result if one of the operands is negative.

The workarounds for the division errata rely on special functions in libgcc.a. At present, those functions are only provided by the "mips64vr*-elf" configurations 👑

Other VR4120 errata require a NOP to be inserted between certain pairs of instructions. Those errata are handled by the assembler, not by GCC itself.

• -mfix-vr4130 — works around the VR4130 "mflo"/"mfhi" errata. These workarounds are implemented by the assembler rather than by GCC, although GCC will avoid using "mflo" and "mfhi" if the VR4130 "macc", "macchi", "dmacc" and "dmacchi" instructions are available instead.
• -mfix-sb1, -mno-fix-sb1 — works around certain SB-1 CPU core errata. (This flag currently works around the SB-1 revision 2 "F1" and "F2" floating-point errata.) 🦋
• -mr10k-cache-barrier=setting — tell GCC whether it should insert cache barriers to avoid the side effects of speculation on R10K processors.

In common with many processors, the R10K tries to predict the outcome of a conditional branch and speculatively executes instructions from the taken branch. It later aborts those instructions if the predicted outcome is wrong. But on the R10K, even aborted instructions can have side effects 😭.

This problem only affects kernel stores and, depending on the system, kernel loads. Example: a speculatively-executed store may load the target memory into cache and mark the cache line as dirty, even if the store itself is later aborted. If a DMA operation writes to the same area of memory before that "dirty" line is flushed, the cached data overwrites the DMA-ed data 💀. See the R10K processor manual for a full description, including other potential problems.

One workaround is to insert cache barrier instructions before every memory access that might be speculatively executed and that might have side effects even if aborted. -mr10k-cache-barrier=setting controls GCC's implementation of this workaround. It assumes that aborted accesses to any byte in the following regions does not have side effects:

1. the memory occupied by the current function's stack frame;
2. the memory occupied by an incoming stack argument;
3. the memory occupied by an object with a link-time-constant address.

It is the kernel's responsibility to ensure that speculative accesses to these regions are indeed safe.

If the input program contains a function declaration such as:

void foo (void);

then the implementation of "foo" must allow "j foo" and "jal foo" to be executed speculatively. GCC honors this restriction for functions it compiles itself. It expects non-GCC functions (such as hand-written assembly code) to do the same.

The option has three forms:

• -mr10k-cache-barrier=load-store — insert a cache barrier before a load or store that might be speculatively executed and that might have side effects even if aborted.
• -mr10k-cache-barrier=store — insert a cache barrier before a store that might be speculatively executed and that might have side effects even if aborted.
• -mr10k-cache-barrier=none — disable the insertion of cache barriers. This is the default setting.

• -mflush-func=func, -mno-flush-func — specifies the function to call to flush the I and D caches, or says to not call any such function at all. If it does get called, the function must take the same arguments as the common "_flush_func", that is: the address of the memory range for which the cache is being flushed, the size of the memory range, and the number 3 (to flush both caches). The default depends on the target GCC was configured for, but commonly it's either "_flush_func" or "__cpu_flush" 🎀.
• -mbranch-cost=num — set the cost of branches to roughly num "simple" instructions. This cost is only a heuristic and is not guaranteed to produce consistent results across releases. A zero cost redundantly selects the default, which is based on the -mtune setting.
• -mbranch-likely, -mno-branch-likely — enable or disable use of Branch Likely instructions, regardless of the default for the selected architecture. By default, Branch Likely instructions may be generated if they're supported by the selected architecture. One exception: for the MIPS32 and MIPS64 architectures and the processors that implement those architectures, Branch Likely instructions are not generated by default, because the MIPS32 and MIPS64 architectures specifically deprecate their use 🚫.
• -mcompact-branches=never, -mcompact-branches=optimal, -mcompact-branches=always — these control which form of branches will be generated. The default is -mcompact-branches=optimal ✨.
  • -mcompact-branches=never ensures that compact branch instructions will never be generated.
  • -mcompact-branches=always ensures that a compact branch instruction will be generated if available for MIPS Release 6 onwards. If a compact branch instruction is not available (or pre-R6), a delay slot form of the branch will be used instead.
  • If it's used for MIPS16/microMIPS targets, it will just be ignored for now 👀. The behavior for MIPS16/microMIPS may change in future, since they do have some compact branch instructions.
  • -mcompact-branches=optimal will cause a delay slot branch to be used if one is available in the current ISA and the delay slot is successfully filled. If the delay slot is not filled, a compact branch will be chosen if one is available.
• -mfp-exceptions, -mno-fp-exceptions — specifies whether FP exceptions are enabled. This affects how FP instructions are scheduled for some processors. The default is that FP exceptions are enabled.

For instance, on the SB-1, if FP exceptions are disabled and we're emitting 64-bit code, then we can use both FP pipes 🔥. Otherwise, we can only use one FP pipe.

• -mvr4130-align, -mno-vr4130-align — the VR4130 pipeline is two-way superscalar, but it can only issue two instructions together if the first one is 8-byte aligned. When this option is enabled, GCC aligns pairs of instructions that it thinks should execute in parallel.

This option only has an effect when optimizing for the VR4130. It normally makes code faster, but at the expense of making it bigger. It's enabled by default at optimization level -O3.

• -msynci, -mno-synci — enable (or disable) generation of "synci" instructions on architectures that support them. The "synci" instructions (if enabled) are generated when "__builtin___clear_cache" is compiled.

This option defaults to -mno-synci, but the default can be overridden by configuring GCC with --with-synci. 🌈

⚠️ When compiling code for single processor systems, it's generally safe to use "synci". However, on many multi-core (SMP) systems, it does not invalidate the instruction caches on all cores and may lead to undefined behavior 💀. Be careful, bestie.

• -mrelax-pic-calls, -mno-relax-pic-calls — try to turn PIC calls that are normally dispatched via register $25 into direct calls. This is only possible if the linker can resolve the destination at link time and if the destination is within range for a direct call 👑.

Okay so real talk, bestie: -mrelax-pic-calls is the default if GCC was configured to use an assembler and a linker that support the .reloc assembly directive and -mexplicit-relocs is in effect. And with -mno-explicit-relocs, this optimization can be performed by the assembler and the linker alone without help from the compiler. Slay. 💅

• -mmcount-ra-address, -mno-mcount-ra-address — Emit / do not emit code that allows _mcount to modify the calling function's return address. When enabled, this option extends the usual _mcount interface with a new ra-address parameter, which has type intptr_t * and is passed in register $12. _mcount can then modify the return address by doing both of the following:
  • Returning the new address in register $31.
  • Storing the new address in *ra-address, if ra-address is nonnull.

The default is -mno-mcount-ra-address.

• -mframe-header-opt, -mno-frame-header-opt — Enable / disable frame header optimization in the o32 ABI. When using the o32 ABI, calling functions will allocate 16 bytes on the stack for the called function to write out register arguments. When enabled, this optimization will suppress the allocation of the frame header if it can be determined that it is unused.

This optimization is off by default at all optimization levels.

• -mlxc1-sxc1, -mno-lxc1-sxc1 — When applicable, enable / disable the generation of lwxc1, swxc1, ldxc1, sdxc1 instructions. Enabled by default.
• -mmadd4, -mno-madd4 — When applicable, enable / disable the generation of 4-operand madd.s, madd.d and related instructions. Enabled by default.

---

✨ MMIX Options ✨

These options are defined for the MMIX, and they're lowkey iconic:

• -mlibfuncs, -mno-libfuncs — Specify that intrinsic library functions are being compiled, passing all values in registers, no matter the size.
• -mepsilon, -mno-epsilon — Generate floating-point comparison instructions that compare with respect to the rE epsilon register.
• -mabi=mmixware, -mabi=gnu — Generate code that passes function parameters and return values that (in the called function) are seen as registers $0 and up, as opposed to the GNU ABI which uses global registers $231 and up.
• -mzero-extend, -mno-zero-extend — When reading data from memory in sizes shorter than 64 bits, use / do not use zero-extending load instructions by default, rather than sign-extending ones.
• -mknuthdiv, -mno-knuthdiv — Make the result of a division yielding a remainder have the same sign as the divisor. With the default, -mno-knuthdiv, the sign of the remainder follows the sign of the dividend. Both methods are arithmetically valid, the latter being almost exclusively used.
• -mtoplevel-symbols, -mno-toplevel-symbols — Prepend / do not prepend a : to all global symbols, so the assembly code can be used with the PREFIX assembly directive.
• -melf — Generate an executable in the ELF format, rather than the default mmo format used by the mmix simulator.
• -mbranch-predict, -mno-branch-predict — Use / do not use the probable-branch instructions, when static branch prediction indicates a probable branch.
• -mbase-addresses, -mno-base-addresses — Generate / do not generate code that uses base addresses. Using a base address automatically generates a request (handled by the assembler and the linker) for a constant to be set up in a global register. The register is used for one or more base address requests within the range 0 to 255 from the value held in the register. This generally leads to short and fast code, but the number of different data items that can be addressed is limited. This means that a program that uses lots of static data may require -mno-base-addresses.
• -msingle-exit, -mno-single-exit — Force / do not force generated code to have a single exit point in each function.

---

✨ MN10300 Options ✨

These -m options are defined for Matsushita MN10300 architectures:

• -mmult-bug — Generate code to avoid bugs in the multiply instructions for the MN10300 processors. This is the default.
• -mno-mult-bug — Do not generate code to avoid bugs in the multiply instructions for the MN10300 processors.
• -mam33 — Generate code using features specific to the AM33 processor.
• -mno-am33 — Do not generate code using features specific to the AM33 processor. This is the default.
• -mam33-2 — Generate code using features specific to the AM33/2.0 processor.
• -mam34 — Generate code using features specific to the AM34 processor.
• -mtune=cpu-type — Use the timing characteristics of the indicated CPU type when scheduling instructions. This does not change the targeted processor type. The CPU type must be one of mn10300, am33, am33-2 or am34.
• -mreturn-pointer-on-d0 — When generating a function that returns a pointer, return the pointer in both a0 and d0. Otherwise, the pointer is returned only in a0, and attempts to call such functions without a prototype result in errors. Note that this option is on by default; use -mno-return-pointer-on-d0 to disable it.
• -mno-crt0 — Do not link in the C run-time initialization object file.
• -mrelax — Indicate to the linker that it should perform a relaxation optimization pass to shorten branches, calls and absolute memory addresses. This option only has an effect when used on the command line for the final link step.

⚠️ This option makes symbolic debugging impossible.

• -mliw — Allow the compiler to generate Long Instruction Word instructions if the target is the AM33 or later. This is the default. This option defines the preprocessor macro __LIW__.
• -mno-liw — Do not allow the compiler to generate Long Instruction Word instructions. This option defines the preprocessor macro __NO_LIW__.
• -msetlb — Allow the compiler to generate the SETLB and Lcc instructions if the target is the AM33 or later. This is the default. This option defines the preprocessor macro __SETLB__.
• -mno-setlb — Do not allow the compiler to generate SETLB or Lcc instructions. This option defines the preprocessor macro __NO_SETLB__.

---

✨ Moxie Options ✨

• -meb — Generate big-endian code. This is the default for moxie-*-* configurations.
• -mel — Generate little-endian code.
• -mmul.x — Generate mul.x and umul.x instructions. This is the default for moxiebox-*-* configurations.
• -mno-crt0 — Do not link in the C run-time initialization object file.

---

✨ MSP430 Options ✨

These options are defined for the MSP430, and they're kind of a lot but we got this:

• -masm-hex — Force assembly output to always use hex constants. Normally such constants are signed decimals, but this option is available for testsuite and/or aesthetic purposes.
• -mmcu= — Select the MCU to target. This is used to create a C preprocessor symbol based upon the MCU name, converted to upper case and pre- and post-fixed with __. This in turn is used by the msp430.h header file to select an MCU-specific supplementary header file.

The option also sets the ISA to use. If the MCU name is one that is known to only support the 430 ISA then that is selected, otherwise the 430X ISA is selected. A generic MCU name of msp430 can also be used to select the 430 ISA. Similarly the generic msp430x MCU name selects the 430X ISA.

In addition an MCU-specific linker script is added to the linker command line. The script's name is the name of the MCU with .ld appended. Thus specifying -mmcu=xxx on the gcc command line defines the C preprocessor symbol __XXX__ and cause the linker to search for a script called xxx.ld.

The ISA and hardware multiply supported for the different MCUs is hard-coded into GCC. However, an external devices.csv file can be used to extend device support beyond those that have been hard-coded.

GCC searches for the devices.csv file using the following methods in the given precedence order, where the first method takes precedence over the second which takes precedence over the third.

• Include path specified with -I and -L — devices.csv is searched for in each of the directories specified by include paths and linker library search paths.
• Path specified by the environment variable MSP430_GCC_INCLUDE_DIR — Define the value of the global environment variable MSP430_GCC_INCLUDE_DIR to the full path to the directory containing devices.csv, and GCC will search this directory for devices.csv. If devices.csv is found, this directory is also registered as an include path and linker library path. Header files and linker scripts in this directory can therefore be used without manually specifying -I and -L on the command line.
• The msp430-elf{,bare}/include/devices directory — Finally, GCC examines msp430-elf{,bare}/include/devices from the toolchain root directory. This directory does not exist in a default installation, but if the user has created it and copied devices.csv there, then the MCU data is read. As above, this directory is also registered as an include path and linker library path.

If none of the above search methods find devices.csv, then the hard-coded MCU data is used.

• -mwarn-mcu, -mno-warn-mcu — This option enables or disables warnings about conflicts between the MCU name specified by the -mmcu option and the ISA set by the -mcpu option and/or the hardware multiply support set by the -mhwmult option. It also toggles warnings about unrecognized MCU names. This option is on by default.
• -mcpu= — Specifies the ISA to use. Accepted values are msp430, msp430x and msp430xv2. This option is deprecated. The -mmcu= option should be used to select the ISA.
• -msim — Link to the simulator runtime libraries and linker script. Overrides any scripts that would be selected by the -mmcu= option.
• -mlarge — Use large-model addressing (20-bit pointers, 20-bit size_t).
• -msmall — Use small-model addressing (16-bit pointers, 16-bit size_t).
• -mrelax — This option is passed to the assembler and linker, and allows the linker to perform certain optimizations that cannot be done until the final link.
• mhwmult= — tells GCC what kind of hardware multiply your target actually has, bestie. Accepted values: none (no hardware multiply at all), 16bit (the OG 16-bit-only multiply that early MCUs shipped with), 32bit (the 16/32-bit multiply that later MCUs got), and f5series (the 16/32-bit multiply on F5-series MCUs). You can also pass auto, which makes GCC deduce the hardware multiply support from the MCU name you gave via -mmcu. And if you didn't pass a -mmcu at all, or GCC doesn't recognize the MCU name? Then no hardware multiply support is assumed. "auto" is the default setting, iykyk.

So how do these multiplies actually go down? Normally they're performed by calling a library routine — that saves space in the generated code. But when you're compiling at -O3 or higher, the hardware multiplier gets invoked inline. Bigger code, but faster. It's giving tradeoff 💅

Also important: the hardware multiply routines disable interrupts while they're running and restore the previous interrupt state when they finish. That makes them safe to use inside interrupt handlers and in normal code. Slay.

• -minrt — turn on a minimum runtime environment: no static initializers, no constructors. Made for memory-constrained devices. The compiler sticks special symbols into some objects that tell the linker and the runtime which code fragments are actually needed. Tiny footprint energy 🎀
• -mtiny-printf — enables the reduced-code-size "printf" and "puts" library functions. Heads up though: the tiny implementations are not reentrant, so use them with caution in multi-threaded applications. 😭

Support for streams is gone, and the string to be printed always goes to stdout via the "write" syscall. The string is not buffered before it's sent to write. No chill, no buffer.

This option requires Newlib Nano IO, so GCC must be configured with --enable-newlib-nano-formatted-io.

• -mmax-inline-shift= — takes an integer between 0 and 64 inclusive and sets the max number of inline shift instructions that should be emitted to perform a shift operation by a constant amount. Once you need to exceed that value, an mspabi helper function gets used instead. Default value is 4.

This only affects cases where a shift by multiple positions can't be completed with a single instruction (e.g. all shifts >1 on the 430 ISA).

And lowkey a key detail: shifts of a 32-bit value are at least twice as costly, so the value you pass for this option gets divided by 2 and that resulting value is used instead. Math is mathing 📐

• -mcode-region=, -mdata-region= — these tell the compiler where to place functions and data that don't have one of the "lower", "upper", "either" or "section" attributes. Possible values are "lower", "upper", "either" or "any". The first three behave like the corresponding attribute. The fourth possible value — "any" — is the default. It leaves placement entirely up to the linker script and how it assigns the standard sections (".text", ".data", etc) to the memory regions. Let the linker cook 👑
• -msilicon-errata= — passes on a request to the assembler to enable the fixes for the named silicon errata.
• -msilicon-errata-warn= — passes on a request to the assembler to enable warning messages when a silicon errata might need to be applied.
• -mwarn-devices-csv, -mno-warn-devices-csv — warn if devices.csv is not found or there are problems parsing it (default: on).

---

🌈 NDS32 Options 🌈

These options are defined for NDS32 implementations:

• -mbig-endian — generate code in big-endian mode.
• -mlittle-endian — generate code in little-endian mode.
• -mreduced-regs — use reduced-set registers for register allocation.
• -mfull-regs — use full-set registers for register allocation.
• -mcmov — generate conditional move instructions.
• -mno-cmov — do not generate conditional move instructions.
• -mext-perf — generate performance extension instructions.
• -mno-ext-perf — do not generate performance extension instructions.
• -mext-perf2 — generate performance extension 2 instructions.
• -mno-ext-perf2 — do not generate performance extension 2 instructions.
• -mext-string — generate string extension instructions.
• -mno-ext-string — do not generate string extension instructions.
• -mv3push — generate v3 push25/pop25 instructions.
• -mno-v3push — do not generate v3 push25/pop25 instructions.
• -m16-bit — generate 16-bit instructions.
• -mno-16-bit — do not generate 16-bit instructions.
• *-misr-vector-size=num* — specify the size of each interrupt vector, which must be 4 or 16.
• *-mcache-block-size=num* — specify the size of each cache block, which must be a power of 2 between 4 and 512.
• *-march=arch* — specify the name of the target architecture.
• *-mcmodel=code-model* — set the code model to one of:
  • small — all the data and read-only data segments must be within 512KB addressing space. The text segment must be within 16MB addressing space.
  • medium — the data segment must be within 512KB while the read-only data segment can be within 4GB addressing space. The text segment should be still within 16MB addressing space.
  • large — all the text and data segments can be within 4GB addressing space.
• -mctor-dtor — enable constructor/destructor feature.
• -mrelax — guide linker to relax instructions.

---

🧚 Nvidia PTX Options 🧚

These options are defined for Nvidia PTX:

• -m64 — ignored, but preserved for backward compatibility. Only 64-bit ABI is supported. Ghost of a flag, still haunting the codebase 💀
• *-march=architecture-string* — generate code for the specified PTX ISA target architecture. Valid architecture strings are sm_30, sm_35, sm_37, sm_52, sm_53, sm_61, sm_70, sm_75, sm_80, and sm_89. The default depends on how the compiler has been configured, see --with-arch.

This option sets the value of the preprocessor macro "__PTX_SM__"; for instance, for sm_35, it has the value 350. Numbers, not vibes ✨

• *-misa=architecture-string* — alias of -march=.
• -march-map=architecture-string — select the closest available -march= value that is **not more capable. For instance, for -march-map=sm_50 select -march=sm_37, and for -march-map=sm_53 select -march=sm_53.
• *-mptx=version-string* — generate code for the specified PTX ISA version. Valid version strings are 3.1, 4.1, 4.2, 5.0, 6.0, 6.3, 7.0, 7.3, and 7.8. The default PTX ISA version is the one that added support for the selected PTX ISA target architecture, see -march=, but at least 6.3, or 7.3 for -march=sm_52 and higher.

This option sets the values of the preprocessor macros "__PTX_ISA_VERSION_MAJOR__" and "__PTX_ISA_VERSION_MINOR__"; for instance, for 3.1 the macros have the values 3 and 1, respectively.

• -mmainkernel — link in code for a __main kernel. This is for stand-alone instead of offloading execution.
• -moptimize — apply partitioned execution optimizations. This is the default when any level of optimization is selected. Optimize era 🔥
• -msoft-stack, -mno-soft-stack — for -mno-soft-stack (the default, unless -mgomp has been specified), use PTX "native" stacks — that is, generate code that uses ".local" memory or PTX "alloca" directly for stack storage. Unless -mptx=7.3 or higher and -march=sm_52 or higher are active, variable-length arrays and dynamically allocating memory on the stack with "alloca" are not supported.

For -msoft-stack (implied by -mgomp), generate code that does not use ".local" memory or PTX "alloca" directly for stack storage. Instead, a per-warp stack pointer is maintained explicitly. This enables variable-length stack allocation (with variable-length arrays or "alloca"), and when global memory is used for underlying storage, makes it possible to access automatic variables from other threads, or with atomic instructions. This code generation variant is used for OpenMP offloading, but the option is exposed on its own for the purpose of testing the compiler; to generate code suitable for linking into programs using OpenMP offloading, use option -mgomp. Careful, bestie — testing purposes only for the standalone one 🫶

• -muniform-simt — switch to a code generation variant that allows executing all threads in each warp, while maintaining memory state and side effects as if only one thread in each warp was active outside of OpenMP SIMD regions. All atomic operations and calls to runtime (malloc, free, vprintf) are conditionally executed (iff current lane index equals the master lane index), and the register being assigned is copied via a shuffle instruction from the master lane. Outside of SIMD regions lane 0 is the master; inside, each thread sees itself as the master. Shared memory array "int __nvptx_uni[]" stores all-zeros or all-ones bitmasks for each warp, indicating current mode (0 outside of SIMD regions). Each thread can bitwise-and the bitmask at position "tid.y" with current lane index to compute the master lane index. Genuinely iconic logic ngl 👑
• -mgomp — generate code for use in OpenMP offloading: enables -msoft-stack and -muniform-simt options, and selects corresponding multilib variant.

---

🌸 OpenRISC Options 🌸

These options are defined for OpenRISC:

• *-mboard=name* — configure a board specific runtime. This will be passed to the linker for newlib board library linking. The default is "or1ksim".
• -mnewlib — this option is ignored; it is for compatibility purposes only. This used to select linker and preprocessor options for use with newlib. Rip to a legend, it's just vibes now 🙈
• -msoft-div, -mhard-div — select software or hardware divide ("l.div", "l.divu") instructions. This default is hardware divide.
• -msoft-mul, -mhard-mul — select software or hardware multiply ("l.mul", "l.muli") instructions. This default is hardware multiply.
• -msoft-float, -mhard-float — select software or hardware for floating point operations. The default is software.
• -mdouble-float — when -mhard-float is selected, enables generation of double-precision floating point instructions. By default functions from libgcc are used to perform double-precision floating point operations.
• -munordered-float — when you've already picked -mhard-float, this turns ON the generation of unordered floating point compare and set flag instructions, aka lf.sfun*. By default, GCC just calls into libgcc to do those unordered FP compare + set flag ops instead. 💅
• -mcmov — turn ON generation of conditional move instructions (l.cmov). By default you get the equivalent behaviour made out of set + branch instead, which is lowkey more instructions 🫠
• -mror — turn ON generation of rotate right instructions (l.ror). By default, functions from libgcc handle rotate right for you.
• -mrori — turn ON generation of rotate right with immediate instructions (l.rori). By default, libgcc functions do the rotate right with immediate work. ✨
• -msext — turn ON generation of sign extension instructions (l.ext*). By default, sign extension gets done with memory loads.
• -msfimm — turn ON generation of compare and set flag with immediate instructions (l.sf*i). By default, GCC generates extra instructions first to shove the immediate into a register 😭
• -mshftimm — turn ON generation of shift with immediate instructions (l.srai, l.srli, l.slli). By default, again, extra instructions get generated to store the immediate into a register first.
• -mcmodel=small — generate OpenRISC code for the small model: the GOT is capped at 64k. This is the default model, bestie, so if you didn't ask, this is what you got 👑
• -mcmodel=large — generate OpenRISC code for the large model: the GOT may grow all the way up to 4G in size. Big model energy fr fr.

---

PDP-11 Options 🕹️

These options are defined for the PDP-11:

• -mfpu — use hardware FPP floating point. This is the default. (Also note: FIS floating point on the PDP-11/40 is not supported.) It implies -m45.
• -msoft-float — do not use hardware floating point.
• -mac0 — return floating-point results in ac0 (fr0 in Unix assembler syntax).
• -mno-ac0 — return floating-point results in memory. This is the default.
• -m40 — generate code for a PDP-11/40. Implies -msoft-float -mno-split.
• -m45 — generate code for a PDP-11/45. This is the default.
• -m10 — generate code for a PDP-11/10. Implies -msoft-float -mno-split.
• -mint16, -mno-int32 — use 16-bit int. This is the default.
• -mint32, -mno-int16 — use 32-bit int.
• -msplit — the target has split instruction and data space. Implies -m45.
• -munix-asm — use Unix assembler syntax.
• -mdec-asm — use DEC assembler syntax.
• -mgnu-asm — use GNU assembler syntax. This is the default.
• -mlra — use the shiny new LRA register allocator. By default, the old reload allocator is the one doing the work 💀

---

PowerPC Options

These are listed under

---

PRU Options 🌱

These command-line options are defined for the PRU target:

• -minrt — link with a minimum runtime environment. This can significantly reduce the size of the final ELF binary, but heads up bestie: some standard C runtime features get removed 😳

This option disables support for static initializers and constructors. Beware that the compiler could still generate code with static initializers and constructors. It's up to you, the programmer, to make sure your source program won't use those features. No cap, that's on you.

The minimal startup code would not pass argc and argv arguments to main, so main must be declared as int main (void). Which, tbh, is already the norm for most firmware projects. ✨

• -mmcu=mcu — specify the PRU hardware variant to use. A correspondingly named spec file gets loaded, passing the memory region sizes to the linker and defining hardware-specific C macros.

Newlib provides only the sim spec, which is intended for running regression tests using a simulator. Specs for real hardware can be obtained by installing the GnuPruMcu (https://github.com/dinuxbg/gnuprumcu/) package.

• -mno-relax — make GCC pass the --no-relax command-line option to the linker instead of the --relax option.
• -mloop — allow (or do not allow) GCC to use the LOOP instruction.
• -mabi=variant — specify the ABI variant to output code for. -mabi=ti selects the unmodified TI ABI, while -mabi=gnu selects a GNU variant that copes more naturally with certain GCC assumptions. Here's the tea on the differences ☕⬇️

Function Pointer Size : TI ABI specifies that function (code) pointers are 16-bit, whereas GNU supports only 32-bit data and code pointers.

Optional Return Value Pointer : Function return values larger than 64 bits are passed by using a hidden pointer as the first argument of the function. TI ABI, though, mandates that the pointer can be NULL in case the caller is not using the returned value. GNU always passes and expects a valid return value pointer.

The current -mabi=ti implementation simply raises a compile error when any of the above code constructs is detected. As a consequence the standard C library cannot be built and it is omitted when linking with -mabi=ti. 😭

Relaxation is a GNU feature and for safety reasons is disabled when using -mabi=ti. The TI toolchain does not emit relocations for QBBx instructions, so the GNU linker cannot adjust them when shortening adjacent LDI32 pseudo instructions.

---

RISC-V Options 🦋

These command-line options are defined for RISC-V targets:

• -mbranch-cost=n — set the cost of branches to roughly n instructions.
• -mplt, -mno-plt — when generating PIC code, do or don't allow the use of PLTs. Ignored for non-PIC. The default is -mplt.
• -mabi=ABI-string — specify integer and floating-point calling convention. ABI-string contains two parts: the size of integer types and the registers used for floating-point types. For example -march=rv64ifd -mabi=lp64d means that long and pointers are 64-bit (implicitly defining int to be 32-bit), and that floating-point values up to 64 bits wide are passed in F registers. Contrast this with -march=rv64ifd -mabi=lp64f, which still allows the compiler to generate code that uses the F and D extensions but only allows floating-point values up to 32 bits long to be passed in registers; or -march=rv64ifd -mabi=lp64, in which no floating-point arguments will be passed in registers.

The default for this argument is system dependent, so users who want a specific calling convention should specify one explicitly. The valid calling conventions are: ilp32, ilp32f, ilp32d, lp64, lp64f, and lp64d. Some calling conventions are impossible to implement on some ISAs: for example, -march=rv32if -mabi=ilp32d is invalid because the ABI requires 64-bit values be passed in F registers, but F registers are only 32 bits wide. There are also the ilp32e ABI that can only be used with the rv32e architecture and the lp64e ABI that can only be used with the rv64e. Those ABIs are not well specified at present, and are subject to change. ⚠️

• -mfdiv, -mno-fdiv — do or don't use hardware floating-point divide and square root instructions. This requires the F or D extensions for floating-point registers. The default is to use them if the specified architecture has these instructions.
• -mfence-tso, -mno-fence-tso — do or don't use the fence.tso instruction, which is unimplemented on some processors (including those from T-Head). If the fence.tso instruction is not available then a stronger fence is used instead.
• -mdiv, -mno-div — do or don't use hardware instructions for integer division. This requires the M extension. The default is to use them if the specified architecture has these instructions.
• -misa-spec=ISA-spec-string — specify the version of the RISC-V Unprivileged (formerly User-Level) ISA specification to produce code conforming to. The possibilities for ISA-spec-string are:
  • 2.2 — produce code conforming to version 2.2.
  • 20190608 — produce code conforming to version 20190608.
  • 20191213 — produce code conforming to version 20191213.

The default is -misa-spec=20191213 unless GCC has been configured with --with-isa-spec= specifying a different default version.

• -march=ISA-string — generate code for given RISC-V ISA (e.g. rv64im). ISA strings must be lower-case. Examples include rv64i, rv32g, rv32e, and rv32imaf. Additionally, a special value help (-march=help) is accepted to list all supported extensions. 🎀

The syntax of the ISA string is defined as follows:

• "The string must start with rv32 or rv64, followed by" i, e, or g, referred to as the base ISA.
• "The subsequent part of the string is a list of extension names. Extension" names can be categorized as multi-letter (e.g. zba) and single-letter (e.g. v). Single-letter extensions can appear consecutively, but multi-letter extensions must be separated by underscores.
• "An underscore can appear anywhere after the base ISA. It has no specific" effect but is used to improve readability and can act as a separator.
• "Extension names may include an optional version number, following the" syntax majorpminor or major, (e.g. m2p1 or m2).

Supported extension are listed below:

• i — 2.0, 2.1 — Base integer extension.
• e — 2.0 — Reduced base integer extension.
• g — - — General-purpose computing base extension, g will expand to i, m, a, f, d, zicsr and zifencei.
• m — 2.0 — Integer multiplication and division extension.
• a — 2.0, 2.1 — Atomic extension.
• f — 2.0, 2.2 — Single-precision floating-point extension.
• d — 2.0, 2.2 — Double-precision floating-point extension.
• c — 2.0 — Compressed extension.
• h — 1.0 — Hypervisor extension.
• v — 1.0 — Vector extension.
• zicsr — 2.0 — Control and status register access extension.
• zifencei — 2.0 — Instruction-fetch fence extension.
• zicond — 1.0 — Integer conditional operations extension.
• za64rs — 1.0 — Reservation set size of 64 bytes.
• za128rs — 1.0 — Reservation set size of 128 bytes.
• zawrs — 1.0 — Wait-on-reservation-set extension.
• zba — 1.0 — Address calculation extension.

• zbb — version 1.0 — Basic bit manipulation extension.
• zbc — version 1.0 — Carry-less multiplication extension.
• zbs — version 1.0 — Single-bit operation extension.
• zfinx — version 1.0 — Single-precision floating-point in integer registers extension.
• zdinx — version 1.0 — Double-precision floating-point in integer registers extension.
• zhinx — version 1.0 — Half-precision floating-point in integer registers extension.
• zhinxmin — version 1.0 — Minimal half-precision floating-point in integer registers extension.
• zbkb — version 1.0 — Cryptography bit-manipulation extension.
• zbkc — version 1.0 — Cryptography carry-less multiply extension.
• zbkx — version 1.0 — Cryptography crossbar permutation extension.
• zkne — version 1.0 — AES Encryption extension.
• zknd — version 1.0 — AES Decryption extension.
• zknh — version 1.0 — Hash function extension.
• zkr — version 1.0 — Entropy source extension.
• zksed — version 1.0 — SM4 block cipher extension.
• zksh — version 1.0 — SM3 hash function extension.
• zkt — version 1.0 — Data independent execution latency extension.
• zk — version 1.0 — Standard scalar cryptography extension.
• zkn — version 1.0 — NIST algorithm suite extension.
• zks — version 1.0 — ShangMi algorithm suite extension.
• zihintntl — version 1.0 — Non-temporal locality hints extension.
• zihintpause — version 1.0 — Pause hint extension.
• zicboz — version 1.0 — Cache-block zero extension.
• zicbom — version 1.0 — Cache-block management extension.
• zicbop — version 1.0 — Cache-block prefetch extension.
• zic64b — version 1.0 — Cache block size isf 64 bytes.
• ziccamoa — version 1.0 — Main memory supports all atomics in A.
• ziccif — version 1.0 — Main memory supports instruction fetch with atomicity requirement.
• zicclsm — version 1.0 — Main memory supports misaligned loads/stores.
• ziccrse — version 1.0 — Main memory supports forward progress on LR/SC sequences.
• zicntr — version 2.0 — Standard extension for base counters and timers.
• zihpm — version 2.0 — Standard extension for hardware performance counters.
• ztso — version 1.0 — Total store ordering extension.
• zve32x — version 1.0 — Vector extensions for embedded processors.
• zve32f — version 1.0 — Vector extensions for embedded processors.
• zve64x — version 1.0 — Vector extensions for embedded processors.
• zve64f — version 1.0 — Vector extensions for embedded processors.
• zve64d — version 1.0 — Vector extensions for embedded processors.
• zvl32b — version 1.0 — Minimum vector length standard extensions
• zvl64b — version 1.0 — Minimum vector length standard extensions
• zvl128b — version 1.0 — Minimum vector length standard extensions
• zvl256b — version 1.0 — Minimum vector length standard extensions
• zvl512b — version 1.0 — Minimum vector length standard extensions
• zvl1024b — version 1.0 — Minimum vector length standard extensions
• zvl2048b — version 1.0 — Minimum vector length standard extensions
• zvl4096b — version 1.0 — Minimum vector length standard extensions
• zvbb — version 1.0 — Vector basic bit-manipulation extension.
• zvbc — version 1.0 — Vector carryless multiplication extension.
• zvkb — version 1.0 — Vector cryptography bit-manipulation extension.
• zvkg — version 1.0 — Vector GCM/GMAC extension.
• zvkned — version 1.0 — Vector AES block cipher extension.
• zvknha — version 1.0 — Vector SHA-2 secure hash extension.
• zvknhb — version 1.0 — Vector SHA-2 secure hash extension.
• zvksed — version 1.0 — Vector SM4 Block Cipher extension.
• zvksh — version 1.0 — Vector SM3 Secure Hash extension.
• zvkn — version 1.0 — Vector NIST Algorithm Suite extension, zvkn will expand to zvkned, zvknhb, zvkb and zvkt.
• zvknc — version 1.0 — Vector NIST Algorithm Suite with carryless multiply extension, zvknc will expand to zvkn and zvbc.
• zvkng — version 1.0 — Vector NIST Algorithm Suite with GCM extension, zvkng will expand to zvkn and zvkg.
• zvks — version 1.0 — Vector ShangMi algorithm suite extension, zvks will expand to zvksed, zvksh, zvkb and zvkt.
• zvksc — version 1.0 — Vector ShangMi algorithm suite with carryless multiplication extension, zvksc will expand to zvks and zvbc.
• zvksg — version 1.0 — Vector ShangMi algorithm suite with GCM extension, zvksg will expand to zvks and zvkg.
• zvkt — version 1.0 — Vector data independent execution latency extension.
• zfh — version 1.0 — Half-precision floating-point extension.
• zfhmin — version 1.0 — Minimal half-precision floating-point extension.
• zvfh — version 1.0 — Vector half-precision floating-point extension.
• zvfhmin — version 1.0 — Vector minimal half-precision floating-point extension.
• zvfbfmin — version 1.0 — Vector BF16 converts extension.
• zfa — version 1.0 — Additional floating-point extension.
• zmmul — version 1.0 — Integer multiplication extension.
• zca — version 1.0 — Integer compressed instruction extension.
• zcf — version 1.0 — Compressed single-precision floating point loads and stores extension.
• zcd — version 1.0 — Compressed double-precision floating point loads and stores extension.
• zcb — version 1.0 — Simple compressed instruction extension.
• zce — version 1.0 — Compressed instruction extensions for embedded processors.
• zcmp — version 1.0 — Compressed push pop extension.
• zcmt — version 1.0 — Table jump instruction extension.
• smaia — version 1.0 — Advanced interrupt architecture extension.
• smepmp — version 1.0 — PMP Enhancements for memory access and execution prevention on Machine mode.
• smstateen — version 1.0 — State enable extension.
• ssaia — version 1.0 — Advanced interrupt architecture extension for supervisor-mode.
• sscofpmf — version 1.0 — Count overflow & filtering extension.
• ssstateen — version 1.0 — State-enable extension for supervisor-mode.
• sstc — version 1.0 — Supervisor-mode timer interrupts extension.
• svinval — version 1.0 — Fine-grained address-translation cache invalidation extension.
• svnapot — version 1.0 — NAPOT translation contiguity extension.
• svpbmt — version 1.0 — Page-based memory types extension.
• xcvmac — version 1.0 — Core-V multiply-accumulate extension.
• xcvalu — version 1.0 — Core-V miscellaneous ALU extension.
• xcvelw — version 1.0 — Core-V event load word extension.
• xtheadba — version 1.0 — T-head address calculation extension.
• xtheadbb — version 1.0 — T-head basic bit-manipulation extension.
• xtheadbs — version 1.0 — T-head single-bit instructions extension.
• xtheadcmo — version 1.0 — T-head cache management operations extension.
• xtheadcondmov — version 1.0 — T-head conditional move extension.
• xtheadfmemidx — version 1.0 — T-head indexed memory operations for floating-point registers extension.
• xtheadfmv — version 1.0 — T-head double floating-point high-bit data transmission extension.
• xtheadint — version 1.0 — T-head acceleration interruption extension.
• xtheadmac — version 1.0 — T-head multiply-accumulate extension.
• xtheadmemidx — version 1.0 — T-head indexed memory operation extension.
• xtheadmempair — version 1.0 — T-head two-GPR memory operation extension.
• xtheadsync — version 1.0 — T-head multi-core synchronization extension.
• xventanacondops — version 1.0 — Ventana integer conditional operations extension.

---

okay bestie, housekeeping time 🫶

When -march= is not specified, use the setting from -mcpu.

If both -march and -mcpu= are not specified, the default for this argument is system dependent, users who want a specific architecture extensions should specify one explicitly.

When the RISC-V specifications define an extension as depending on other extensions, GCC will implicitly add the dependent extensions to the enabled extension set if they weren't added explicitly.

so if you're out here just vibing with defaults and wondering why your binary behaves different on your laptop vs your friend's Pi situation… that's the "system dependent" part. just say it explicitly, bestie. 🌸

• -mcpu=processor-string — Use architecture of and optimize the output for the given processor, specified by particular CPU name. Permissible values for this option are: sifive-e20, sifive-e21, sifive-e24, sifive-e31, sifive-e34, sifive-e76, sifive-s21, sifive-s51, sifive-s54, sifive-s76, sifive-u54, sifive-u74, sifive-x280, sifive-xp450, sifive-x670, thead-c906, tt-ascalon-x, tt-ascalon-d8, tt-ascalon-xg, xiangshan-nanhu.

Note that -mcpu does not override -march or -mtune.

THE WAY THAT it does not override -march or -mtune 😭 read that twice, it will save you a whole afternoon of confusion, iykyk

• -mtune=processor-string — Optimize the output for the given processor, specified by microarchitecture or particular CPU name. Permissible values for this option are: generic-ooo, rocket, sifive-3-series, sifive-5-series, sifive-7-series, size, sifive-p400-series, sifive-p600-series, and all valid options for -mcpu=.

When -mtune= is not specified, use the setting from -mcpu, the default is rocket if both are not specified.

The size choice is not intended for use by end-users. This is used when -Os is specified. It overrides the instruction cost info provided by -mtune=, but does not override the pipeline info. This helps reduce code size while still giving good performance.

tl;dr: size is not for you, it's a little secret between GCC and -Os. Don't come for it. 💅

• -mpreferred-stack-boundary=num — Attempt to keep the stack boundary aligned to a 2 raised to num byte boundary. If -mpreferred-stack-boundary is not specified, the default is 4 (16 bytes or 128-bits).

math check bestie ✨ 2 to the num, so num 4 gives you 16 bytes, which is 128 bits. Iconic power-of-two behaviour, no cap.

⚠️ Warning, bestie, read this twice: if you use this switch, then you must build all modules with the same value, including any libraries. That means the system libraries and the startup modules too. Everything. Same value. No rogue modules out here doing their own thing. 😭

• -msmall-data-limit=n — puts global and static data that's smaller than n bytes into a special section (on some targets).
• -msave-restore, -mno-save-restore — do / don't use smaller-but-slower prologue and epilogue code that relies on library function calls. The default is fast inline prologues and epilogues, so speed stays the main character unless you say otherwise. 💅
• -mmovcc, -mno-movcc — do / don't produce branchless conditional-move code sequences, even on targets that don't have special instructions for conditional operations. If it's enabled, sequences of ALU operations get produced using base integer ISA instructions where that's profitable.
• -minline-atomics, -mno-inline-atomics — do / don't use smaller-but-slower subword atomic emulation code that calls libatomic functions. The default is fast inline subword atomics that need no libatomic at all. obsessed with not linking extra stuff tbh ✨
• -minline-strlen, -mno-inline-strlen — do / do not attempt to inline strlen calls if possible. Inlining will only be done if the string is properly aligned and instructions for accelerated processing are available. The default is to not inline strlen calls.
• -minline-strcmp, -mno-inline-strcmp — do / do not attempt to inline strcmp calls if possible. Inlining will only be done if the strings are properly aligned and instructions for accelerated processing are available. The default is to not inline strcmp calls.

The --param riscv-strcmp-inline-limit=n parameter controls the maximum number of bytes compared by the inlined code. The default value is 64.

• -minline-strncmp, -mno-inline-strncmp — do / do not attempt to inline strncmp calls if possible. Inlining will only be done if the strings are properly aligned and instructions for accelerated processing are available. The default is to not inline strncmp calls.

The --param riscv-strcmp-inline-limit=n parameter controls the maximum number of bytes compared by the inlined code. The default value is 64.

• -mshorten-memrefs, -mno-shorten-memrefs — do / do not attempt to make more use of compressed load/store instructions by replacing a load/store of base register + large offset with a new load/store of new base + small offset. If the new base gets stored in a compressed register, then the new load/store can be compressed. Currently targets 32-bit integer load/stores only.
• -mstrict-align, -mno-strict-align — do not / do generate unaligned memory accesses. The default is set depending on whether the processor we're optimizing for supports fast unaligned access or not. 🦋
• -mscalar-strict-align, -mno-scalar-strict-align — do not / do generate unaligned memory accesses. The default is set depending on whether the processor we're optimizing for supports fast unaligned access or not. And plot twist: this is an alias for -mstrict-align.
• -mvector-strict-align, -mno-vector-strict-align — do not / do generate unaligned vector memory accesses. The default is set to off unless the processor we're optimizing for explicitly supports element-misaligned vector memory access.
• -mcmodel=medlow — generate code for the medium-low code model. The program and its statically defined symbols must lie within a single 2 GiB address range and must lie between absolute addresses -2 GiB and +2 GiB. Programs can be statically or dynamically linked. This is the default code model unless GCC has been configured with --with-cmodel= specifying a different default code model.
• -mcmodel=medany — generate code for the medium-any code model. The program and its statically defined symbols must be within any single 2 GiB address range. Programs can be statically or dynamically linked.

The code generated by the medium-any code model is position-independent, but it is not guaranteed to function correctly when linked into position-independent executables or libraries. So like… read that again before you ship it. 🙈

• -mcmodel=large — generate code for a large code model, which has no restrictions on size or placement of symbols. Free-range symbols, iconic.
• -mexplicit-relocs, -mno-exlicit-relocs — use or do not use assembler relocation operators when dealing with symbolic addresses. The alternative is to use assembler macros instead, which may limit optimization. (Yes, the second one is spelled -mno-exlicit-relocs in the original — that's not a typo of mine, do not "fix" it. 💀)
• -mrelax, -mno-relax — take advantage of linker relaxations to reduce the number of instructions required to materialize symbol addresses. The default is to take advantage of linker relaxations. Fewer instructions = the vibe. ✨
• -mriscv-attribute, -mno-riscv-attribute — emit (do not emit) RISC-V attribute to record extra information into ELF objects. This feature requires at least binutils 2.32.
• -mcsr-check, -mno-csr-check` — enables or disables the CSR checking.
• -malign-data=type — control how GCC aligns variables and constants of array, structure, or union types. Supported values for type are xlen, which uses x register width as the alignment value, and natural, which uses natural alignment. xlen is the default.
• -mbig-endian — generate big-endian code. This is the default when GCC is configured for a riscv64be-*-* or riscv32be-*-* target.
• -mlittle-endian — generate little-endian code. This is the default when GCC is configured for a riscv64-*-* or riscv32-*-* but not a riscv64be-*-* or riscv32be-*-* target.
• -mstack-protector-guard=guard, -mstack-protector-guard-reg=reg, -mstack-protector-guard-offset=offset — generate stack protection code using a canary at guard. Supported locations are global for a global canary or tls for a per-thread canary in the TLS block.

With the latter choice, the options -mstack-protector-guard-reg=reg and -mstack-protector-guard-offset=offset furthermore specify which register to use as base register for reading the canary, and from what offset from that base register. There is no default register or offset, as this is entirely for use within the Linux kernel. Kernel girlies, this one's for you. 🫶

• -mtls-dialect=desc — use TLS descriptors as the thread-local storage mechanism for dynamic accesses of TLS variables.
• -mtls-dialect=trad — use traditional TLS as the thread-local storage mechanism for dynamic accesses of TLS variables. This is the default.

---

🎀 RL78 Options 🎀

• -msim — links in additional target libraries to support operation within a simulator.
• -mmul=none, -mmul=g10, -mmul=g13, -mmul=g14, -mmul=rl78 — specifies the type of hardware multiplication and division support to be used. The simplest is "none", which uses software for both multiplication and division. This is the default. The "g13" value is for the hardware multiply/divide peripheral found on the RL78/G13 (S2 core) targets. The "g14" value selects the use of the multiplication and division instructions supported by the RL78/G14 (S3 core) parts. The value "rl78" is an alias for "g14" and the value "mg10" is an alias for "none".

In addition, a C preprocessor macro is defined, based upon the setting of this option. Possible values are: __RL78_MUL_NONE__, __RL78_MUL_G13__ or __RL78_MUL_G14__.

• -mcpu=g10, -mcpu=g13, -mcpu=g14, -mcpu=rl78 — specifies the RL78 core to target. The default is the G14 core, also known as an S3 core or just RL78. The G13 or S2 core does not have multiply or divide instructions; instead it uses a hardware peripheral for these operations. The G10 or S1 core does not have register banks, so it uses a different calling convention.

If this option is set, it also selects the type of hardware multiply support to use, unless this is overridden by an explicit -mmul=none option on the command line. So specifying -mcpu=g13 enables the use of the G13 hardware multiply peripheral, and specifying -mcpu=g10 disables the use of hardware multiplications altogether.

Note: although the RL78/G14 core is the default target, specifying -mcpu=g14 or -mcpu=rl78 on the command line does change the behavior of the toolchain, since it also enables G14 hardware multiply support. If these options are not specified on the command line, then software multiplication routines will be used even though the code targets the RL78 core. This is for backwards compatibility with older toolchains which did not have hardware multiply and divide support. History really said "hold my behavior." 😭

In addition, a C preprocessor macro is defined, based upon the setting of this option. Possible values are: __RL78_G10__, __RL78_G13__ or __RL78_G14__.

• -mg10, -mg13, -mg14, -mrl78 — these are aliases for the corresponding -mcpu= option. They are provided for backwards compatibility. Legacy but still here, we love a survivor. 👑
• -mallregs — allow the compiler to use all of the available registers. By default registers r24..r31 are reserved for use in interrupt handlers. With this option enabled, these registers can be used in ordinary functions as well.
• -m64bit-doubles, -m32bit-doubles — make the double data type be 64 bits (-m64bit-doubles) or 32 bits (-m32bit-doubles) in size. The default is -m32bit-doubles.
• -msave-mduc-in-interrupts, -mno-save-mduc-in-interrupts — specifies that interrupt handler functions should preserve the MDUC registers. This is only necessary if normal code might use the MDUC registers, for example because it performs multiplication and division operations. The default is to ignore the MDUC registers, as this makes the interrupt handlers faster. The target option -mg13 needs to be passed for this to work, since this feature is only available on the G13 target (S2 core). The MDUC registers will only be saved if the interrupt handler performs a multiplication or division operation or it calls another function.

---

💜 IBM RS/6000 and PowerPC Options 💜

These -m options are defined for the IBM RS/6000 and PowerPC:

• -mpowerpc-gpopt, -mno-powerpc-gpopt
• -mpowerpc-gfxopt, -mno-powerpc-gfxopt
• -mpowerpc64, -mno-powerpc64
• -mmfcrf, -mno-mfcrf
• -mpopcntb, -mno-popcntb
• -mpopcntd, -mno-popcntd
• -mfprnd, -mno-fprnd
• -mcmpb, -mno-cmpb
• -mhard-dfp, -mno-hard-dfp — okay so these are literally your "which instructions does my processor even have bestie" knobs 💅. The default value of each one gets decided back when GCC is being configured, so it's giving ✨mystery✨. But if you pass -mcpu=cpu_type, that overrides all of these. And fr fr, the docs are like "please just use -mcpu=cpu_type, stop touching these individually", so take the hint 🫶.

Let's go one by one, no skipping:

• -mpowerpc-gpopt lets GCC use the optional PowerPC architecture instructions in the General Purpose group, floating-point square root included 🌸.
• -mpowerpc-gfxopt lets GCC use the optional PowerPC architecture instructions in the Graphics group, floating-point select included.
• -mmfcrf lets GCC generate that move-from-condition-register-field instruction, the one living on the POWER4 processor and any other processor that supports the PowerPC V2.01 architecture 👑.
• -mpopcntb lets GCC generate the popcount and double-precision FP reciprocal estimate instruction, implemented on the POWER5 processor and other PowerPC V2.02 architecture-supporting processors.
• -mpopcntd lets GCC generate the popcount instruction implemented on the POWER7 processor and other PowerPC V2.06 architecture processors.
• -mfprnd lets GCC generate the FP round-to-integer instructions implemented on the POWER5+ processor and other PowerPC V2.03 architecture processors.
• -mcmpb lets GCC generate the compare bytes instruction implemented on the POWER6 processor and other PowerPC V2.05 architecture processors.
• -mhard-dfp lets GCC generate the decimal floating-point instructions implemented on some POWER processors 🌈.

And -mpowerpc64? That one lets GCC generate the extra 64-bit instructions found in the full PowerPC64 architecture and treat GPRs as 64-bit, doubleword quantities. GCC defaults to -mno-powerpc64, bestie, so don't come for me when it's off 🙈.

• -mcpu=cpu_type — sets the architecture type, register usage, AND instruction scheduling parameters for machine type cpu_type. It's the whole package 💖. The supported values for cpu_type are:

401, 403, 405, 405fp, 440, 440fp, 464, 464fp, 476, 476fp, 505,
601, 602, 603, 603e, 604, 604e, 620, 630, 740, 7400, 7450, 750, 801,
821, 823, 860, 970, 8540, a2, e300c2, e300c3, e500mc, e500mc64,
e5500, e6500, ec603e, G3, G4, G5, titan, power3, power4, power5,
power5+, power6, power6x, power7, power8, power9, power10, power11,
powerpc, powerpc64, powerpc64le, rs64, and native

Now let's break it down, because some of these are special ⭐:

• -mcpu=powerpc, -mcpu=powerpc64, and -mcpu=powerpc64le specify pure 32-bit PowerPC (either endian), 64-bit big endian PowerPC, and 64-bit little endian PowerPC architecture machine types respectively, and they assume an appropriate, generic processor model for scheduling purposes 🎀.
• Specifying native as the cpu type detects and selects the architecture option matching the host processor of the system doing the compilation. So iconic. But heads up: -mcpu=native has no effect if GCC doesn't recognize the processor 💀.
• Every other option names a specific processor. Code generated under those options runs best on that processor, and may not run at all on others. Not to be dramatic but it really said "this is my house" 😭.

The -mcpu options automatically enable or disable the following options:

-maltivec  -mfprnd  -mhard-float  -mmfcrf  -mmultiple -mpopcntb
-mpopcntd  -mpowerpc64 -mpowerpc-gpopt  -mpowerpc-gfxopt -mmulhw
-mdlmzb  -mmfpgpr  -mvsx -mcrypto  -mhtm  -mpower8-fusion
-mquad-memory  -mquad-memory-atomic  -mfloat128 -mfloat128-hardware
-mprefixed -mpcrel -mmma -mrop-protect

Real talk though: the particular set of options for any given CPU varies between compiler versions, depending on whatever setting seems to produce optimal code for that CPU. It doesn't necessarily reflect the actual hardware's capabilities, iykyk 🙈. So don't @ me.

If you wanna force one individual option to a particular value, you can specify it after the -mcpu option, like this:

-mcpu=970 -mno-altivec

And on AIX specifically? The -maltivec and -mpowerpc64 options are not enabled or disabled by -mcpu at present, because AIX doesn't have full support for them. You may still enable or disable them individually if you're sure it'll work in your environment 🫶.

• -mtune=cpu_type — sets the instruction scheduling parameters for machine type cpu_type, but does not set the architecture type or register usage, unlike what -mcpu=cpu_type does. Same values for cpu_type apply here, they match -mcpu. And if you specify both? The generated code uses the architecture and registers from -mcpu, but the scheduling parameters from -mtune. Teamwork bestie 🤝.
• -mcmodel=small — generates PowerPC64 code for the small model: the TOC is limited to 64k.
• -mcmodel=medium — generates PowerPC64 code for the medium model: the TOC and other static data may be up to a total of 4G in size. This is the default for 64-bit Linux ✨.
• -mcmodel=large — generates PowerPC64 code for the large model: the TOC may be up to 4G in size. Other data and code is only limited by the 64-bit address space. Go big or go home 💅.
• -maltivec, -mno-altivec — generate code that uses (or doesn't use) AltiVec instructions, and also toggle the built-in functions that allow more direct access to the AltiVec instruction set. You may also need to set -mabi=altivec to adjust the current ABI with AltiVec ABI enhancements 🎀.

When -maltivec is used, the element order for AltiVec intrinsics like vec_splat, vec_extract, and vec_insert matches array element order corresponding to the endianness of the target 🌈. Translation: element zero identifies the leftmost element in a vector register when targeting a big-endian platform, and the rightmost element in a vector register when targeting a little-endian platform. Endianness really said "I contain multitudes" 😭.

• -mvrsave, -mno-vrsave — generate VRSAVE instructions when generating AltiVec code.
• -msecure-plt — generates code that lets ld and ld.so build executables and shared libraries with non-executable .plt and .got sections. This is a PowerPC 32-bit SYSV ABI option 💖.
• -mbss-plt — generates code that uses a BSS .plt section that ld.so fills in, and requires .plt and .got sections that are both writable and executable 😬. Also a PowerPC 32-bit SYSV ABI option.
• -msplit-patch-nops — when you're adding NOPs for a patchable area via the -fpatchable-function-entry option, this emits the "before" NOPs in front of the global entry point and the "after" NOPs after the local entry point. This makes the sequence of NOPs not consecutive when a global entry point is generated ✨. Without this option, the NOPs get emitted directly before and after the local entry point, which makes them consecutive but moves the global and local entry points further apart. And if only a single entry point is generated? This option has no effect. Filler arc, no plot 🙈.
• -misel, -mno-isel — this switch enables or disables the generation of ISEL instructions.
• -mvsx, -mno-vsx — generate code that uses (does not use) vector/scalar (VSX) instructions, and also enable the built-in functions that allow more direct access to the VSX instruction set 🔥.
• -mcrypto, -mno-crypto — enable (disable) the built-in functions that allow direct access to the cryptographic instructions added in version 2.07 of the PowerPC ISA 🔐.
• -mhtm, -mno-htm — enable (disable) the built-in functions that allow direct access to the Hardware Transactional Memory (HTM) instructions added in version 2.07 of the PowerPC ISA.
• -mpower8-fusion, -mno-power8-fusion — generate code that keeps (does not keeps, yes that's the manual's own wording, love that for them 🫶) some integer operations adjacent so the instructions can be fused together on power8 and later processors.
• -mquad-memory, -mno-quad-memory — generate code that uses (does not use) the non-atomic quad word memory instructions. Careful bestie: -mquad-memory requires use of 64-bit mode ⚠️.
• -mquad-memory-atomic, -mno-quad-memory-atomic — generate code that uses (does not use) the atomic quad word memory instructions. Same deal: -mquad-memory-atomic requires use of 64-bit mode.
• -mfloat128, -mno-float128 — enable/disable the __float128 keyword for IEEE 128-bit floating point, and use either software emulation or hardware instructions for IEEE 128-bit floating point 💫.

The VSX instruction set (-mvsx) must be enabled to use the IEEE 128-bit floating point support. And IEEE 128-bit floating point is only supported on Linux.

The default for -mfloat128 is enabled on PowerPC Linux systems using the VSX instruction set, and disabled on other systems.

If you use the ISA 3.0 instruction set (-mcpu=power9) on a 64-bit system, the IEEE 128-bit floating point support will also enable generation of ISA 3.0 IEEE 128-bit floating point instructions ⭐. Otherwise — if you don't specify generating ISA 3.0 instructions, or you're targeting a 32-bit big endian system — IEEE 128-bit floating point gets done with software emulation 😭.

• -mfloat128-hardware, -mno-float128-hardware — enable/disable using ISA 3.0 hardware instructions to support the __float128 data type 🔥.

The default for -mfloat128-hardware is enabled on PowerPC Linux systems using the ISA 3.0 instruction set, and disabled on other systems.

• -m32, -m64 — generate code for 32-bit or 64-bit environments of Darwin and SVR4 targets (GNU/Linux included 💖). The 32-bit environment sets int, long and pointer to 32 bits and generates code that runs on any PowerPC variant — queen of compatibility honestly 👑. The 64-bit environment sets int to 32 bits and long and pointer to 64 bits, and generates code for PowerPC64, same as -mpowerpc64.
• -mfull-toc, -mno-fp-in-toc, -mno-sum-in-toc, -mminimal-toc — these four are all about how the TOC (Table Of Contents) gets built, and bestie, every single executable file gets one of these. -mfull-toc is the default, and it means GCC sets aside at least one TOC entry for every unique non-automatic variable reference in your program. It also drops floating-point constants in there too. Cute! Except… only 16,384 entries fit in the TOC. That's the whole budget. 😭

So if the linker starts yelling at you that you've overflowed the available TOC space, you've got two levers: -mno-fp-in-toc stops GCC from putting floating-point constants in the TOC, and -mno-sum-in-toc makes GCC generate code to calculate the sum of an address and a constant at run time instead of parking that sum in the TOC. You can use one or both, and each one makes GCC spit out code that's very slightly slower and larger — the trade is conserving TOC space.

Still overflowing even with both? Okay, bring out -mminimal-toc. That one makes GCC create only one TOC entry for every file. Downside: the code gets slower and bigger, but it uses extremely little TOC space. You probably want this only on files that contain less frequently-executed code. iykyk 💅

• -maix64, -maix32 — turn on the 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit long type, and all the infrastructure needed to support them. -maix64 implies -mpowerpc64, and -maix32 kills the 64-bit ABI and implies -mno-powerpc64. GCC defaults to -maix32 btw.
• -mxl-compat, -mno-xl-compat — this one makes your code cosplay the IBM XL compiler semantics more closely when you're using the AIX-compatible ABI. It passes floating-point arguments to prototyped functions beyond the register save area (RSA) on the stack, on top of the argument FPRs. It does not assume that the most significant double in a 128-bit long double value is properly rounded when comparing values and converting to double. And it uses XL symbol names for long double support routines.

Okay so here's the lore 🫶: the AIX calling convention got extended — but not initially documented — to handle an obscure K&R C case where you call a function that takes the address of its arguments with fewer arguments than declared. IBM XL compilers read floating-point arguments that don't fit in the RSA from the stack when a subroutine is compiled without optimization. Since always stashing floating-point arguments on the stack is inefficient and rarely needed, this option is not enabled by default and only matters when you're calling subroutines compiled by IBM XL compilers without optimization.

• -mpe — support for the IBM RS/6000 SP Parallel Environment (PE). It links an application written to use message passing with special startup code so the app can actually run. For this to work the system must have PE installed in the standard location /usr/lpp/ppe.poe/, or the specs file has to be overridden with the -specs= option to point at the right directory. Heads up though — the Parallel Environment does not support threads, so -mpe and -pthread are incompatible. 💀
• -malign-natural, -malign-power — on AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux: -malign-natural overrides the ABI-defined alignment of larger types (floating-point doubles, thinking of you 🌸) so they sit on their natural size-based boundary. -malign-power tells GCC to follow the ABI-specified alignment rules instead. GCC defaults to whatever standard alignment the ABI defines.

On 64-bit Darwin, natural alignment is the default, and -malign-power is not supported there.

• -msoft-float, -mhard-float — generate code that does not use (-msoft-float) or does use (-mhard-float) the floating-point register set. Software floating-point emulation is provided if you use -msoft-float, and you've gotta pass the option to GCC when linking too.
• -mmultiple, -mno-multiple — generate code that uses (-mmultiple) or does not use (-mno-multiple) the load multiple word instructions and the store multiple word instructions. These are generated by default on POWER systems, and not generated on PowerPC systems. Do not use -mmultiple on little-endian PowerPC systems, since those instructions don't work when the processor is in little-endian mode. The exceptions are PPC740 and PPC750, which permit these instructions in little-endian mode. ⭐
• -mupdate, -mno-update — generate code that uses (or doesn't use) the load or store instructions that update the base register to the address of the calculated memory location. These are generated by default. If you use -mno-update, there's a small window between the time the stack pointer gets updated and the time the address of the previous frame is stored — which means code that walks the stack frame across interrupts or signals may get corrupted data. Watch out for that one 😭
• -mavoid-indexed-addresses, -mno-avoid-indexed-addresses — generate code that tries to avoid (or doesn't try to avoid) using indexed load or store instructions. These instructions can take a performance hit on Power6 processors in certain situations, like when stepping through large arrays that cross a 16M boundary. This option is enabled by default when targeting Power6 and disabled otherwise.
• -mfused-madd, -mno-fused-madd — generate code that uses (or doesn't use) the floating-point multiply and accumulate instructions. These are generated by default if hardware floating point is used. Fun fact: the machine-dependent -mfused-madd option is now mapped to the machine-independent -ffp-contract=fast option, and -mno-fused-madd is mapped to -ffp-contract=off. ✨
• -mmulhw, -mno-mulhw — generate code that uses (or doesn't use) the half-word multiply and multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors. These are generated by default when targeting those processors.
• -mdlmzb, -mno-dlmzb — generate code that uses (or doesn't use) the string-search dlmzb instruction on the IBM 405, 440, 464 and 476 processors. Generated by default when targeting those processors.
• -mno-bit-align, -mbit-align — on System V.4 and embedded PowerPC systems, don't (or do) force structures and unions that contain bit-fields to be aligned to the base type of the bit-field.

For example, by default a structure containing nothing but 8 unsigned bit-fields of length 1 is aligned to a 4-byte boundary and has a size of 4 bytes. But using -mno-bit-align, the structure gets aligned to a 1-byte boundary and is 1 byte in size. Slay 💅

• -mno-strict-align, -mstrict-align — on System V.4 and embedded PowerPC systems, don't (or do) assume that unaligned memory references are handled by the system.
• -mrelocatable, -mno-relocatable — generate code that allows (or doesn't allow) a static executable to be relocated to a different address at run time. A simple embedded PowerPC system loader should relocate the entire contents of .got2 and the 4-byte locations listed in the .fixup section — which is a table of 32-bit addresses generated by this option. For this to work, all objects linked together must be compiled with -mrelocatable or -mrelocatable-lib. Also, -mrelocatable code aligns the stack to an 8-byte boundary.
• -mrelocatable-lib, -mno-relocatable-lib — like -mrelocatable, -mrelocatable-lib generates a .fixup section to let static executables be relocated at run time, but it does not use the smaller stack alignment of -mrelocatable. The cute part: objects compiled with -mrelocatable-lib may be linked with objects compiled with any combination of the -mrelocatable options. 🫶
• -mno-toc, -mtoc — on System V.4 and embedded PowerPC systems, don't (or do) assume that register 2 contains a pointer to a global area pointing to the addresses used in the program.
• -mlittle, -mlittle-endian — on System V.4 and embedded PowerPC systems, compile code for the processor in little-endian mode. -mlittle-endian is the same as -mlittle.
• -mbig, -mbig-endian — on System V.4 and embedded PowerPC systems, compile code for the processor in big-endian mode. -mbig-endian is the same as -mbig.
• -mdynamic-no-pic — on Darwin / macOS systems, compile code so that it is not relocatable, but its external references are relocatable. The resulting code is suitable for applications, but not shared libraries. 🙈
• -msingle-pic-base — treat the register used for PIC addressing as read-only, rather than loading it in the prologue for each function. The runtime system is responsible for initializing this register with an appropriate value before execution begins.
• *-mprioritize-restricted-insns=priority* — this option controls the priority assigned to dispatch-slot restricted instructions during the second scheduling pass. The argument priority takes the value 0, 1, or 2 to assign no, highest, or second-highest (respectively) priority to dispatch-slot restricted instructions.
• *-msched-costly-dep=dependence_type* — this option controls which dependences are considered costly by the target during instruction scheduling. The argument dependence_type takes one of the following values:
  • no — no dependence is costly.
  • all — all dependences are costly.
  • true_store_to_load — a true dependence from store to load is costly.
  • store_to_load — any dependence from store to load is costly.
  • number — any dependence for which the latency is greater than or equal to number is costly.
• *-minsert-sched-nops=scheme* — this option controls which NOP insertion scheme is used during the second scheduling pass. The argument scheme takes one of the following values:
  • no — don't insert NOPs.
  • pad — pad with NOPs any dispatch group that has vacant issue slots, according to the scheduler's grouping.
  • regroup_exact — insert NOPs to force costly dependent insns into separate groups. Insert exactly as many NOPs as needed to force an insn to a new group, according to the estimated processor grouping.
• number Insert NOPs to force costly dependent insns into separate groups. Insert number NOPs to force an insn to a new group.
• -mcall-sysv — on System V.4 and embedded PowerPC systems, compile code using calling conventions that stick to the March 1995 draft of the System V Application Binary Interface, PowerPC processor supplement. This is the default unless you configured GCC using powerpc-*-eabiaix. 💖
• -mcall-sysv-eabi, -mcall-eabi — specify both -mcall-sysv and -meabi. two-for-one, slay. ✨
• -mcall-sysv-noeabi — specify both -mcall-sysv and -mno-eabi.
• -mcall-aixdesc — on System V.4 and embedded PowerPC systems, compile code for the AIX operating system. 🎀
• -mcall-linux — on System V.4 and embedded PowerPC systems, compile code for the Linux-based GNU system.
• -mcall-freebsd — on System V.4 and embedded PowerPC systems, compile code for the FreeBSD operating system.
• -mcall-netbsd — on System V.4 and embedded PowerPC systems, compile code for the NetBSD operating system.
• -mcall-openbsd — on System V.4 and embedded PowerPC systems, compile code for the OpenBSD operating system.
• -mtraceback=traceback_type — pick the type of traceback table. valid values for traceback_type are full, part, and no. three options, no notes. 🦋
• -maix-struct-return — return all structures in memory, exactly like the AIX ABI says to.
• -msvr4-struct-return — return structures smaller than 8 bytes in registers, exactly like the SVR4 ABI says to.
• -mabi=abi-type — extend the current ABI with a particular extension, or yeet such extension back off. valid values are: altivec, no-altivec, ibmlongdouble, ieeelongdouble, elfv1, elfv2, and for AIX: vec-extabi, vec-default.
• -mabi=ibmlongdouble — change the current ABI to use IBM extended-precision long double. bestie, this is not likely to work if your system defaults to using IEEE extended-precision long double. 😭 and if you change the long double type away from IEEE extended-precision, the compiler throws a warning unless you use the -Wno-psabi option. requires -mlong-double-128 to be enabled. ⚠️
• -mabi=ieeelongdouble — change the current ABI to use IEEE extended-precision long double. same energy as above: not likely to work if your system defaults to using IBM extended-precision long double. if you change the long double type away from IBM extended-precision, the compiler issues a warning unless you use the -Wno-psabi option. requires -mlong-double-128 to be enabled. ⚠️
• -mabi=elfv1 — change the current ABI to use the ELFv1 ABI. this is the default ABI for big-endian PowerPC 64-bit Linux. overriding the default ABI requires special system support and is likely to fail in spectacular ways. 💀
• -mabi=elfv2 — change the current ABI to use the ELFv2 ABI. this is the default ABI for little-endian PowerPC 64-bit Linux. overriding the default ABI requires special system support and is likely to fail in spectacular ways. 💀
• -mgnu-attribute, -mno-gnu-attribute — emit .gnu_attribute assembly directives to set tag/value pairs in a .gnu.attributes section that spell out ABI variations in function parameters or return values. 🌸
• -mprototype, -mno-prototype — on System V.4 and embedded PowerPC systems, assume that all calls to variable argument functions are properly prototyped. otherwise the compiler must insert an instruction before every non-prototyped call to set or clear bit 6 of the condition code register ("CR") to say whether floating-point values are passed in the floating-point registers, in case the function takes variable arguments. with -mprototype, only calls to prototyped variable argument functions set or clear that bit. 👑
• -msim — on embedded PowerPC systems, assume that the startup module is called sim-crt0.o and that the standard C libraries are libsim.a and libc.a. this is the default for powerpc-*-eabisim configurations. ✨
• -mmvme — on embedded PowerPC systems, assume that the startup module is called crt0.o and the standard C libraries are libmvme.a and libc.a.
• -mads — on embedded PowerPC systems, assume that the startup module is called crt0.o and the standard C libraries are libads.a and libc.a.
• -myellowknife — on embedded PowerPC systems, assume that the startup module is called crt0.o and the standard C libraries are libyk.a and libc.a. 🍓
• -mvxworks — on System V.4 and embedded PowerPC systems, specify that you are compiling for a VxWorks system.
• -memb — on embedded PowerPC systems, set the "PPC_EMB" bit in the ELF flags header to indicate that eabi extended relocations are used.
• -meabi, -mno-eabi — on System V.4 and embedded PowerPC systems, do (or do not) adhere to the Embedded Applications Binary Interface (EABI), which is a set of modifications to the System V.4 specifications. selecting -meabi means the stack is aligned to an 8-byte boundary, a function "__eabi" is called from "main" to set up the EABI environment, and the -msdata option can use both "r2" and "r13" to point to two separate small data areas. selecting -mno-eabi means the stack is aligned to a 16-byte boundary, no EABI initialization function is called from "main", and the -msdata option only uses "r13" to point to a single small data area. the -meabi option is on by default if you configured GCC using one of the powerpc*-*-eabi* options. 💅
• -msdata=eabi — on System V.4 and embedded PowerPC systems, put small initialized "const" global and static data in the .sdata2 section, which is pointed to by register "r2". put small initialized non-"const" global and static data in the .sdata section, which is pointed to by register "r13". put small uninitialized global and static data in the .sbss section, which sits adjacent to the .sdata section. the -msdata=eabi option is incompatible with the -mrelocatable option. the -msdata=eabi option also sets the -memb option. 🫶
• -msdata=sysv — on System V.4 and embedded PowerPC systems, put small global and static data in the .sdata section, pointed to by register "r13". put small uninitialized global and static data in the .sbss section, adjacent to .sdata. the -msdata=sysv option is incompatible with the -mrelocatable option.
• -msdata=default, -msdata — on System V.4 and embedded PowerPC systems, if -meabi is used, compile code the same as -msdata=eabi; otherwise compile code the same as -msdata=sysv.
• -msdata=data — on System V.4 and embedded PowerPC systems, put small global data in the .sdata section. put small uninitialized global data in the .sbss section. but do not use register "r13" to address small data. this is the default behavior unless other -msdata options are used.
• -msdata=none, -mno-sdata — on embedded PowerPC systems, put all initialized global and static data in the .data section, and all uninitialized data in the .bss section.
• -mreadonly-in-sdata — put read-only objects in the .sdata section as well. this is the default. 🧚
• -mblock-move-inline-limit=num — inline all block moves (think calls to "memcpy" or structure copies) less than or equal to num bytes. the minimum value for num is 32 bytes on 32-bit targets and 64 bytes on 64-bit targets. the default value is target-specific. ✨
• -mblock-compare-inline-limit=num — generate non-looping inline code for all block compares (think calls to "memcmp" or structure compares) less than or equal to num bytes. if num is 0, all inline expansion (non-loop and loop) of block compare is disabled. the default value is target-specific.
• -mblock-compare-inline-loop-limit=num — generate an inline expansion using loop code for all block compares that are less than or equal to num bytes, but greater than the limit for non-loop inline block compare expansion. if the block length isn't constant, at most num bytes will be compared before "memcmp" is called to compare the rest of the block. the default value is target-specific.
• -mstring-compare-inline-limit=num — compare at most num string bytes with inline code. if the difference or end of string isn't found at the end of the inline compare, a call to "strcmp" or "strncmp" handles the rest of the comparison. the default is 64 bytes. 💖
• -G num — on embedded PowerPC systems, put global and static items less than or equal to num bytes into the small data or BSS sections instead of the normal data or BSS section. by default, num is 8. the -G num switch is also passed to the linker. all modules should be compiled with the same -G num value. 👑
• -mregnames, -mno-regnames — on System V.4 and embedded PowerPC systems, do (or do not) emit register names in the assembly language output using symbolic forms.
• -mlongcall, -mno-longcall — by default, assume that all calls are far away, so a longer and more expensive calling sequence is required. this is required for calls farther than 32 megabytes (33,554,432 bytes) from the current location. a short call is generated if the compiler knows the call cannot be that far away. this setting can be overridden by the "shortcall" function attribute, or by #pragma longcall(0).

some linkers can detect out-of-range calls and generate glue code on the fly. on those systems, long calls are unnecessary and generate slower code. as of this writing, the AIX linker can do this, as can the GNU linker for PowerPC/64. it's planned to add this feature to the GNU linker for 32-bit PowerPC systems as well. 🔥

on PowerPC64 ELFv2 and 32-bit PowerPC systems with newer GNU linkers, GCC can generate long calls using an inline PLT call sequence (see -mpltseq). PowerPC with -mbss-plt and PowerPC64 ELFv1 (big-endian) do not support inline PLT calls.

on Darwin/PPC systems, #pragma longcall generates jbsr callee, L42, plus a branch island (glue code). the two target addresses represent the callee and the branch island. the Darwin/PPC linker prefers the first address and generates a bl callee if the PPC bl instruction reaches the callee directly; otherwise, the linker generates bl L42 to call the branch island. the branch island is appended to the body of the calling function; it computes the full 32-bit address of the callee and jumps to it. 🌈

On Mach-O (Darwin) systems, this option tells the compiler to emit the glue for every direct call, and then the Darwin linker is the one who decides whether to actually use it or just yeet it into the void. 💅

Ngl, plot twist for the future: GCC may straight up ignore all longcall specifications when the linker is known to generate glue. So like, don't get too attached to your longcall vibes. 🦋

• -mpltseq, -mno-pltseq — implement (or don't implement, bestie, your call) -fno-plt and long calls using an inline PLT call sequence that supports lazy linking and long calls to functions in dlopen'd shared libraries. Inline PLT calls are only supported on PowerPC64 ELFv2 and 32-bit PowerPC systems with newer GNU linkers, and they're on by default if the support is detected when configuring GCC — plus, for 32-bit PowerPC, if GCC is configured with --enable-secureplt. ⚠️ Heads up: -mpltseq code and -mbss-plt 32-bit PowerPC relocatable objects may not be linked together. It's giving incompatible.
• -mtls-markers, -mno-tls-markers — mark (or don't mark) calls to "__tls_get_addr" with a relocation specifying the function argument. That relocation lets the linker reliably associate the function call with the argument setup instructions for TLS optimization, which in turn lets GCC schedule the sequence better. Teamwork, honestly. 🫶
• -mrecip, -mno-recip — turns on the reciprocal estimate and reciprocal square root estimate instructions, with extra Newton-Raphson steps to bump the precision, instead of doing a divide (or a square root and divide) for floating-point arguments. You should use -ffast-math when you use -mrecip — or at minimum -funsafe-math-optimizations, -ffinite-math-only, -freciprocal-math and -fno-trapping-math. 💀 Note that while the throughput of the sequence is generally higher than the throughput of the non-reciprocal instruction, the precision can drop by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994) for reciprocal square roots. So like, speedy but sloppy.
• -mrecip=opt — this one controls which reciprocal estimate instructions may be used. opt is a comma-separated list of options, and any of them may be preceded by a "!" to invert it:
  • all — enable all estimate instructions. the whole buffet ✨
  • default — enable the default instructions, equivalent to -mrecip.
  • none — disable all estimate instructions, equivalent to -mno-recip.
  • div — enable the reciprocal approximation instructions for both single and double precision.
  • divf — enable the single-precision reciprocal approximation instructions.
  • divd — enable the double-precision reciprocal approximation instructions.
  • rsqrt — enable the reciprocal square root approximation instructions for both single and double precision.
  • rsqrtf — enable the single-precision reciprocal square root approximation instructions.
  • rsqrtd — enable the double-precision reciprocal square root approximation instructions.

So for example, -mrecip=all,!rsqrtd enables all of the reciprocal estimate instructions, except for the "FRSQRTE", "XSRSQRTEDP", and "XVRSQRTEDP" instructions which handle the double-precision reciprocal square root calculations. See how the ! just flips it? obsessed. 💖

• -mrecip-precision, -mno-recip-precision — assume (or don't assume) that the reciprocal estimate instructions provide higher-precision estimates than is mandated by the PowerPC ABI. Selecting -mcpu=power6, -mcpu=power7 or -mcpu=power8 automatically selects -mrecip-precision. The double-precision square root estimate instructions are not generated by default on low-precision machines, since they don't provide an estimate that converges after three steps. no cap.
• -mveclibabi=type — specifies the ABI type to use for vectorizing intrinsics using an external library. The only type supported at present is mass, which specifies to use IBM's Mathematical Acceleration Subsystem (MASS) libraries for vectorizing intrinsics using external libraries. GCC currently emits calls to "acosd2", "acosf4", "acoshd2", "acoshf4", "asind2", "asinf4", "asinhd2", "asinhf4", "atan2d2", "atan2f4", "atand2", "atanf4", "atanhd2", "atanhf4", "cbrtd2", "cbrtf4", "cosd2", "cosf4", "coshd2", "coshf4", "erfcd2", "erfcf4", "erfd2", "erff4", "exp2d2", "exp2f4", "expd2", "expf4", "expm1d2", "expm1f4", "hypotd2", "hypotf4", "lgammad2", "lgammaf4", "log10d2", "log10f4", "log1pd2", "log1pf4", "log2d2", "log2f4", "logd2", "logf4", "powd2", "powf4", "sind2", "sinf4", "sinhd2", "sinhf4", "sqrtd2", "sqrtf4", "tand2", "tanf4", "tanhd2", and "tanhf4" when generating code for power7. Both -ftree-vectorize and -funsafe-math-optimizations must also be enabled. The MASS libraries must be specified at link time. ☝️ don't forget that last bit.
• -mfriz, -mno-friz — generate (or don't generate) the "friz" instruction when the -funsafe-math-optimizations option is used to optimize rounding of floating-point values to 64-bit integer and back to floating point. ⚠️ The "friz" instruction does not return the same value if the floating-point number is too large to fit in an integer. so like, only if you're brave.
• -mpointers-to-nested-functions, -mno-pointers-to-nested-functions — generate (or don't generate) code to load up the static chain register ("r11") when calling through a pointer on AIX and 64-bit Linux systems, where a function pointer points to a 3-word descriptor giving the function address, TOC value to be loaded in register "r2", and static chain value to be loaded in register "r11". The -mpointers-to-nested-functions is on by default. 🚨 You cannot call through pointers to nested functions, or pointers to functions compiled in other languages that use the static chain, if you use -mno-pointers-to-nested-functions. like, that's just the deal.
• -msave-toc-indirect, -mno-save-toc-indirect — generate (or don't generate) code to save the TOC value in the reserved stack location in the function prologue if the function calls through a pointer on AIX and 64-bit Linux systems. If the TOC value is not saved in the prologue, it's saved just before the call through the pointer. The -mno-save-toc-indirect option is the default. 🎀
• -mcompat-align-parm, -mno-compat-align-parm — generate (or don't generate) code to pass structure parameters with a maximum alignment of 64 bits, for compatibility with older versions of GCC.

Older versions of GCC (prior to 4.9.0) incorrectly did not align a structure parameter on a 128-bit boundary when that structure contained a member requiring 128-bit alignment. This is corrected in more recent versions of GCC. This option may be used to generate code that is compatible with functions compiled with older versions of GCC. 🕰️ retro vibes only.

The -mno-compat-align-parm option is the default.

• -mstack-protector-guard=guard, -mstack-protector-guard-reg=reg, -mstack-protector-guard-offset=offset — generate stack protection code using canary at guard. Supported locations are global for global canary, or tls for per-thread canary in the TLS block (the default with GNU libc version 2.4 or later). 🐤

With the latter choice, the options -mstack-protector-guard-reg=reg and -mstack-protector-guard-offset=offset furthermore specify which register to use as base register for reading the canary, and from what offset from that base register. The default for those is as specified in the relevant ABI.

• -mpcrel, -mno-pcrel — generate (or don't generate) pc-relative addressing. The -mpcrel option requires that the medium code model (-mcmodel=medium) and prefixed addressing (-mprefixed) options are enabled. chain reaction, fr fr.
• -mprefixed, -mno-prefixed — generate (or don't generate) addressing modes using prefixed load and store instructions. The -mprefixed option requires that the option -mcpu=power10 (or later) is enabled. 💅
• -mmma, -mno-mma — generate (or don't generate) the MMA instructions. The -mma option requires that the option -mcpu=power10 (or later) is enabled. (yes bestie, that's -mma with one m — that's literally what it says, don't come for me.)
• -mrop-protect, -mno-rop-protect — generate (or don't generate) ROP protection instructions when the target processor supports them. Currently this option disables the shrink-wrap optimization (-fshrink-wrap). 🛡️
• -mprivileged, -mno-privileged — generate (or don't generate) code that will run in privileged state. main character energy only. 👑
• -mblock-ops-unaligned-vsx, -mno-block-ops-unaligned-vsx — generate (or don't generate) unaligned vsx loads and stores for inline expansion of "memcpy" and "memmove".
• --param rs6000-vect-unroll-limit= — the vectorizer will check with target information to determine whether it would be beneficial to unroll the main vectorized loop, and by how much. This parameter sets the upper bound of how much the vectorizer will unroll the main loop. The default value is four. ✨

---

🌸 RX Options 🌸

These command-line options are defined for RX targets, bestie — niche but iconic:

• -m64bit-doubles, -m32bit-doubles — make the "double" data type be 64 bits (-m64bit-doubles) or 32 bits (-m32bit-doubles) in size. The default is -m32bit-doubles. 📝 Note: RX floating-point hardware only works on 32-bit values, which is why the default is -m32bit-doubles. it's giving hardware limitations.
• -fpu, -nofpu — enables (-fpu) or disables (-nofpu) the use of RX floating-point hardware. The default is enabled for the RX600 series and disabled for the RX200 series.

Floating-point instructions are only generated for 32-bit floating-point values, however, so the FPU hardware is not used for doubles if the -m64bit-doubles option is used.

⚠️ Note: If the -fpu option is enabled then -funsafe-math-optimizations is also enabled automatically. This is because the RX FPU instructions are themselves unsafe. 😭 yes, literally unsafe, that's the whole reason.

• -mcpu=name — selects the type of RX CPU to be targeted. Currently three types are supported: the generic RX600 and RX200 series hardware, and the specific RX610 CPU. The default is RX600.

The only difference between RX600 and RX610 is that the RX610 does not support the "MVTIPL" instruction. 📋

The RX200 series does not have a hardware floating-point unit and so -nofpu is enabled by default when this type is selected.

• -mbig-endian-data, -mlittle-endian-data — store data (but not code 🤨) in the big-endian format. The default is -mlittle-endian-data, i.e. to store data in the little-endian format.
• -msmall-data-limit=N — sets the biggest size, in bytes, that a global or static variable can be and still get yeeted into the small data area. Using that area can give you smaller + faster code, but bestie, listen: the area is limited, and it is 100% on the programmer to make sure it does not overflow. Also, the moment you use the small data area, one of the RX's registers (usually r13) gets reserved to point at it, so the compiler can't use it anymore. That could make your code slower and/or bigger if variables end up pushed onto the stack instead of chillin' in that register.

Heads up: common variables (the ones that were never initialized) and constants do not go into the small data area, because they get assigned to other sections in the output executable.

The default is zero, which turns the whole feature off. And no, this is not enabled by default at higher optimization levels (-O2 etc.) because reserving a register can be so detrimental. It's fully on the programmer to experiment and figure out whether this feature actually helps their program. For how the actual register holding the small data area pointer gets picked, go read the description of the -mpid option. 👀

• -msim, -mno-sim — use the simulator runtime. Default is the libgloss board-specific runtime instead.
• -mas100-syntax, -mno-as100-syntax — when you're generating assembler output, use a syntax that plays nice with Renesas's AS100 assembler. The GAS assembler can handle this syntax too, but it comes with some restrictions, so it is not what gets generated by default.
• -mmax-constant-size=N — sets the maximum size, in bytes, of a constant that's allowed to be used as an operand in an RX instruction. So like — the RX instruction set does allow constants up to 4 bytes long in instructions, but a longer value = a longer instruction. So sometimes it's genuinely beneficial to cap how big the constants in your instructions are allowed to be. Constants that are too big get placed into a constant pool instead and referenced via register indirection.

N can be between 0 and 4. A value of 0 (the default) or 4 means constants of any size are allowed.

• -mrelax — enable linker relaxation. Linker relaxation is when the linker tries to shrink your program by finding shorter versions of various instructions. Disabled by default.
• -mint-register=N — pick how many registers to reserve for fast interrupt handler functions. N can be between 0 and 4. 1 means register r13 is reserved for the exclusive use of fast interrupt handlers. 2 reserves r13 and r12. 3 reserves r13, r12 and r11. 4 reserves r13 all the way through r10. And 0, the default, reserves nothing.
• -msave-acc-in-interrupts — says that interrupt handler functions should preserve the accumulator register. This is only necessary if normal code might actually use the accumulator register, like when it does 64-bit multiplications. Default is to ignore the accumulator, because that makes interrupt handlers faster.
• -mpid, -mno-pid — turns on generation of position independent data. When it's enabled, any access to constant data happens via an offset from a base address held in a register. That lets the location of constant data be worked out at run time without needing the executable to be relocated, which is a big win for embedded applications with tight memory constraints. Data that can be modified is not affected by this option at all.

Note: using this feature reserves a register, usually r13, for that constant data base address. This can result in slower and/or larger code, especially in complicated functions.

Which register actually gets picked to hold the constant data base address depends on whether -msmall-data-limit and/or -mint-register are enabled on the command line. Starting at register r13 and going downwards, registers get allocated first to satisfy -mint-register, then -mpid, and finally -msmall-data-limit. So the small data area register totally could end up being r8 if you specify both -mint-register=4 and -mpid on the command line. (iykyk)

By default this feature is off. You can get the default back with the -mno-pid command-line option.

• -mno-warn-multiple-fast-interrupts, -mwarn-multiple-fast-interrupts — stops GCC from throwing a warning message when it finds more than one fast interrupt handler while compiling a file. The default is to warn for every extra fast interrupt handler it finds, because the RX only supports one such interrupt. 💀
• -mallow-string-insns, -mno-allow-string-insns — enables or disables the use of the string manipulation instructions SMOVF, SCMPU, SMOVB, SMOVU, SUNTIL, SWHILE, and also the RMPA instruction. These instructions may prefetch data, which is not safe if you're accessing an I/O register. (See section 12.2.7 of the RX62N Group User's Manual for more info.)

Default is to allow these instructions — but GCC cannot reliably detect every single circumstance where a string instruction might be used to access an I/O register, so their use cannot be disabled automatically. Instead it's on the programmer to use -mno-allow-string-insns if their program accesses I/O space. 🫡

When the instructions are enabled, GCC defines the C preprocessor symbol __RX_ALLOW_STRING_INSNS__; otherwise it defines the symbol __RX_DISALLOW_STRING_INSNS__.

• -mjsr, -mno-jsr — use only (or not only) JSR instructions to access functions. You'd reach for this option when code size exceeds the range of BSR instructions. Note that -mno-jsr does not mean "don't use JSR" — it means any type of branch may be used.

Note: the generic GCC command-line option -ffixed-reg has special significance to the RX port when it's used with the interrupt function attribute. That attribute marks a function as one intended to process fast interrupts. GCC makes sure it only uses the registers r10, r11, r12 and/or r13, and only as long as the normal use of the corresponding registers has been restricted via the -ffixed-reg or -mint-register command-line options.

---

🌈 S/390 and zSeries Options 🎀

These are the -m options defined for the S/390 and zSeries architecture.

• -mhard-float, -msoft-float — use (do not use) the hardware floating-point instructions and registers for floating-point operations. When -msoft-float is specified, functions in libgcc.a are used to perform floating-point operations. When -mhard-float is specified, the compiler generates IEEE floating-point instructions. This is the default. ✨
• -mhard-dfp, -mno-hard-dfp — use (do not use) the hardware decimal-floating-point instructions for decimal-floating-point operations. When -mno-hard-dfp is specified, functions in libgcc.a are used to perform decimal-floating-point operations. When -mhard-dfp is specified, the compiler generates decimal-floating-point hardware instructions. This is the default for -march=z9-ec or higher.
• -mlong-double-64, -mlong-double-128 — these switches control the size of the long double type. A size of 64 bits makes the long double type equivalent to the double type. This is the default.
• -mbackchain, -mno-backchain — store (do not store) the address of the caller's frame as a backchain pointer into the callee's stack frame. A backchain may be needed so debugging tools that don't understand DWARF call frame information can still work. When -mno-packed-stack is in effect, the backchain pointer is stored at the bottom of the stack frame; when -mpacked-stack is in effect, the backchain goes into the topmost word of the 96/160 byte register save area.

Generally speaking, code compiled with -mbackchain is call-compatible with code compiled with -mno-backchain; however, using the backchain for debugging purposes usually requires that the whole binary be built with -mbackchain. Note that the combination of -mbackchain, -mpacked-stack and -mhard-float is not supported. And to build a linux kernel, use -msoft-float. 🌸

The default is to not maintain the backchain.

• -mpacked-stack, -mno-packed-stack — use (do not use) the packed stack layout. When -mno-packed-stack is specified, the compiler uses all fields of the 96/160 byte register save area only for their default purpose; unused fields still take up stack space. When -mpacked-stack is specified, register save slots are densely packed at the top of the register save area; unused space gets reused for other purposes, letting you use the available stack space more efficiently. However, when -mbackchain is also in effect, the topmost word of the save area is always used to store the backchain, and the return address register is always saved two words below the backchain.

As long as the stack frame backchain isn't used, code generated with -mpacked-stack is call-compatible with code generated with -mno-packed-stack. But note: some non-FSF releases of GCC 2.95 for S/390 or zSeries generated code that uses the stack frame backchain at run time, not just for debugging purposes. That kind of code is not call-compatible with code compiled with -mpacked-stack. Also, the combination of -mbackchain, -mpacked-stack and -mhard-float is not supported. To build a linux kernel, use -msoft-float.

The default is to not use the packed stack layout.

• -msmall-exec, -mno-small-exec — generate (or do not generate) code using the bras instruction to do subroutine calls. This only works reliably if the total executable size does not exceed 64k. The default is to use the basr instruction instead, which does not have this limitation. 👑
• -m64, -m31 — okay bestie, so when you say -m31 you're telling GCC "make me code that plays nice with the GNU/Linux for S/390 ABI," and when you say -m64 it's giving GNU/Linux for zSeries ABI. This is literally how GCC gets to spit out 64-bit instructions, iconic. For s390 targets the default is -m31, but for s390x targets it defaults to -m64. 💅
• -mzarch, -mesa — with -mzarch you get code using the instructions available on z/Architecture ✨ and with -mesa you get code using the instructions available on ESA/390. Big heads up tho: -mesa is not possible with -m64, sorry bestie 😭. When generating code that's compliant with the GNU/Linux for S/390 ABI, the default is -mesa; when it's compliant with the GNU/Linux for zSeries ABI, the default is -mzarch. Slay.
• -mhtm, -mno-htm — -mhtm enables a whole set of builtins that make use of instructions available with the transactional execution facility that came in with the IBM zEnterprise EC12 machine generation S/390 System z Built-in Functions 🦋. And -mhtm is enabled by default when you're using -march=zEC12. Periodt.
• -mvx, -mno-vx — when you pass -mvx, you generate code using the instructions available with the vector extension facility introduced with the IBM z13 machine generation. ⚠️ This option changes the ABI for some vector type values when it comes to alignment and calling conventions. And in case vector type values are being used in an ABI-relevant context, a GAS .gnu_attribute command gets added to mark the resulting binary with the ABI used. -mvx is enabled by default when using -march=z13. 💖
• -mzvector, -mno-zvector — the -mzvector option enables vector language extensions and builtins using instructions available with the vector extension facility from the IBM z13 machine generation. This one adds support for vector to be used as a keyword to define vector type variables and arguments. Note: vector is only available when GNU extensions are enabled, and it won't be expanded when you're requesting strict standard compliance, e.g. with -std=c99. 🙅 On top of the GCC low-level builtins, -mzvector enables a set of builtins added for compatibility with AltiVec-style implementations like Power and Cell. To make use of these builtins you need to include the header file vecintrin.h. And bestie, -mzvector is disabled by default.
• -mmvcle, -mno-mvcle — generate (or don't generate 👀) code using the "mvcle" instruction to do block moves. When you specify -mno-mvcle, it uses a "mvc" loop instead. That's the default unless you're optimizing for size.
• -mdebug, -mno-debug — print (or don't print) extra debug info while compiling. Default is to not print debug info. Simple as that. ✨
• -march=cpu-type — generate code that runs on cpu-type, which is the name of a system representing a certain processor type. Possible values for cpu-type are z900/arch5, z990/arch6, z9-109, z9-ec/arch7, z10/arch8, z196/arch9, zEC12, z13/arch11, z14/arch12, z15/arch13, z16/arch14, z17/arch15, and native.

The default is -march=z900.

Specifying native as the cpu type can be used to select the best architecture option for the host processor. But ngl, -march=native has no effect if GCC doesn't recognize the processor. 💀

• -mtune=cpu-type — tune everything applicable about the generated code to cpu-type, except for the ABI and the set of available instructions. The list of cpu-type values is the same as for -march. And the default? It's the value used for -march. Matching energy. 🫶
• -mtpf-trace, -mno-tpf-trace — generate code that adds (or doesn't add) TPF OS specific branches to trace routines in the operating system. This option is off by default, even when you're compiling for the TPF OS. 🙈
• -mtpf-trace-skip, -mno-tpf-trace-skip — generate code that changes (or doesn't change) the default branch targets enabled by -mtpf-trace so they point to specialized trace routines that give you the ability to selectively skip function trace entries for the TPF OS. This option is off by default, even when compiling for the TPF OS and specifying -mtpf-trace. Yeah, it's like that.
• -mfused-madd, -mno-fused-madd — generate code that uses (or doesn't use) the floating-point multiply and accumulate instructions. These get generated by default if hardware floating point is used. 🍓
• -mwarn-framesize=framesize — emit a warning if the current function goes over the given frame size. Since this is a compile-time check, it doesn't have to be a real problem when the program actually runs — it's just vibes-based detection for functions that most probably cause a stack overflow. Super useful in an environment with limited stack size, e.g. the linux kernel. ⭐
• -mwarn-dynamicstack — emit a warning if the function calls "alloca" or uses dynamically-sized arrays. Ngl this is generally a bad idea with a limited stack size. 💀
• -mstack-guard=stack-guard, -mstack-size=stack-size — if these options are provided, the S/390 back end emits additional instructions in the function prologue that trigger a trap if the stack size is stack-guard bytes above the stack-size (remember, the stack on S/390 grows downward, lowercase bestie 📉). If the stack-guard option is omitted, the smallest power of 2 larger than the frame size of the compiled function is chosen. These options are intended to help debug stack overflow problems. The additionally emitted code causes only a little overhead, so it can also be used in production-like systems without greater performance degradation. The given values have to be exact powers of 2, and stack-size has to be greater than stack-guard without exceeding 64k. And in order to be efficient, the extra code makes the assumption that the stack starts at an address aligned to the value given by stack-size. Heads up: the stack-guard option can only be used in conjunction with stack-size. 👑
• -mhotpatch=pre-halfwords,post-halfwords — if the hotpatch option is enabled, a "hot-patching" function prologue is generated for all functions in the compilation unit. The funtion label is prepended with the given number of two-byte NOP instructions (pre-halfwords, maximum 1000000). After the label, 2 * post-halfwords bytes are appended, using the largest NOP like instructions the architecture allows (maximum 1000000).

If both arguments are zero, hotpatching is disabled.

This option can be overridden for individual functions with the "hotpatch" attribute. 💅

---

🌸 SH Options 🌸

These -m options are defined for the SH implementations:

• -m1 — Generate code for the SH1. ✨
• -m2 — Generate code for the SH2. ✨
• -m2e — Generate code for the SH2e. 🦋
• -m2a-nofpu — Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way that the floating-point unit is not used.
• -m2a-single-only — Generate code for the SH2a-FPU, in such a way that no double-precision floating-point operations are used.
• -m2a-single — Generate code for the SH2a-FPU assuming the floating-point unit is in single-precision mode by default.
• -m2a — Generate code for the SH2a-FPU assuming the floating-point unit is in double-precision mode by default.
• -m3 — Generate code for the SH3. 💜
• -m3e — Generate code for the SH3e. 💜
• -m4-nofpu — Generate code for the SH4 without a floating-point unit.
• -m4-single-only — Generate code for the SH4 with a floating-point unit that only supports single-precision arithmetic.
• -m4-single — Generate code for the SH4 assuming the floating-point unit is in single-precision mode by default.
• -m4 — Generate code for the SH4. 🔥
• -m4-100 — Generate code for SH4-100.
• -m4-100-nofpu — Generate code for SH4-100 in such a way that the floating-point unit is not used.
• -m4-100-single — Generate code for SH4-100 assuming the floating-point unit is in single-precision mode by default.
• -m4-100-single-only — Generate code for SH4-100 in such a way that no double-precision floating-point operations are used.
• -m4-200 — Generate code for SH4-200.
• -m4-200-nofpu — Generate code for SH4-200 without in such a way that the floating-point unit is not used.
• -m4-200-single — Generate code for SH4-200 assuming the floating-point unit is in single-precision mode by default.
• -m4-200-single-only — Generate code for SH4-200 in such a way that no double-precision floating-point operations are used.
• -m4-300 — Generate code for SH4-300.
• -m4-300-nofpu — Generate code for SH4-300 without in such a way that the floating-point unit is not used.
• -m4-300-single — Generate code for SH4-300 in such a way that no double-precision floating-point operations are used.
• -m4-300-single-only — Generate code for SH4-300 in such a way that no double-precision floating-point operations are used.
• -m4-340 — Generate code for SH4-340 (no MMU, no FPU).
• -m4-500 — Generate code for SH4-500 (no FPU). Passes -isa=sh4-nofpu to the assembler.
• -m4a-nofpu — Generate code for the SH4al-dsp, or for a SH4a in such a way that the floating-point unit is not used.
• -m4a-single-only — Generate code for the SH4a, in such a way that no double-precision floating-point operations are used.
• -m4a-single — Generate code for the SH4a assuming the floating-point unit is in single-precision mode by default.
• -m4a — Generate code for the SH4a. 👑
• -m4al — Same as -m4a-nofpu, except that it implicitly passes -dsp to the assembler. GCC doesn't generate any DSP instructions at the moment. 🙈
• -mb — Compile code for the processor in big-endian mode.
• -ml — Compile code for the processor in little-endian mode.
• -mdalign — Align doubles at 64-bit boundaries. ⚠️ Heads up, this changes the calling conventions, and thus some functions from the standard C library do not work unless you recompile it first with -mdalign.
• -mrelax — Shorten some address references at link time, when possible; uses the linker option -relax.
• -mbigtable — Use 32-bit offsets in "switch" tables. The default is to use 16-bit offsets.
• -mbitops — Enable the use of bit manipulation instructions on SH2A. ✨
• -mfmovd — Enable the use of the instruction "fmovd". Check -mdalign for alignment constraints.
• -mrenesas — Comply with the calling conventions defined by Renesas. 💖
• -mno-renesas — okay so this one says: comply with the calling conventions that GCC had before the Renesas conventions showed up. It's the default for every target in the SH toolchain. periodt.
• -mnomacsave — mark the "MAC" register as call-clobbered, even if -mrenesas is given. no cap.
• -mieee / -mno-ieee — these control IEEE compliance for floating-point comparisons, which changes how we handle cases where a comparison result is unordered. By default -mieee is implicitly enabled. If -ffinite-math-only is enabled, -mno-ieee is implicitly set, and that gives faster floating-point greater-equal and less-equal comparisons. You can override those implicit settings by specifying either -mieee or -mno-ieee.
• -minline-ic_invalidate — inline the code that invalidates instruction cache entries after setting up nested function trampolines. This option does nothing if -musermode is in effect and the selected code generation option (e.g. -m4) doesn't allow the "icbi" instruction. If the selected code generation option doesn't allow "icbi", and -musermode is not in effect, the inlined code directly manipulates the instruction cache address array with an associative write. That not only needs privileged mode at run time, it also fails if the cache line was mapped via the TLB and has become unmapped. yikes.
• -misize — dump instruction size and location in the assembly code. cute.
• -mpadstruct — this one is deprecated, bestie. It pads structures to a multiple of 4 bytes, which is incompatible with the SH ABI.
• -matomic-model=model — sets the atomic operations model plus extra parameters as a comma-separated list. For details on the atomic built-in functions, see __atomic Builtins. These are the supported models and parameters:
  • none — disable compiler-generated atomic sequences and emit library calls for atomic operations. This is the default if the target is not "sh*-*-linux*".
  • soft-gusa — generate GNU/Linux-compatible gUSA software atomic sequences for the atomic built-in functions. Those generated sequences need extra support from the system's interrupt/exception handling code, and they're only suitable for SH3* and SH4* single-core systems. Enabled by default when the target is "sh*-*-linux*" and SH3* or SH4*. When the target is SH4A, this option also partially uses the hardware atomic instructions "movli.l" and "movco.l" to make more efficient code, unless strict is specified.
  • soft-tcb — generate software atomic sequences that use a variable in the thread control block. It's a variation of the gUSA sequences and can also be used on SH1* and SH2* targets. The generated sequences need extra support from the system's interrupt/exception handling code and are only suitable for single-core systems. When you use this model, the gbr-offset= parameter has to be specified as well.
  • soft-imask — generate software atomic sequences that temporarily disable interrupts by setting "SR.IMASK = 1111". This model works only when the program runs in privileged mode and is only suitable for single-core systems. Extra support from the system's interrupt/exception handling code is not required. Enabled by default when the target is "sh*-*-linux*" and SH1* or SH2*.
  • hard-llcs — generate hardware atomic sequences using only the "movli.l" and "movco.l" instructions. Only available on SH4A and suitable for multi-core systems. Since those hardware instructions support only 32 bit atomic variables, access to 8 or 16 bit variables is emulated with 32 bit accesses. Code compiled with this option is also compatible with other software atomic model interrupt/exception handling systems if executed on an SH4A system. Extra support from the system's interrupt/exception handling code is not required for this model.
  • gbr-offset= — this parameter specifies the offset in bytes of the variable in the thread control block structure that the generated atomic sequences should use when the soft-tcb model has been selected. For other models this parameter is ignored. The value you specify must be an integer multiple of four and in the range 0-1020.
  • strict — this parameter prevents mixed usage of multiple atomic models, even if they're compatible, and makes the compiler generate atomic sequences of the specified model only.
• -mtas — generate the "tas.b" opcode for "__atomic_test_and_set". Heads up: depending on the exact hardware and software configuration, this can degrade overall performance because of the operand cache line flushes that the "tas.b" instruction implies. On multi-core SH4A processors, the "tas.b" instruction must be used with caution since it can result in data corruption for certain cache configurations. 💀
• -mprefergot — when generating position-independent code, emit function calls using the Global Offset Table instead of the Procedure Linkage Table.
• -musermode / -mno-usermode — don't allow (or allow) the compiler to generate privileged mode code. Specifying -musermode also implies -mno-inline-ic_invalidate if the inlined code wouldn't work in user mode. -musermode is the default when the target is "sh*-*-linux*". If the target is SH1* or SH2*, -musermode has no effect, because there is no user mode.
• -multcost=number — set the cost to assume for a multiply insn.
• -mdiv=strategy — set the division strategy used for integer division operations. strategy can be one of:
  • call-div1 — call a library function that uses the single-step division instruction "div1" to do the operation. Division by zero calculates an unspecified result and does not trap. This is the default except for SH4, SH2A and SHcompact.
  • call-fp — call a library function that does the operation in double precision floating point. Division by zero causes a floating-point exception. This is the default for SHcompact with FPU. If you specify this for targets that don't have a double precision FPU, it defaults to "call-div1".
  • call-table — call a library function that uses a lookup table for small divisors and the "div1" instruction with case distinction for larger divisors. Division by zero calculates an unspecified result and does not trap. This is the default for SH4. If you specify this for targets that don't have dynamic shift instructions, it defaults to "call-div1".

When a division strategy has not been specified, the default strategy is selected based on the current target. For SH2A, the default strategy is to use the "divs" and "divu" instructions instead of library function calls.

• -maccumulate-outgoing-args — reserve space once for outgoing arguments in the function prologue instead of around each call. Generally good for performance and size. Also needed for unwinding so the stack frame doesn't change around conditional code.
• -mdivsi3_libfunc=name — set the name of the library function used for 32-bit signed division to name. This only affects the name used in the call division strategies, and the compiler still expects the same sets of input/output/clobbered registers as if this option weren't present.
• -mfixed-range=register-range — generate code that treats the given register range as fixed registers. A fixed register is one the register allocator cannot use. This is useful when compiling kernel code. A register range is specified as two registers separated by a dash. Multiple register ranges can be specified separated by a comma.
• -mbranch-cost=num — assume num is the cost for a branch instruction. Higher numbers make the compiler try to generate more branch-free code if possible. If you don't specify it, the value is selected depending on the processor type being compiled for.
• -mzdcbranch / -mno-zdcbranch — assume (or don't assume) that zero displacement conditional branch instructions "bt" and "bf" are fast. If -mzdcbranch is specified, the compiler prefers zero displacement branch code sequences. Enabled by default when generating code for SH4 and SH4A. It can be explicitly disabled by specifying -mno-zdcbranch.
• -mcbranch-force-delay-slot — force the usage of delay slots for conditional branches, which stuffs the delay slot with a "nop" if a suitable instruction can't be found. By default this option is disabled. It can be enabled to work around hardware bugs like the ones found in the original SH7055.
• -mfused-madd / -mno-fused-madd — generate code that uses (or does not use) the floating-point multiply and accumulate instructions. These instructions are generated by default if hardware floating point is used. The machine-dependent -mfused-madd option is now mapped to the machine-independent -ffp-contract=fast option, and -mno-fused-madd is mapped to -ffp-contract=off.
• -mfsca / -mno-fsca — allow or disallow the compiler to emit the "fsca" instruction for sine and cosine approximations. The -mfsca option must be used in combination with -funsafe-math-optimizations. It is enabled by default when generating code for SH4A. Using -mno-fsca disables sine and cosine approximations even if -funsafe-math-optimizations is in effect.
• -mfsrra / -mno-fsrra — allow or disallow the compiler to emit the "fsrra" instruction for reciprocal square root approximations. The -mfsrra option must be used in combination with -funsafe-math-optimizations and -ffinite-math-only. It is enabled by default when generating code for SH4A. Using -mno-fsrra disables reciprocal square root approximations even if -funsafe-math-optimizations and -ffinite-math-only are in effect.
• -mpretend-cmove — prefer zero-displacement conditional branches for conditional move instruction patterns. This can result in faster code on the SH4 processor.
• -mfdpic — generate code using the FDPIC ABI.

🌞 Solaris 2 Options 🌞

These -m options are supported on Solaris 2, bestie — tiny family, big opinions:

• -mclear-hwcap — tells the compiler to yeet the hardware capabilities that the Solaris assembler generated. You only need this when your object files use ISA extensions that the machine you're currently on does not support, but they check at runtime whether or not to actually use them. So like, it's fine, it's checking, chill 💅
• -mimpure-text — used in addition to -shared, tells the compiler to not pass -z text to the linker when linking a shared object. With this option you can link position-dependent code into a shared object.

-mimpure-text also suppresses that "relocations remain against allocatable but non-writable sections" linker error message. However (and this is the important bit, no cap 😭): the necessary relocations trigger copy-on-write, and the shared object is not actually shared across processes. So instead of using -mimpure-text, you should compile all source code with -fpic or -fPIC. Periodt.

These switches are supported in addition to the above on Solaris 2:

• -pthreads — this is literally just a synonym for -pthread. Same vibe, different name 🫶

---

✨ SPARC Options ✨

These -m options are supported on the SPARC:

• -mno-app-regs, -mapp-regs — specify -mapp-regs to generate output using the global registers 2 through 4, which the SPARC SVR4 ABI reserves for applications. Like global register 1, each global register 2 through 4 is then treated as an allocable register that gets clobbered by function calls. This is the default. 🔥

To be fully SVR4 ABI-compliant at the cost of some performance loss, specify -mno-app-regs. You should compile libraries and system software with this option. That's the "trust me bestie" recommendation.

• -mflat, -mno-flat — with -mflat, the compiler does not generate save/restore instructions and instead uses a "flat" or single register window model. This model is compatible with the regular register window model. The local registers and the input registers (0--5) are still treated as "call-saved" registers and are saved on the stack as needed.

With -mno-flat (the default), the compiler generates save/restore instructions (except for leaf functions). This is the normal operating mode. The main character mode, if you will 👑

• -mfpu, -mhard-float — generate output containing floating-point instructions. This is the default.
• -mno-fpu, -msoft-float — generate output containing library calls for floating point.

Warning: the requisite libraries are not available for all SPARC targets. Normally the facilities of the machine's usual C compiler are used, but this cannot be done directly in cross-compilation. You must make your own arrangements to provide suitable library functions for cross-compilation. The embedded targets sparc-*-aout and sparclite-*-* do provide software floating-point support.

-msoft-float changes the calling convention in the output file, so it's only useful if you compile all of a program with this option. In particular, you need to compile libgcc.a — the library that comes with GCC — with -msoft-float in order for this to work. All of it. The whole thing. 😤

• -mhard-quad-float — generate output containing quad-word (long double) floating-point instructions.
• -msoft-quad-float — generate output containing library calls for quad-word (long double) floating-point instructions. The functions called are those specified in the SPARC ABI. This is the default.

As of this writing, there are no SPARC implementations that have hardware support for the quad-word floating-point instructions. They all invoke a trap handler for one of these instructions, and then the trap handler emulates the effect of the instruction. Because of the trap handler overhead, this is much slower than calling the ABI library routines. Thus the -msoft-quad-float option is the default. It's giving "slow but honest" 🐢

• -mno-unaligned-doubles, -munaligned-doubles — assume that doubles have 8-byte alignment. This is the default.

With -munaligned-doubles, GCC assumes that doubles have 8-byte alignment only if they are contained in another type, or if they have an absolute address. Otherwise, it assumes they have 4-byte alignment. Specifying this option avoids some rare compatibility problems with code generated by other compilers. It is not the default because it results in a performance loss, especially for floating-point code. So, vibe trade-off 📉

• -muser-mode, -mno-user-mode — do not generate code that can only run in supervisor mode. This is relevant only for the "casa" instruction emitted for the LEON3 processor. This is the default.
• -mfaster-structs, -mno-faster-structs — with -mfaster-structs, the compiler assumes that structures should have 8-byte alignment. This enables the use of pairs of "ldd" and "std" instructions for copies in structure assignment, in place of twice as many "ld" and "st" pairs. However, the use of this changed alignment directly violates the SPARC ABI. So it's intended only for use on targets where the developer acknowledges that their resulting code is not directly in line with the rules of the ABI. You're on your own, and that's the deal 🙈
• -mstd-struct-return, -mno-std-struct-return — with -mstd-struct-return, the compiler generates checking code in functions returning structures or unions to detect size mismatches between the two sides of function calls, as per the 32-bit ABI.

The default is -mno-std-struct-return. This option has no effect in 64-bit mode.

• -mcpu=cpu_type — set the instruction set, register set, and instruction scheduling parameters for machine type cpu_type. Supported values for cpu_type are v7, cypress, v8, supersparc, hypersparc, leon, leon3, leon3v7, leon5, sparclite, f930, f934, sparclite86x, sparclet, tsc701, v9, ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4, niagara7 and m8.

Native Solaris and GNU/Linux toolchains also support the value native, which selects the best architecture option for the host processor. -mcpu=native has no effect if GCC does not recognize the processor.

Default instruction scheduling parameters are used for values that select an architecture and not an implementation. These are v7, v8, sparclite, sparclet, v9.

Here's the whole registry, bestie — each supported architecture and its supported implementations:

• v7 — cypress, leon3v7
• v8 — supersparc, hypersparc, leon, leon3, leon5
• sparclite — f930, f934, sparclite86x
• sparclet — tsc701
• v9 — ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4, niagara7, m8

By default (unless configured otherwise), GCC generates code for the V7 variant of the SPARC architecture. With -mcpu=cypress, the compiler additionally optimizes it for the Cypress CY7C602 chip, as used in the SPARCStation/SPARCServer 3xx series. This is also appropriate for the older SPARCStation 1, 2, IPX etc.

With -mcpu=v8, GCC generates code for the V8 variant of the SPARC architecture. The only difference from V7 code is that the compiler emits the integer multiply and integer divide instructions which exist in SPARC-V8 but not in SPARC-V7. With -mcpu=supersparc, the compiler additionally optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and 2000 series.

With -mcpu=sparclite, GCC generates code for the SPARClite variant of the SPARC architecture. This adds the integer multiply, integer divide step and scan ("ffs") instructions which exist in SPARClite but not in SPARC-V7. With -mcpu=f930, the compiler additionally optimizes it for the Fujitsu MB86930 chip, which is the original SPARClite, with no FPU. With -mcpu=f934, the compiler additionally optimizes it for the Fujitsu MB86934 chip, which is the more recent SPARClite with FPU.

With -mcpu=sparclet, GCC generates code for the SPARClet variant of the SPARC architecture. This adds the integer multiply, multiply/accumulate, integer divide step and scan ("ffs") instructions which exist in SPARClet but not in SPARC-V7. With -mcpu=tsc701, the compiler additionally optimizes it for the TEMIC SPARClet chip.

With -mcpu=v9, GCC generates code for the V9 variant of the SPARC architecture. This adds 64-bit integer and floating-point move instructions, 3 additional floating-point condition code registers and conditional move instructions. With -mcpu=ultrasparc, the compiler additionally optimizes it for the Sun UltraSPARC I/II/IIi chips. With -mcpu=ultrasparc3, the compiler additionally optimizes it for the Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With -mcpu=niagara, the compiler additionally optimizes it for Sun UltraSPARC T1 chips. With -mcpu=niagara2, the compiler additionally optimizes it for Sun UltraSPARC T2 chips. With -mcpu=niagara3, the compiler additionally optimizes it for Sun UltraSPARC T3 chips. With -mcpu=niagara4, the compiler additionally optimizes it for Sun UltraSPARC T4 chips. With -mcpu=niagara7, the compiler additionally optimizes it for Oracle SPARC M7 chips. With -mcpu=m8, the compiler additionally optimizes it for Oracle M8 chips. Look at that glow-up timeline, honestly ✨

• -mtune=cpu_type — set the instruction scheduling parameters for machine type cpu_type, but do not set the instruction set or register set that the option -mcpu=cpu_type does.

The same values for -mcpu=cpu_type can be used for -mtune=cpu_type, but the only useful values are those that select a particular CPU implementation. Those are cypress, supersparc, hypersparc, leon, leon3, leon3v7, leon5, f930, f934, sparclite86x, tsc701, ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4, niagara7 and m8. With native Solaris and GNU/Linux toolchains, native can also be used.

• -mv8plus, -mno-v8plus — with -mv8plus, GCC generates code for the SPARC-V8+ ABI. The difference from the V8 ABI is that the global and out registers are considered 64 bits wide. This is enabled by default on Solaris in 32-bit mode for all SPARC-V9 processors.
• -mvis, -mno-vis — with -mvis, GCC generates code that takes advantage of the UltraSPARC Visual Instruction Set extensions. The default is -mno-vis.
• -mvis2, -mno-vis2 — with -mvis2, GCC generates code that takes advantage of version 2.0 of the UltraSPARC Visual Instruction Set extensions. The default is -mvis2 when targeting a cpu that supports such instructions, such as UltraSPARC-III and later. Setting -mvis2 also sets -mvis. Combo deal, no extra charge 💖
• -mvis3, -mno-vis3 — turn this on and GCC drops code that flexes version 3.0 of the UltraSPARC Visual Instruction Set extensions. The default is -mvis3 whenever you're targeting a cpu that's got those instructions in it, like niagara-3 and anything later. Turning on -mvis3 also turns on -mvis2 and -mvis, bestie, they come as a lil bundle.
• -mvis3b, -mno-vis3b — same version 3.0 UltraSPARC VIS energy, but plus the extra VIS instructions that showed up in the Oracle SPARC Architecture 2011. Default is -mvis3b when the target cpu supports them, like niagara-7 and later. Setting -mvis3b also sets -mvis3, -mvis2 and -mvis. ✨
• -mvis4, -mno-vis4 — version 4.0 of the UltraSPARC Visual Instruction Set extensions. Default is -mvis4 when targeting a cpu that supports such instructions, like niagara-7 and later. Setting -mvis4 also sets -mvis3b, -mvis3, -mvis2 and -mvis — it's giving ✨cascade✨.
• -mvis4b, -mno-vis4b — version 4.0 UltraSPARC VIS plus the additional VIS instructions introduced in the Oracle SPARC Architecture 2017. Default is -mvis4b when the cpu supports them, such as m8 and later. Setting -mvis4b also sets -mvis4, -mvis3b, -mvis3, -mvis2 and -mvis. The whole chain, no cap. 💅
• -mcbcond, -mno-cbcond — with -mcbcond, GCC generates code that takes advantage of the UltraSPARC Compare-and-Branch-on-Condition instructions. Default is -mcbcond when you're targeting a CPU that supports such instructions, like Niagara-4 and later.
• -mfmaf, -mno-fmaf — with -mfmaf, GCC uses the UltraSPARC Fused Multiply-Add Floating-point instructions. Default is -mfmaf when the CPU supports them, such as Niagara-3 and later. 🫶
• -mfsmuld, -mno-fsmuld — with -mfsmuld, GCC generates code that takes advantage of the Floating-point Multiply Single to Double (FsMULd) instruction. Default is -mfsmuld when targeting a CPU supporting the architecture versions V8 or V9 with FPU — except -mcpu=leon. (leon said no thanks 😭)
• -mpopc, -mno-popc — with -mpopc, GCC generates code that uses the UltraSPARC Population Count instruction. Default is -mpopc when the CPU supports such an instruction, like Niagara-2 and later.
• -msubxc, -mno-subxc — with -msubxc, GCC takes advantage of the UltraSPARC Subtract-Extended-with-Carry instruction. Default is -msubxc when targeting a CPU that supports it, such as Niagara-7 and later.
• -mfix-at697f — enables the documented workaround for the single erratum of the Atmel AT697F processor (which corresponds to erratum #13 of the AT697E processor).
• -mfix-ut699 — enables the documented workarounds for the floating-point errata and the data cache nullify errata of the UT699 processor. two for one, iconic.
• -mfix-ut700 — enables the documented workaround for the back-to-back store errata of the UT699E/UT700 processor.
• -mfix-gr712rc — enables the documented workaround for the back-to-back store errata of the GR712RC processor.

---

okay so these -m options are supported in addition to the above on SPARC-V9 processors in 64-bit environments:

• -m32, -m64 — generate code for a 32-bit or 64-bit environment. The 32-bit environment sets int, long and pointer to 32 bits. The 64-bit environment sets int to 32 bits and long and pointer to 64 bits. 🧚
• -mcmodel=which — set the code model to one of:
  • medlow — the Medium/Low code model: 64-bit addresses, programs must be linked in the low 32 bits of memory. Programs can be statically or dynamically linked.
  • medmid — the Medium/Middle code model: 64-bit addresses, programs must be linked in the low 44 bits of memory, the text and data segments must be less than 2GB in size and the data segment must be located within 2GB of the text segment.
  • medany — the Medium/Anywhere code model: 64-bit addresses, programs may be linked anywhere in memory, the text and data segments must be less than 2GB in size and the data segment must be located within 2GB of the text segment.
  • embmedany — the Medium/Anywhere code model for embedded systems: 64-bit addresses, the text and data segments must be less than 2GB in size, both starting anywhere in memory (determined at link time). The global register %g4 points to the base of the data segment. Programs are statically linked and PIC is not supported. (no PIC?? noted 💀)
• -mmemory-model=mem-model — set the memory model in force on the processor to one of:
  • default — the default memory model for the processor and operating system.
  • rmo — Relaxed Memory Order
  • pso — Partial Store Order
  • tso — Total Store Order
  • sc — Sequential Consistency

These memory models are formally defined in Appendix D of the SPARC-V9 architecture manual, as set in the processor's PSTATE.MM field. 📖

• -mstack-bias, -mno-stack-bias — with -mstack-bias, GCC assumes the stack pointer, and the frame pointer if present, are offset by -2047, which must be added back when making stack frame references. This is the default in 64-bit mode. Otherwise, assume no such offset is present.

---

🌈 Options for System V 🌈

these extra options are available on System V Release 4 for compatibility with other compilers on those systems, bestie:

• -G — create a shared object. It is recommended that -symbolic or -shared be used instead. (lowkey deprecated energy, use the other ones 🥲)
• -Qy — identify the versions of each tool used by the compiler, in a .ident assembler directive in the output.
• -Qn — refrain from adding .ident directives to the output file (this is the default).
• -YP,dirs — search the directories dirs, and no others, for libraries specified with -l.
• -Ym,dir — look in the directory dir to find the M4 preprocessor. The assembler uses this option.

---

💜 V850 Options 💜

these -m options are defined for V850 implementations:

• -mlong-calls, -mno-long-calls — treat all calls as being far away (near). If calls are assumed to be far away, the compiler always loads the function's address into a register, and calls indirect through the pointer.
• -mno-ep, -mep — do not optimize (do optimize) basic blocks that use the same index pointer 4 or more times to copy pointer into the ep register, and use the shorter sld and sst instructions. The -mep option is on by default if you optimize.
• -mno-prolog-function, -mprolog-function — do not use (do use) external functions to save and restore registers at the prologue and epilogue of a function. The external functions are slower, but use less code space if more than one function saves the same number of registers. The -mprolog-function option is on by default if you optimize.
• -mspace — try to make the code as small as possible. At present, this just turns on the -mep and -mprolog-function options. smol code supremacy 🎀
• -mtda=n — put static or global variables whose size is n bytes or less into the tiny data area that register ep points to. The tiny data area can hold up to 256 bytes in total (128 bytes for byte references).
• -msda=n — put static or global variables whose size is n bytes or less into the small data area that register gp points to. The small data area can hold up to 64 kilobytes.
• -mzda=n — put static or global variables whose size is n bytes or less into the first 32 kilobytes of memory.
• -mv850 — specify that the target processor is the V850.
• -mv850e3v5 — specify that the target processor is the V850E3V5. The preprocessor constant __v850e3v5__ is defined if this option is used.
• -mv850e2v4 — specify that the target processor is the V850E3V5. This is an alias for the -mv850e3v5 option. (yes, the name says e2v4 and it means e3v5, we don't ask questions 🙈)
• -mv850e2v3 — specify that the target processor is the V850E2V3. The preprocessor constant __v850e2v3__ is defined if this option is used.
• -mv850e2 — specify that the target processor is the V850E2. The preprocessor constant __v850e2__ is defined if this option is used.
• -mv850e1 — specify that the target processor is the V850E1. The preprocessor constants __v850e1__ and __v850e__ are defined if this option is used.
• -mv850es — specify that the target processor is the V850ES. This is an alias for the -mv850e1 option.
• -mv850e — specify that the target processor is the V850E. The preprocessor constant __v850e__ is defined if this option is used.

If neither -mv850 nor -mv850e nor -mv850e1 nor -mv850e2 nor -mv850e2v3 nor -mv850e3v5 are defined then a default target processor is chosen and the relevant __v850*__ preprocessor constant is defined.

The preprocessor constants __v850 and __v851__ are always defined, regardless of which processor variant is the target. every single time, no cap. 💖

• -mdisable-callt, -mno-disable-callt — this option suppresses generation of the CALLT instruction for the v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the v850 architecture.

This option is enabled by default when the RH850 ABI is in use (see -mrh850-abi), and disabled by default when the GCC ABI is in use. If CALLT instructions are being generated then the C preprocessor symbol __V850_CALLT__ is defined.

• -mrelax, -mno-relax — pass on (or do not pass on) the -mrelax command-line option to the assembler.
• -mlong-jumps, -mno-long-jumps — disable (or re-enable) the generation of PC-relative jump instructions.
• -msoft-float, -mhard-float — disable (or re-enable) the generation of hardware floating point instructions. This option is only significant when the target architecture is V850E2V3 or higher. If hardware floating point instructions are being generated then the C preprocessor symbol __FPU_OK__ is defined, otherwise the symbol __NO_FPU__ is defined.
• -mloop — enables the use of the e3v5 LOOP instruction. The use of this instruction is not enabled by default when the e3v5 architecture is selected, because its use is still experimental. (experimental?? proceed with caution bestie ⚠️)
• -mrh850-abi, -mghs — turns on support for the RH850 flavored version of the V850 ABI. this is the default bestie, no cap. When this ABI is live, here's the whole vibe:
  • Integer sized structures and unions get returned via a memory pointer, not a register.
  • Large structures and unions (more than 8 bytes in size) are passed by value.
  • Functions are aligned to 16-bit boundaries.
  • The -m8byte-align command-line option is supported. ✨
  • The -mdisable-callt command-line option is enabled by default. The -mno-disable-callt command-line option is not supported. 💀

When this version of the ABI is enabled the C preprocessor symbol "__V850_RH850_ABI__" is defined.

• -mgcc-abi — turns on support for the old GCC version of the V850 ABI. Rules for this one, periodt:
  • Integer sized structures and unions are returned in register "r10".
  • Large structures and unions (more than 8 bytes in size) are passed by reference.
  • Functions are aligned to 32-bit boundaries, unless optimizing for size.
  • The -m8byte-align command-line option is not supported.
  • The -mdisable-callt command-line option is supported but not enabled by default.

When this version of the ABI is enabled the C preprocessor symbol "__V850_GCC_ABI__" is defined.

• -m8byte-align, -mno-8byte-align — lets "double" and "long long" types be aligned on 8-byte boundaries. The default is to restrict the alignment of all objects to at most 4-bytes. When -m8byte-align is in effect the C preprocessor symbol "__V850_8BYTE_ALIGN__" is defined.
• -mbig-switch — generates code suitable for big switch tables. Use this option only if the assembler/linker complain about out of range branches within a switch table. 🙈
• -mapp-regs — this option causes r2 and r5 to be used in the code generated by the compiler. This setting is the default.
• -mno-app-regs — this option causes r2 and r5 to be treated as fixed registers.

---

🌈 VAX Options 🌈

These -m options are defined for the VAX:

• -munix — do not output certain jump instructions ("aobleq" and so on) that the Unix assembler for the VAX cannot handle across long ranges.
• -mgnu — do output those jump instructions, on the assumption that the GNU assembler is being used.
• -md, -md-float — use the D_floating data format for double-precision floating-point numbers instead of G_floating.
• -mg, -mg-float — use the G_floating data format for double-precision floating-point numbers instead of D_floating.
• -mlra, -mno-lra — enable Local Register Allocation. This is still experimental for the VAX, so by default the compiler uses standard reload.

---

🔭 Visium Options 🔭

• -mdebug — a program which performs file I/O and is destined to run on an MCM target should be linked with this option. It causes the libraries libc.a and libdebug.a to be linked. The program should be run on the target under the control of the GDB remote debugging stub.
• -msim — a program which performs file I/O and is destined to run on the simulator should be linked with option. This causes libraries libc.a and libsim.a to be linked.
• -mfpu, -mhard-float — generate code containing floating-point instructions. This is the default.
• -mno-fpu, -msoft-float — generate code containing library calls for floating-point.

-msoft-float changes the calling convention in the output file; therefore, it is only useful if you compile all of a program with this option. In particular, you need to compile libgcc.a, the library that comes with GCC, with -msoft-float in order for this to work. ⚠️

• *-mcpu=cpu_type* — set the instruction set, register set, and instruction scheduling parameters for machine type cpu_type. Supported values for cpu_type are mcm, gr5 and gr6.

mcm is a synonym of gr5 present for backward compatibility.

By default (unless configured otherwise), GCC generates code for the GR5 variant of the Visium architecture.

With -mcpu=gr6, GCC generates code for the GR6 variant of the Visium architecture. The only difference from GR5 code is that the compiler will generate block move instructions.

• *-mtune=cpu_type* — set the instruction scheduling parameters for machine type cpu_type, but do not set the instruction set or register set that the option -mcpu=cpu_type would.
• -msv-mode — generate code for the supervisor mode, where there are no restrictions on the access to general registers. This is the default.
• -muser-mode — generate code for the user mode, where the access to some general registers is forbidden: on the GR5, registers r24 to r31 cannot be accessed in this mode; on the GR6, only registers r29 to r31 are affected.

---

💜 VMS Options 💜

These -m options are defined for the VMS implementations:

• -mvms-return-codes — return VMS condition codes from "main". The default is to return POSIX-style condition (e.g. error) codes.
• *-mdebug-main=prefix* — flag the first routine whose name starts with prefix as the main routine for the debugger.
• -mmalloc64 — default to 64-bit memory allocation routines.
• *-mpointer-size=size* — set the default size of pointers. Possible options for size are 32 or short for 32 bit pointers, 64 or long for 64 bit pointers, and no for supporting only 32 bit pointers. The later option disables "pragma pointer_size".

---

🌸 VxWorks Options 🌸

The options in this section are defined for all VxWorks targets. Options specific to the target hardware are listed with the other options for that target.

• -mrtp — GCC can generate code for both VxWorks kernels and real time processes (RTPs). This option switches from the former to the latter. It also defines the preprocessor macro "__RTP__".
• -msmp — select SMP runtimes for linking. Not available on architectures other than PowerPC, nor on VxWorks version 7 or later, in which the selection is part of the VxWorks build configuration and the library paths are the same for either choice. 🙈
• -non-static — link an RTP executable against shared libraries rather than static libraries. The options -static and -shared can also be used for RTPs; -static is the default.
• -Bstatic, -Bdynamic — these options are passed down to the linker. They are defined for compatibility with Diab.
• -Xbind-lazy — enable lazy binding of function calls. This option is equivalent to -Wl,-z,now and is defined for compatibility with Diab.
• -Xbind-now — disable lazy binding of function calls. This option is the default and is defined for compatibility with Diab.

---

💅 x86 Options 💅

These -m options are defined for the x86 family of computers. slay.

• *-march=cpu-type* — generate instructions for the machine type cpu-type. In contrast to -mtune=cpu-type, which merely tunes the generated code for the specified cpu-type, -march=cpu-type allows GCC to generate code that may not run at all on processors other than the one indicated. Specifying -march=cpu-type implies -mtune=cpu-type, except where noted otherwise. 😭

The choices for cpu-type are:

• native — this selects the CPU to generate code for at compilation time by determining the processor type of the compiling machine. Using -march=native enables all instruction subsets supported by the local machine (hence the result might not run on different machines). Using -mtune=native produces code optimized for the local machine under the constraints of the selected instruction set.
• x86-64 — a generic CPU with 64-bit extensions, MMX, SSE, SSE2, and FXSR instruction set support.
• x86-64-v2, x86-64-v3, x86-64-v4 — these choices for cpu-type select the corresponding micro-architecture level from the x86-64 psABI. On ABIs other than the x86-64 psABI they select the same CPU features as the x86-64 psABI documents for the particular micro-architecture level.

Since these cpu-type values do not have a corresponding -mtune setting, using -march with these values enables generic tuning. Specific tuning can be enabled using the -mtune=other-cpu-type option with an appropriate other-cpu-type value.

• i386 — original Intel i386 CPU.
• i486 — Intel i486 CPU. (No scheduling is implemented for this chip.)
• i586, pentium — Intel Pentium CPU with no MMX support.
• lakemont — Intel Lakemont MCU, based on Intel Pentium CPU.
• pentium-mmx — Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support.
• pentiumpro — Intel Pentium Pro CPU with no MMX support.
• i686 — when used with -march, the Pentium Pro instruction set is used, so the code runs on all i686 family chips. When used with -mtune, it has the same meaning as generic.
• pentium2 — Intel Pentium II CPU, based on Pentium Pro core with MMX and FXSR instruction set support.
• pentium3, pentium3m — Intel Pentium III CPU, based on Pentium Pro core with MMX, FXSR and SSE instruction set support.
• pentium-m — Intel Pentium M; low-power version of Intel Pentium III CPU with MMX, SSE, SSE2 and FXSR instruction set support. Used by Centrino notebooks. 💻
• pentium4, pentium4m — Intel Pentium 4 CPU with MMX, SSE, SSE2 and FXSR instruction set support.
• prescott — improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2, SSE3 and FXSR instruction set support.
• nocona — improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and FXSR instruction set support.
• core2 — Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, CX16, SAHF and FXSR instruction set support.
• nehalem, corei7 — Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, CX16, SAHF and FXSR instruction set support.

it's the CPU roster bestie 💅 every name below is a thing you can pass as the cpu-type for the architecture-tuning flags, and each one drags along its own lil instruction-set entourage. no skips, no merges, we're doing all of them 🌸

---

westmere 💜

Intel Westmere CPU, 64-bit extensions in the bag, and the crew is MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, CX16, SAHF, FXSR and PCLMUL instruction set support. iconic, classic, done. ✨

sandybridge, corei7-avx 🦋

Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, CX16, SAHF, FXSR, AVX, XSAVE and PCLMUL instruction set support. two names, one vibe, fr fr.

ivybridge, core-avx-i 🎀

Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, CX16, SAHF, FXSR, AVX, XSAVE, PCLMUL, FSGSBASE, RDRND and F16C instruction set support.

haswell, core-avx2 🔥

Intel Haswell CPU, 64-bit extensions, and now we're cooking: MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, CX16, SAHF, FXSR, AVX, XSAVE, PCLMUL, FSGSBASE, RDRND, F16C, AVX2, BMI, BMI2, LZCNT, FMA, MOVBE and HLE instruction set support. she ate.

broadwell 🌈

Intel Broadwell CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, CX16, SAHF, FXSR, AVX, XSAVE, PCLMUL, FSGSBASE, RDRND, F16C, AVX2, BMI, BMI2, LZCNT, FMA, MOVBE, HLE, RDSEED, ADCX and PREFETCHW instruction set support.

skylake ⭐

Intel Skylake CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, CX16, SAHF, FXSR, AVX, XSAVE, PCLMUL, FSGSBASE, RDRND, F16C, AVX2, BMI, BMI2, LZCNT, FMA, MOVBE, HLE, RDSEED, ADCX, PREFETCHW, AES, CLFLUSHOPT, XSAVEC, XSAVES and SGX instruction set support. the guest list is growing, lowkey iconic.

skylake-avx512 💖

Intel Skylake Server CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, CX16, SAHF, FXSR, AVX, XSAVE, PCLMUL, FSGSBASE, RDRND, F16C, AVX2, BMI, BMI2, LZCNT, FMA, MOVBE, HLE, RDSEED, ADCX, PREFETCHW, AES, CLFLUSHOPT, XSAVEC, XSAVES, SGX, AVX512F, CLWB, AVX512VL, AVX512BW, AVX512DQ and AVX512CD instruction set support. AVX512 has entered the chat 😭

cascadelake 💅

Intel Cascade Lake CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, CX16, SAHF, FXSR, AVX, XSAVE, PCLMUL, FSGSBASE, RDRND, F16C, AVX2, BMI, BMI2, LZCNT, FMA, MOVBE, HLE, RDSEED, ADCX, PREFETCHW, AES, CLFLUSHOPT, XSAVEC, XSAVES, SGX, AVX512F, CLWB, AVX512VL, AVX512BW, AVX512DQ, AVX512CD and AVX512VNNI instruction set support.

cannonlake 🧚

Intel Cannon Lake Server CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, CX16, SAHF, FXSR, AVX, XSAVE, PCLMUL, FSGSBASE, RDRND, F16C, AVX2, BMI, BMI2, LZCNT, FMA, MOVBE, HLE, RDSEED, ADCX, PREFETCHW, AES, CLFLUSHOPT, XSAVEC, XSAVES, SGX, AVX512F, AVX512VL, AVX512BW, AVX512DQ, AVX512CD, PKU, AVX512VBMI, AVX512IFMA and SHA instruction set support.

cooperlake 🍓

Intel Cooper Lake CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, CX16, SAHF, FXSR, AVX, XSAVE, PCLMUL, FSGSBASE, RDRND, F16C, AVX2, BMI, BMI2, LZCNT, FMA, MOVBE, HLE, RDSEED, ADCX, PREFETCHW, AES, CLFLUSHOPT, XSAVEC, XSAVES, SGX, AVX512F, CLWB, AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VNNI and AVX512BF16 instruction set support.

icelake-client ❄️

Intel Ice Lake Client CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, CX16, SAHF, FXSR, AVX, XSAVE, PCLMUL, FSGSBASE, RDRND, F16C, AVX2, BMI, BMI2, LZCNT, FMA, MOVBE, HLE, RDSEED, ADCX, PREFETCHW, AES, CLFLUSHOPT, XSAVEC, XSAVES, SGX, AVX512F, AVX512VL, AVX512BW, AVX512DQ, AVX512CD, PKU, AVX512VBMI, AVX512IFMA, SHA, AVX512VNNI, GFNI, VAES, AVX512VBMI2 , VPCLMULQDQ, AVX512BITALG, RDPID and AVX512VPOPCNTDQ instruction set support.

icelake-server 💙

Intel Ice Lake Server CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, CX16, SAHF, FXSR, AVX, XSAVE, PCLMUL, FSGSBASE, RDRND, F16C, AVX2, BMI, BMI2, LZCNT, FMA, MOVBE, HLE, RDSEED, ADCX, PREFETCHW, AES, CLFLUSHOPT, XSAVEC, XSAVES, SGX, AVX512F, AVX512VL, AVX512BW, AVX512DQ, AVX512CD, PKU, AVX512VBMI, AVX512IFMA, SHA, AVX512VNNI, GFNI, VAES, AVX512VBMI2 , VPCLMULQDQ, AVX512BITALG, RDPID, AVX512VPOPCNTDQ, PCONFIG, WBNOINVD and CLWB instruction set support.

tigerlake 🐯

Intel Tiger Lake CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, CX16, SAHF, FXSR, AVX, XSAVE, PCLMUL, FSGSBASE, RDRND, F16C, AVX2, BMI, BMI2, LZCNT, FMA, MOVBE, HLE, RDSEED, ADCX, PREFETCHW, AES, CLFLUSHOPT, XSAVEC, XSAVES, SGX, AVX512F, AVX512VL, AVX512BW, AVX512DQ, AVX512CD PKU, AVX512VBMI, AVX512IFMA, SHA, AVX512VNNI, GFNI, VAES, AVX512VBMI2, VPCLMULQDQ, AVX512BITALG, RDPID, AVX512VPOPCNTDQ, MOVDIRI, MOVDIR64B, CLWB, AVX512VP2INTERSECT and KEYLOCKER instruction set support. she's long, she's powerful, she's everything 💅

rocketlake 🚀

Intel Rocket Lake CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, CX16, SAHF, FXSR, AVX, XSAVE, PCLMUL, FSGSBASE, RDRND, F16C, AVX2, BMI, BMI2, LZCNT, FMA, MOVBE, HLE, RDSEED, ADCX, PREFETCHW, AES, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F, AVX512VL, AVX512BW, AVX512DQ, AVX512CD PKU, AVX512VBMI, AVX512IFMA, SHA, AVX512VNNI, GFNI, VAES, AVX512VBMI2, VPCLMULQDQ, AVX512BITALG, RDPID and AVX512VPOPCNTDQ instruction set support.

alderlake, raptorlake, meteorlake, gracemont 🦖

four names, one entry, obsessed — Intel Alder Lake/Raptor Lake/Meteor Lake/Gracemont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, AES, PREFETCHW, PCLMUL, RDRND, XSAVE, XSAVEC, XSAVES, XSAVEOPT, FSGSBASE, PTWRITE, RDPID, SGX, GFNI-SSE, CLWB, MOVDIRI, MOVDIR64B, WAITPKG, ADCX, AVX, AVX2, BMI, BMI2, F16C, FMA, LZCNT, PCONFIG, PKU, VAES, VPCLMULQDQ, SERIALIZE, HRESET, KL, WIDEKL and AVX-VNNI instruction set support. no cap.

arrowlake 🏹

Intel Arrow Lake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, AES, PREFETCHW, PCLMUL, RDRND, XSAVE, XSAVEC, XSAVES, XSAVEOPT, FSGSBASE, PTWRITE, RDPID, SGX, GFNI-SSE, CLWB, MOVDIRI, MOVDIR64B, WAITPKG, ADCX, AVX, AVX2, BMI, BMI2, F16C, FMA, LZCNT, PCONFIG, PKU, VAES, VPCLMULQDQ, SERIALIZE, HRESET, KL, WIDEKL, AVX-VNNI, UINTR, AVXIFMA, AVXVNNIINT8, AVXNECONVERT and CMPCCXADD instruction set support.

arrowlake-s, lunarlake 🌙

Intel Arrow Lake S/Lunar Lake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, AES, PREFETCHW, PCLMUL, RDRND, XSAVE, XSAVEC, XSAVES, XSAVEOPT, FSGSBASE, PTWRITE, RDPID, SGX, GFNI-SSE, CLWB, MOVDIRI, MOVDIR64B, WAITPKG, ADCX, AVX, AVX2, BMI, BMI2, F16C, FMA, LZCNT, PCONFIG, PKU, VAES, VPCLMULQDQ, SERIALIZE, HRESET, KL, WIDEKL, AVX-VNNI, UINTR, AVXIFMA, AVXVNNIINT8, AVXNECONVERT, CMPCCXADD, AVXVNNIINT16, SHA512, SM3 and SM4 instruction set support. 🌸

pantherlake 🐆

Intel Panther Lake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, AES, PREFETCHW, PCLMUL, RDRND, XSAVE, XSAVEC, XSAVES, XSAVEOPT, FSGSBASE, PTWRITE, RDPID, SGX, GFNI-SSE, CLWB, MOVDIRI, MOVDIR64B, WAITPKG, ADCX, AVX, AVX2, BMI, BMI2, F16C, FMA, LZCNT, PCONFIG, PKU, VAES, VPCLMULQDQ, SERIALIZE, HRESET, AVX-VNNI, UINTR, AVXIFMA, AVXVNNIINT8, AVXNECONVERT, CMPCCXADD, AVXVNNIINT16, SHA512, SM3 and SM4 instruction set support.

sapphirerapids, emeraldrapids 💎

Intel Sapphire Rapids/Emerald Rapids CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, CX16, SAHF, FXSR, AVX, XSAVE, PCLMUL, FSGSBASE, RDRND, F16C, AVX2, BMI, BMI2, LZCNT, FMA, MOVBE, HLE, RDSEED, ADCX, PREFETCHW, AES, CLFLUSHOPT, XSAVEC, XSAVES, SGX, AVX512F, AVX512VL, AVX512BW, AVX512DQ, AVX512CD, PKU, AVX512VBMI, AVX512IFMA, SHA, AVX512VNNI, GFNI, VAES, AVX512VBMI2, VPCLMULQDQ, AVX512BITALG, RDPID, AVX512VPOPCNTDQ, PCONFIG, WBNOINVD, CLWB, MOVDIRI, MOVDIR64B, ENQCMD, CLDEMOTE, PTWRITE, WAITPKG, SERIALIZE, TSXLDTRK, UINTR, AMX-BF16, AMX-TILE, AMX-INT8, AVX-VNNI, AVX512-FP16 and AVX512BF16 instruction set support. the gemstone girlies 😭💖

graniterapids 🪨

Intel Granite Rapids CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, CX16, SAHF, FXSR, AVX, XSAVE, PCLMUL, FSGSBASE, RDRND, F16C, AVX2, BMI, BMI2, LZCNT, FMA, MOVBE, HLE, RDSEED, ADCX, PREFETCHW, AES, CLFLUSHOPT, XSAVEC, XSAVES, SGX, AVX512F, AVX512VL, AVX512BW, AVX512DQ, AVX512CD, PKU, AVX512VBMI, AVX512IFMA, SHA, AVX512VNNI, GFNI, VAES, AVX512VBMI2, VPCLMULQDQ, AVX512BITALG, RDPID, AVX512VPOPCNTDQ, PCONFIG, WBNOINVD, CLWB, MOVDIRI, MOVDIR64B, ENQCMD, CLDEMOTE, PTWRITE, WAITPKG, SERIALIZE, TSXLDTRK, UINTR, AMX-BF16, AMX-TILE, AMX-INT8, AVX-VNNI, AVX512-FP16, AVX512BF16, AMX-FP16 and PREFETCHI instruction set support.

graniterapids-d 🌟

Intel Granite Rapids D CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, CX16, SAHF, FXSR, AVX, XSAVE, PCLMUL, FSGSBASE, RDRND, F16C, AVX2, BMI, BMI2, LZCNT, FMA, MOVBE, HLE, RDSEED, ADCX, PREFETCHW, AES, CLFLUSHOPT, XSAVEC, XSAVES, SGX, AVX512F, AVX512VL, AVX512BW, AVX512DQ, AVX512CD, PKU, AVX512VBMI, AVX512IFMA, SHA, AVX512VNNI, GFNI, VAES, AVX512VBMI2, VPCLMULQDQ, AVX512BITALG, RDPID, AVX512VPOPCNTDQ, PCONFIG, WBNOINVD, CLWB, MOVDIRI, MOVDIR64B, ENQCMD, CLDEMOTE, PTWRITE, WAITPKG, SERIALIZE, TSXLDTRK, UINTR, AMX-BF16, AMX-TILE, AMX-INT8, AVX-VNNI, AVX512FP16, AVX512BF16, AMX-FP16, PREFETCHI and AMX-COMPLEX instruction set support. the way she packed AMX-COMPLEX in at the end and dipped 💀👑

heads up bestie: that's the cpu-type list — spot the one you actually have before you pass it to the tuning flags, okay? 🙈

• diamondrapids — Intel Diamond Rapids CPU, and the 64-bit extensions + instruction set support list is long bestie, breathe: MOVBE, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, CX16, SAHF, FXSR, AVX, XSAVE, PCLMUL, FSGSBASE, RDRND, F16C, AVX2, BMI, BMI2, LZCNT, FMA, MOVBE, HLE, RDSEED, ADCX, PREFETCHW, AES, CLFLUSHOPT, XSAVEC, XSAVES, SGX, AVX512F, AVX512VL, AVX512BW, AVX512DQ, AVX512CD, PKU, AVX512VBMI, AVX512IFMA, SHA, AVX512VNNI, GFNI, VAES, AVX512VBMI2, VPCLMULQDQ, AVX512BITALG, RDPID, AVX512VPOPCNTDQ, PCONFIG, WBNOINVD, CLWB, MOVDIRI, MOVDIR64B, ENQCMD, CLDEMOTE, PTWRITE, WAITPKG, SERIALIZE, TSXLDTRK, UINTR, AMX-BF16, AMX-TILE, AMX-INT8, AVX-VNNI, AVX512FP16, AVX512BF16, AMX-FP16, PREFETCHI, AMX-COMPLEX, AVX10.1-512, AVX-IFMA, AVX-NE-CONVERT, AVX-VNNI-INT16, AVX-VNNI-INT8, CMPccXADD, SHA512, SM3, SM4, AVX10.2-512, APX_F, AMX-AVX512, AMX-FP8, MOVRS and AMX-MOVRS instruction set support. 💅 she ate and left no crumbs.
• bonnell, atom — Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3 and SSSE3 instruction set support. small but mighty fr 🦋
• silvermont, slm — Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, CX16, SAHF, FXSR, PCLMUL, PREFETCHW and RDRND instruction set support. ✨
• goldmont — Intel Goldmont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, CX16, SAHF, FXSR, PCLMUL, PREFETCHW, RDRND, AES, SHA, RDSEED, XSAVE, XSAVEC, XSAVES, XSAVEOPT, CLFLUSHOPT and FSGSBASE instruction set support. 💖
• goldmont-plus — Intel Goldmont Plus CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, CX16, SAHF, FXSR, PCLMUL, PREFETCHW, RDRND, AES, SHA, RDSEED, XSAVE, XSAVEC, XSAVES, XSAVEOPT, CLFLUSHOPT, FSGSBASE, PTWRITE, RDPID and SGX instruction set support. the plus is giving glow-up energy 🎀
• tremont — Intel Tremont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, CX16, SAHF, FXSR, PCLMUL, PREFETCHW, RDRND, AES, SHA, RDSEED, XSAVE, XSAVEC, XSAVES, XSAVEOPT, CLFLUSHOPT, FSGSBASE, PTWRITE, RDPID, SGX, CLWB, GFNI-SSE, MOVDIRI, MOVDIR64B, CLDEMOTE and WAITPKG instruction set support. 🔥
• sierraforest — Intel Sierra Forest CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, AES, PREFETCHW, PCLMUL, RDRND, XSAVE, XSAVEC, XSAVES, XSAVEOPT, FSGSBASE, PTWRITE, RDPID, SGX, GFNI-SSE, CLWB, MOVDIRI, MOVDIR64B, CLDEMOTE, WAITPKG, ADCX, AVX, AVX2, BMI, BMI2, F16C, FMA, LZCNT, PCONFIG, PKU, VAES, VPCLMULQDQ, SERIALIZE, HRESET, KL, WIDEKL, AVX-VNNI, AVXIFMA, AVXVNNIINT8, AVXNECONVERT, CMPCCXADD, ENQCMD and UINTR instruction set support. 🌲 iconic name tbh 🌈
• grandridge — Intel Grand Ridge CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, AES, PREFETCHW, PCLMUL, RDRND, XSAVE, XSAVEC, XSAVES, XSAVEOPT, FSGSBASE, PTWRITE, RDPID, SGX, GFNI-SSE, CLWB, MOVDIRI, MOVDIR64B, CLDEMOTE, WAITPKG, ADCX, AVX, AVX2, BMI, BMI2, F16C, FMA, LZCNT, PCONFIG, PKU, VAES, VPCLMULQDQ, SERIALIZE, HRESET, KL, WIDEKL, AVX-VNNI, AVXIFMA, AVXVNNIINT8, AVXNECONVERT, CMPCCXADD, ENQCMD and UINTR instruction set support. 💜
• clearwaterforest — Intel Clearwater Forest CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, AES, PREFETCHW, PCLMUL, RDRND, XSAVE, XSAVEC, XSAVES, XSAVEOPT, FSGSBASE, PTWRITE, RDPID, SGX, GFNI-SSE, CLWB, MOVDIRI, MOVDIR64B, CLDEMOTE, WAITPKG, ADCX, AVX, AVX2, BMI, BMI2, F16C, FMA, LZCNT, PCONFIG, PKU, VAES, VPCLMULQDQ, SERIALIZE, HRESET, AVX-VNNI, ENQCMD, UINTR, AVXIFMA, AVXVNNIINT8, AVXNECONVERT, CMPCCXADD, AVXVNNIINT16, SHA512, SM3, SM4, USER_MSR and PREFETCHI instruction set support. 💧✨
• k6 — AMD K6 CPU with MMX instruction set support. vintage but valid 🍓
• k6-2, k6-3 — Improved versions of AMD K6 CPU with MMX and 3DNow! instruction set support. glow-up confirmed 💅
• athlon, athlon-tbird — AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow! and SSE prefetch instructions support. 🧚
• athlon-4, athlon-xp, athlon-mp — Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow! and full SSE instruction set support. 👑
• k8, opteron, athlon64, athlon-fx — Processors based on the AMD K8 core with x86-64 instruction set support, including the AMD Opteron, Athlon 64, and Athlon 64 FX processors. (This supersets MMX, SSE, SSE2, 3DNow!, enhanced 3DNow! and 64-bit instruction set extensions.) 💖
• k8-sse3, opteron-sse3, athlon64-sse3 — Improved versions of AMD K8 cores with SSE3 instruction set support. 🌸
• amdfam10, barcelona — CPUs based on AMD Family 10h cores with x86-64 instruction set support. (This supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit instruction set extensions.) 🔥
• bdver1 — CPUs based on AMD Family 15h cores with x86-64 instruction set support. (This supersets FMA4, AVX, XOP, LWP, AES, PCLMUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.) ✨
• bdver2 — AMD Family 15h core based CPUs with x86-64 instruction set support. (This supersets BMI, TBM, F16C, FMA, FMA4, AVX, XOP, LWP, AES, PCLMUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.) 💜
• bdver3 — AMD Family 15h core based CPUs with x86-64 instruction set support. (This supersets BMI, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, XOP, LWP, AES, PCLMUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.) 🦋
• bdver4 — AMD Family 15h core based CPUs with x86-64 instruction set support. (This supersets BMI, BMI2, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, AVX2, XOP, LWP, AES, PCLMUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.) the way they kept stacking features… main character energy fr 👑
• znver1 — AMD Family 17h core based CPUs with x86-64 instruction set support. (This supersets BMI, BMI2, F16C, FMA, FSGSBASE, AVX, AVX2, ADCX, RDSEED, MWAITX, SHA, CLZERO, AES, PCLMUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, and 64-bit instruction set extensions.) ✨ (yes bestie, there's an Oxford comma in there and we are keeping it, no cap)
• znver2 — AMD Family 17h core based CPUs with x86-64 instruction set support. (This supersets BMI, BMI2, CLWB, F16C, FMA, FSGSBASE, AVX, AVX2, ADCX, RDSEED, MWAITX, SHA, CLZERO, AES, PCLMUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, RDPID, WBNOINVD, and 64-bit instruction set extensions.) 💅
• znver3 — AMD Family 19h core based CPUs with x86-64 instruction set support. (This supersets BMI, BMI2, CLWB, F16C, FMA, FSGSBASE, AVX, AVX2, ADCX, RDSEED, MWAITX, SHA, CLZERO, AES, PCLMUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, RDPID, WBNOINVD, PKU, VPCLMULQDQ, VAES, and 64-bit instruction set extensions.) 🌸
• znver4 — AMD Family 19h core based CPUs with x86-64 instruction set support. (This supersets BMI, BMI2, CLWB, F16C, FMA, FSGSBASE, AVX, AVX2, ADCX, RDSEED, MWAITX, SHA, CLZERO, AES, PCLMUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, RDPID, WBNOINVD, PKU, VPCLMULQDQ, VAES, AVX512F, AVX512DQ, AVX512IFMA, AVX512CD, AVX512BW, AVX512VL, AVX512BF16, AVX512VBMI, AVX512VBMI2, AVX512VNNI, AVX512BITALG, AVX512VPOPCNTDQ, GFNI and 64-bit instruction set extensions.) 🔥
• znver5 — AMD Family 1ah core based CPUs with x86-64 instruction set support. (This supersets BMI, BMI2, CLWB, F16C, FMA, FSGSBASE, AVX, AVX2, ADCX, RDSEED, MWAITX, SHA, CLZERO, AES, PCLMUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, RDPID, WBNOINVD, PKU, VPCLMULQDQ, VAES, AVX512F, AVX512DQ, AVX512IFMA, AVX512CD, AVX512BW, AVX512VL, AVX512BF16, AVX512VBMI, AVX512VBMI2, AVX512VNNI, AVX512BITALG, AVX512VPOPCNTDQ, GFNI, AVXVNNI, MOVDIRI, MOVDIR64B, AVX512VP2INTERSECT, PREFETCHI and 64-bit instruction set extensions.) the newest sibling and it's serving 💜
• btver1 — CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit instruction set extensions.) 🧚
• btver2 — CPUs based on AMD Family 16h cores with x86-64 instruction set support. This includes MOVBE, F16C, BMI, AVX, PCLMUL, AES, SSE4.2, SSE4.1, CX16, ABM, SSE4A, SSSE3, SSE3, SSE2, SSE, MMX and 64-bit instruction set extensions. 🦋
• winchip-c6 — IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction set support. 🍓
• winchip2 — IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow! instruction set support. 🎀
• c3 — VIA C3 CPU with MMX and 3DNow! instruction set support. (No scheduling is implemented for this chip.) 🫶
• c3-2 — VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support. (No scheduling is implemented for this chip.) 💖
• c7 — VIA C7 (Esther) CPU with MMX, SSE, SSE2 and SSE3 instruction set support. (No scheduling is implemented for this chip.) ✨
• samuel-2 — VIA Eden Samuel 2 CPU with MMX and 3DNow! instruction set support. (No scheduling is implemented for this chip.) 🌸
• nehemiah — VIA Eden Nehemiah CPU with MMX and SSE instruction set support. (No scheduling is implemented for this chip.) 💅
• esther — VIA Eden Esther CPU with MMX, SSE, SSE2 and SSE3 instruction set support. (No scheduling is implemented for this chip.) 🔥
• eden-x2 — VIA Eden X2 CPU with x86-64, MMX, SSE, SSE2 and SSE3 instruction set support. (No scheduling is implemented for this chip.) 👑
• eden-x4 — VIA Eden X4 CPU, slinging x86-64, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, AVX and AVX2 instruction set support. (Ngl, no scheduling is implemented for this chip. She's giving vibes only.)
• nano — the generic VIA Nano CPU, working with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3 instruction set support. (Again, no scheduling is implemented for this chip.)
• nano-1000 — VIA Nano 1xxx CPU, rockin' x86-64, MMX, SSE, SSE2, SSE3 and SSSE3 instruction set support. (No scheduling implemented for this chip either, bestie.)
• nano-2000 — VIA Nano 2xxx CPU, serving x86-64, MMX, SSE, SSE2, SSE3 and SSSE3 instruction set support. (Also no scheduling. It's a whole thing.)
• nano-3000 — VIA Nano 3xxx CPU, with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1 instruction set support. (And once more: no scheduling is implemented for this chip.)
• nano-x2 — VIA Nano Dual Core CPU, holding x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1 instruction set support. (You guessed it — no scheduling for this chip.)
• nano-x4 — VIA Nano Quad Core CPU, carrying x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1 instruction set support. (No scheduling implemented for this chip. Periodt.)
• lujiazui — ZHAOXIN lujiazui CPU, and she's stacked: x86-64, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, AES, PCLMUL, RDRND, XSAVE, XSAVEOPT, FSGSBASE, CX16, ABM, BMI, BMI2, FXSR, RDSEED instruction set support. Okay but lowkey important: while these CPUs do support AVX and F16C, those aren't enabled by -march=lujiazui for performance reasons.
• yongfeng — ZHAOXIN yongfeng CPU, with x86-64, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, AVX, POPCNT, AES, PCLMUL, RDRND, XSAVE, XSAVEOPT, FSGSBASE, CX16, ABM, BMI, BMI2, F16C, FXSR, RDSEED, AVX2, FMA, SHA, LZCNT instruction set support. Iconic.
• shijidadao — ZHAOXIN shijidadao CPU, matching the same energy: x86-64, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, AVX, POPCNT, AES, PCLMUL, RDRND, XSAVE, XSAVEOPT, FSGSBASE, CX16, ABM, BMI, BMI2, F16C, FXSR, RDSEED, AVX2, FMA, SHA, LZCNT instruction set support.
• geode — AMD Geode embedded processor, keeping it simple with MMX and 3DNow! instruction set support. 🦋
• c86-4g-m4 — HYGON c86-4g-m4 CPU, absolutely serving instruction sets: x86-64, MMX, SSE, SSE2, SSE3, SSE4A, CX16, ABM, SSSE3, SSE4.1, SSE4.2, AES, PCLMUL, AVX, AVX2, BMI, BMI2, F16C, FMA, PRFCHW, FXSR, SHA, XSAVE, XSAVEOPT, XSAVEC, FSGSBASE, RDRND, MOVBE, MWAITX, ADX, RDSEED, CLZERO, CLFLUSHOPT, XSAVES, LZCNT, POPCNT instruction set support.
• c86-4g-m6 — HYGON c86-4g-m6 CPU, and the list is identical bestie: x86-64, MMX, SSE, SSE2, SSE3, SSE4A, CX16, ABM, SSSE3, SSE4.1, SSE4.2, AES, PCLMUL, AVX, AVX2, BMI, BMI2, F16C, FMA, PRFCHW, FXSR, SHA, XSAVE, XSAVEOPT, XSAVEC, FSGSBASE, RDRND, MOVBE, MWAITX, ADX, RDSEED, CLZERO, CLFLUSHOPT, XSAVES, LZCNT, POPCNT instruction set support.
• c86-4g-m7 — HYGON c86-4g-m7 CPU, and she ate down: x86-64, MMX, SSE, SSE2, SSE3, SSE4A, CX16, ABM, SSSE3, SSE4.1, SSE4.2, AES, PCLMUL, AVX, AVX2, BMI, BMI2, F16C, FMA, PRFCHW, FXSR, SHA, XSAVE, XSAVEOPT, XSAVEC, FSGSBASE, RDRND, MOVBE, MWAITX, ADX, RDSEED, CLZERO, CLFLUSHOPT, XSAVES, LZCNT, POPCNT, AVX512F, AVX512DQ, AVX512IFMA, AVX512CD, AVX512BW, AVX512VL, AVX512BF16, AVX512VBMI, AVX512VBMI2, GFNI, AVX512VNNI, VAES, AVX512BITALG, AVX512VPOPCNTDQ, AVX512VP2INTERSECT, AVXVNNI, VPCLMULQDQ, WBNOINVD instruction set support. 💅

---

🎛️ -mtune=cpu-type

Okay so this one tunes everything applicable about the generated code to cpu-type — everything except the ABI and the set of available instructions. The way it works: picking a specific cpu-type schedules things appropriately for that particular chip, but the compiler does not generate any code that cannot run on the default machine type unless you go use a -march=cpu-type option. So like, if GCC is configured for i686-pc-linux-gnu, then -mtune=pentium4 generates code that's tuned for Pentium 4 but still runs on i686 machines. Cute, right?

The choices for cpu-type are the same as for -march. But -mtune also supports 2 extra choices for cpu-type:

• generic — Produce code optimized for the most common IA32/AMD64/EM64T processors. Real talk: if you know the CPU your code is gonna run on, use the corresponding -mtune or -march option instead of -mtune=generic. But if you do not know exactly what CPU users of your application will have? Then yeah, use this one.

As new processors get deployed in the marketplace, the behavior of this option will change. So if you upgrade to a newer version of GCC, code generation controlled by this option will change to reflect the processors that are most common at the time that version of GCC is released.

There's no -march=generic option, and here's why: -march indicates the instruction set the compiler can use, and there is no generic instruction set applicable to all processors. Whereas -mtune indicates the processor (or, in this case, collection of processors) for which the code is optimized.

• intel — Produce code optimized for the most current Intel processors, which are Haswell and Silvermont for this version of GCC. Same deal: if you know the CPU your code will run on, use the corresponding -mtune or -march option instead of -mtune=intel. But if you want your application to perform better on both Haswell and Silvermont, then you should use this option.

As new Intel processors are deployed in the marketplace, the behavior of this option will change. So if you upgrade to a newer version of GCC, code generation controlled by this option will change to reflect the most current Intel processors at the time that version of GCC is released.

There is no -march=intel option because -march indicates the instruction set the compiler can use, and there is no common instruction set applicable to all processors. In contrast, -mtune indicates the processor (or, in this case, collection of processors) for which the code is optimized.

---

💀 -mcpu=cpu-type

A deprecated synonym for -mtune. That's it. That's the whole vibe. Don't come for me, I didn't write it.

---

🧮 -mfpmath=unit

Generate floating-point arithmetic for the selected unit, okay? And the choices for unit are, in order:

• 387 — Use the standard 387 floating-point coprocessor present on the majority of chips, and emulated otherwise. Code compiled with this option runs almost everywhere. The temporary results are computed in 80-bit precision instead of the precision specified by the type, which results in slightly different results compared to most other chips. See -ffloat-store for a more detailed description.

This is the default choice for non-Darwin x86-32 targets.

• sse — Use scalar floating-point instructions present in the SSE instruction set. This instruction set is supported by Pentium III and newer chips, and in the AMD line by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE instruction set supports only single-precision arithmetic, so the double and extended-precision arithmetic are still done using 387. A later version, present only in Pentium 4 and AMD x86-64 chips, supports double-precision arithmetic too.

For the x86-32 compiler, you must use -march=cpu-type, -msse or -msse2 switches to enable SSE extensions and make this option effective. For the x86-64 compiler, these extensions are enabled by default.

The resulting code should be considerably faster in the majority of cases and avoid the numerical instability problems of 387 code, but may break some existing code that expects temporaries to be 80 bits. Bestie, be careful.

This is the default choice for the x86-64 compiler, Darwin x86-32 targets, and the default choice for x86-32 targets with the SSE2 instruction set when -ffast-math is enabled.

• sse,387
• sse+387
• both — Attempt to utilize both instruction sets at once. This effectively doubles the amount of available registers, and on chips with separate execution units for 387 and SSE it doubles the execution resources too. Use this option with care, because it is still experimental — the GCC register allocator does not model separate functional units well, which results in unstable performance. 🫠

---

🗣️ -masm=dialect

Output assembly instructions using the selected dialect. It also affects which dialect is used for basic asm and extended asm. Supported choices (in dialect order) are att or intel. The default is att. And btw, Darwin does not support intel.

---

🌊 -mieee-fp / -mno-ieee-fp

Control whether or not the compiler uses IEEE floating-point comparisons. These correctly handle the case where the result of a comparison is unordered. Very demure, very mindful. 😌

---

🔢 -m80387 / -mhard-float

Generate output containing 80387 instructions for floating point.

---

🧸 -mno-80387 / -msoft-float

Generate output containing library calls for floating point.

Warning: the requisite libraries are not part of GCC. Normally the facilities of the machine's usual C compiler are used, but this cannot be done directly in cross-compilation. You must make your own arrangements to provide suitable library functions for cross-compilation. 😭

On machines where a function returns floating-point results in the 80387 register stack, some floating-point opcodes may be emitted even if -msoft-float is used.

---

🚫 -mno-fp-ret-in-387

Do not use the FPU registers for return values of functions.

The usual calling convention has functions return values of types float and double in an FPU register, even if there is no FPU. The idea is that the operating system should emulate an FPU. Wild, I know.

The option -mno-fp-ret-in-387 causes such values to be returned in ordinary CPU registers instead. 👑

• -mno-fancy-math-387 — okay so some 387 emulators are out here NOT supporting the "sin", "cos" and "sqrt" instructions for the 387 💀 so if that's your situation, specify this option to avoid generating those instructions. Heads up though: this option gets overridden when -march says the target CPU always has an FPU, because then the instruction doesn't need emulation at all, periodt. Also, these instructions are not generated unless you also use the -funsafe-math-optimizations switch. iykyk ✨
• -malign-double, -mno-align-double — these control whether GCC aligns "double", "long double", and "long long" variables on a two-word boundary or a one-word boundary. Aligning "double" variables on a two-word boundary produces code that runs somewhat faster on a Pentium, at the expense of more memory. trade-offs, bestie, we stay making them 💅

On x86-64, -malign-double is enabled by default.

Warning: if you use the -malign-double switch, structures containing the above types are aligned differently than the published application binary interface specifications for the x86-32, and are not binary compatible with structures in code compiled without that switch. she's not compatible, she's not for everybody, no cap 😭

• -m96bit-long-double, -m128bit-long-double — these switches control the size of the "long double" type. The x86-32 application binary interface specifies the size to be 96 bits, so -m96bit-long-double is the default in 32-bit mode.

Modern architectures (Pentium and newer) prefer "long double" to be aligned to an 8- or 16-byte boundary. In arrays or structures conforming to the ABI, this is not possible. So specifying -m128bit-long-double aligns "long double" to a 16-byte boundary by padding the "long double" with an additional 32-bit zero. iconic behavior honestly 🫶

In the x86-64 compiler, -m128bit-long-double is the default choice, as its ABI specifies that "long double" is aligned on a 16-byte boundary.

Notice that neither of these options enable any extra precision over the x87 standard of 80 bits for a "long double". it's giving size, not precision, bestie 🎀

Warning: if you override the default value for your target ABI, this changes the size of structures and arrays containing "long double" variables, as well as modifying the function calling convention for functions taking "long double". Hence they are not binary-compatible with code compiled without that switch. 💀

• -mlong-double-64, -mlong-double-80, -mlong-double-128 — these switches control the size of the "long double" type. A size of 64 bits makes the "long double" type equivalent to the "double" type. This is the default for 32-bit Bionic C library. A size of 128 bits makes the "long double" type equivalent to the "__float128" type. This is the default for 64-bit Bionic C library.

Warning: if you override the default value for your target ABI, this changes the size of structures and arrays containing "long double" variables, as well as modifying the function calling convention for functions taking "long double". Hence they are not binary-compatible with code compiled without that switch. 🙈 same energy, different option, we respect the pattern

• -malign-data=type — controls how GCC aligns variables. Supported values for type are:
  • compat — uses increased alignment value compatible with GCC 4.8 and earlier
  • abi — uses alignment value as specified by the psABI
  • cacheline — uses increased alignment value to match the cache line size

compat is the default. slay, legacy-friendly 💖

• -mlarge-data-threshold=threshold — when -mcmodel=medium or -mcmodel=large is specified, data objects larger than threshold are placed in large data sections. The default is 65535. manifesting big data energy ✨
• -mrtd — use a different function-calling convention, in which functions that take a fixed number of arguments return with the "ret num" instruction, which pops their arguments while returning. This saves one instruction in the caller, since there's no need to pop the arguments there. efficiency queen 👑

You can specify that an individual function is called with this calling sequence with the function attribute "stdcall". You can also override the -mrtd option by using the function attribute "cdecl".

Warning: this calling convention is incompatible with the one normally used on Unix, so you cannot use it if you need to call libraries compiled with the Unix compiler. no mixing, bestie 😭

Also, you must provide function prototypes for all functions that take variable numbers of arguments (including "printf"); otherwise incorrect code is generated for calls to those functions. 💀

In addition, seriously incorrect code results if you call a function with too many arguments. (Normally, extra arguments are harmlessly ignored.) the way that nobody warns you about this… well, the manpage does, actually 🎀

• -mregparm=num — controls how many registers are used to pass integer arguments. By default, no registers are used to pass arguments, and at most 3 registers can be used. You can control this behavior for a specific function by using the function attribute "regparm".

Warning: if you use this switch, and num is nonzero, then you must build all modules with the same value, including any libraries. This includes the system libraries and startup modules. all of them. every single one. no exceptions 💅

• -msseregparm — use SSE register passing conventions for float and double arguments and return values. You can control this behavior for a specific function by using the function attribute "sseregparm". registers... passing... it's giving speed 🌈

Warning: if you use this switch then you must build all modules with the same value, including any libraries. This includes the system libraries and startup modules. yes bestie the system libraries too, I'm so serious 🙈

• -mvect8-ret-in-mem — return 8-byte vectors in memory instead of MMX registers. This is the default on VxWorks, to match the ABI of the Sun Studio compilers until version 12. Only use this option if you need to remain compatible with existing code produced by those previous compiler versions or older versions of GCC. legacy compatibility era, we love a callback 🫶
• -mpc32, -mpc64, -mpc80 — set 80387 floating-point precision to 32, 64 or 80 bits. When -mpc32 is specified, the significands of results of floating-point operations are rounded to 24 bits (single precision); -mpc64 rounds the significands of results of floating-point operations to 53 bits (double precision) and -mpc80 rounds the significands of results of floating-point operations to 64 bits (extended double precision), which is the default. When this option is used, floating-point operations in higher precisions are not available to the programmer without setting the FPU control word explicitly. you gotta do that part yourself, bestie 💀

Setting the rounding of floating-point operations to less than the default 80 bits can speed some programs by 2% or more. Note that some mathematical libraries assume that extended-precision (80-bit) floating-point operations are enabled by default; routines in such libraries could suffer significant loss of accuracy, typically through so-called "catastrophic cancellation", when this option is used to set the precision to less than extended precision. and that name?? catastrophic cancellation?? ate and left no crumbs 😭

• -mdaz-ftz — the flush-to-zero (FTZ) and denormals-are-zero (DAZ) flags in the MXCSR register are used to control floating-point calculations. SSE and AVX instructions, including scalar and vector instructions, could benefit from enabling the FTZ and DAZ flags when -mdaz-ftz is specified. Don't set FTZ/DAZ flags when -mno-daz-ftz or -shared is specified — -mdaz-ftz will set FTZ/DAZ flags even with -shared. yeah she does what she wants with -shared, noted 📝
• -mstackrealign — realign the stack at entry. On the x86, the -mstackrealign option generates an alternate prologue and epilogue that realigns the run-time stack if necessary. This supports mixing legacy codes that keep 4-byte stack alignment with modern codes that keep 16-byte stack alignment, for SSE compatibility. See also the attribute "force_align_arg_pointer", applicable to individual functions. blending eras, we love to see it 💖
• -mpreferred-stack-boundary=num — attempt to keep the stack boundary aligned to a 2 raised to num byte boundary. If -mpreferred-stack-boundary is not specified, the default is 4 (16 bytes or 128 bits).

Warning: when generating code for the x86-64 architecture with SSE extensions disabled, -mpreferred-stack-boundary=3 can be used to keep the stack boundary aligned to an 8 byte boundary. Since x86-64 ABI require 16 byte stack alignment, this is ABI incompatible and intended to be used in controlled environment where stack space is important limitation. This option leads to wrong code when functions compiled with 16 byte stack alignment (such as functions from a standard library) are called with misaligned stack. In this case, SSE instructions may lead to misaligned memory access traps. In addition, variable arguments are handled incorrectly for 16 byte aligned objects (including x87 long double and __int128), leading to wrong results. You must build all modules with -mpreferred-stack-boundary=3, including any libraries. This includes the system libraries and startup modules. 😭 okay so this one is serious serious, controlled environment only

• -mincoming-stack-boundary=num — assume the incoming stack is aligned to a 2 raised to num byte boundary. If -mincoming-stack-boundary is not specified, the one specified by -mpreferred-stack-boundary is used. lowkey just trust the boundary you were given ✨

On Pentium and Pentium Pro, "double" and "long double" values should be aligned to an 8-byte boundary (see -malign-double) or suffer significant run time performance penalties. On Pentium III, the Streaming SIMD Extension (SSE) data type "__m128" may not work properly if it is not 16-byte aligned.

To ensure proper alignment of this values on the stack, the stack boundary must be as aligned as that required by any value stored on the stack. Further, every function must be generated such that it keeps the stack aligned. Thus calling a function compiled with a higher preferred stack boundary from a function compiled with a lower preferred stack boundary most likely misaligns the stack. It is recommended that libraries that use callbacks always use the default setting.

This extra alignment does consume extra stack space, and generally increases code size. Code that is sensitive to stack space usage, such as embedded systems and operating system kernels, may want to reduce the preferred alignment to -mpreferred-stack-boundary=2. tiny stack energy, we respect the grind 🎀

• -mmmx
• -msse
• -msse2
• -msse3
• -mssse3
• -msse4
• -msse4a
• -msse4.1
• -msse4.2
• -mavx
• -mavx2
• -mavx512f
• -mavx512cd
• -mavx512vl
• -mavx512bw
• -mavx512dq
• -mavx512ifma
• -mavx512vbmi
• -msha
• -maes
• -mpclmul
• -mclflushopt
• -mclwb
• -mfsgsbase
• -mptwrite
• -mrdrnd
• -mf16c
• -mfma
• -mpconfig
• -mwbnoinvd
• -mfma4
• -mprfchw
• -mrdpid
• -mrdseed
• -msgx
• -mxop
• -mlwp
• -m3dnow
• -m3dnowa
• -mpopcnt
• -mabm
• -madx
• -mbmi
• -mbmi2
• -mlzcnt
• -mfxsr
• -mxsave
• -mxsaveopt
• -mxsavec
• -mxsaves
• -mrtm
• -mhle
• -mtbm
• -mmwaitx
• -mclzero
• -mpku
• -mavx512vbmi2
• -mavx512bf16
• -mavx512fp16
• -mgfni
• -mvaes
• -mwaitpkg
• -mvpclmulqdq
• -mavx512bitalg
• -mmovdiri
• -mmovdir64b
• -menqcmd
• -muintr
• -mtsxldtrk
• -mavx512vpopcntdq
• -mavx512vp2intersect
• -mavx512vnni
• -mavxvnni
• -mcldemote
• -mserialize
• -mamx-tile
• -mamx-int8
• -mamx-bf16
• -mhreset
• -mkl
• -mwidekl
• -mavxifma
• -mavxvnniint8
• -mavxneconvert
• -mcmpccxadd
• -mamx-fp16
• -mprefetchi
• -mraoint
• -mamx-complex
• -mavxvnniint16
• -msm3
• -msha512
• -msm4
• -mapxf
• -musermsr
• -mavx10.1
• -mavx10.1-256
• -mavx10.1-512
• -mavx10.2
• -mamx-avx512
• -mamx-tf32
• -mamx-transpose
• -mamx-fp8
• -mmovrs
• -mamx-movrs

okay so bestie, these switches are your all-access pass to the MMX, SSE, AVX512CD, AVX512VL, AVX512BW, AVX512DQ, AVX512IFMA, AVX512VBMI, SHA, AES, PCLMUL, CLFLUSHOPT, CLWB, FSGSBASE, PTWRITE, RDRND, F16C, FMA, PCONFIG, WBNOINVD, FMA4, PREFETCHW, RDPID, RDSEED, SGX, XOP, LWP, 3DNow!, enhanced 3DNow!, POPCNT, ABM, ADX, BMI, BMI2, LZCNT, FXSR, XSAVE, XSAVEOPT, XSAVEC, XSAVES, RTM, HLE, TBM, MWAITX, CLZERO, PKU, AVX512VBMI2, GFNI, VAES, WAITPKG, VPCLMULQDQ, AVX512BITALG, MOVDIRI, MOVDIR64B, AVX512BF16, ENQCMD, AVX512VPOPCNTDQ, AVX512VNNI, SERIALIZE, UINTR, HRESET, AMXTILE, AMXINT8, AMXBF16, KL, WIDEKL, AVXVNNI, AVX512-FP16, AVXIFMA, AVXVNNIINT8, AVXNECONVERT, CMPCCXADD, AMX-FP16, PREFETCHI, RAOINT, AMX-COMPLEX, AVXVNNIINT16, SM3, SHA512, SM4, APX_F, USER_MSR, AVX10.1, AVX10.2, AMX-AVX512, AMX-TF32, AMX-TRANSPOSE, AMX-FP8, MOVRS, AMX-MOVRS or CLDEMOTE extended instruction sets. Each has a corresponding -mno- option to disable use of these instructions.

These extensions are also available as built-in functions: see x86 Built-in Functions, for details of the functions enabled and disabled by these switches.

Note that -msse4 enables both SSE4.1 and SSE4.2 support, while -mno-sse4 turns off those features; neither form of the option affects SSE4A support, controlled separately by -msse4a.

To generate SSE/SSE2 instructions automatically from floating-point code (as opposed to 387 instructions), see -mfpmath=sse.

GCC depresses SSEx instructions when -mavx is used. Instead, it generates new AVX instructions or AVX equivalence for all SSEx instructions when needed.

These options enable GCC to use these extended instructions in generated code, even without -mfpmath=sse. Applications that perform run-time CPU detection must compile separate files for each supported architecture, using the appropriate flags. In particular, the file containing the CPU detection code should be compiled without these options.

• -mdump-tune-features — This one tells GCC to spill the tea on all the x86 performance tuning features and their default settings. 🍵 The names can be used in -mtune-ctrl=feature-list.
• -mtune-ctrl=feature-list — This is for fine-grain control of x86 code generation features. feature-list is a comma-separated list of feature names. See also -mdump-tune-features. When you specify a feature, it’s turned on if it is not preceded with ^, otherwise it’s turned off. -mtune-ctrl=feature-list is intended to be used by GCC developers. Using it may lead to code paths not covered by testing and can potentially result in compiler ICEs or runtime errors. 😬
• -mno-default — Tells GCC to turn off all tunable features. See also -mtune-ctrl=feature-list and -mdump-tune-features.
• -mcld — This makes GCC emit a "cld" instruction in the prologue of functions that use string instructions. String instructions depend on the DF flag to choose between autoincrement or autodecrement mode. While the ABI specifies the DF flag to be cleared on function entry, some operating systems violate this specification by not clearing the DF flag in their exception dispatchers. 😤 The exception handler can be invoked with the DF flag set, which leads to wrong direction mode when string instructions are used. This option can be enabled by default on 32-bit x86 targets by configuring GCC with the --enable-cld configure option. Generation of "cld" instructions can be suppressed with the -mno-cld compiler option in this case.
• -mvzeroupper — This tells GCC to emit a "vzeroupper" instruction before a transfer of control flow out of the function. It minimizes the AVX to SSE transition penalty as well as removes unnecessary "zeroupper" intrinsics. ✨
• -mprefer-avx128 — Tells GCC to use 128-bit AVX instructions instead of 256-bit AVX instructions in the auto-vectorizer. 💅
• -mprefer-vector-width=opt — This tells GCC to use opt-bit vector width in instructions instead of default on the selected platform.
• -mpartial-vector-fp-math — This lets GCC generate floating-point operations that might affect the set of floating-point status flags on partial vectors, where vector elements reside in the low part of the 128-bit SSE register. Unless -fno-trapping-math is specified, the compiler guarantees correct behavior by sanitizing all input operands to have zeroes in the unused upper part of the vector register. But heads up: by using built-in functions or inline assembly with partial vector arguments, NaNs, denormal or invalid values can leak into the upper part of the vector, causing possible performance issues when -fno-trapping-math is in effect. 😱 These issues can be mitigated by manually sanitizing the upper part of the partial vector argument register or by using -mdaz-ftz to set denormals-are-zero (DAZ) flag in the MXCSR register. This option is enabled by default.
• -mmove-max=bits — This sets the maximum number of bits that can be moved from memory to memory efficiently to bits. The valid bits are 128, 256 and 512.
• -mstore-max=bits — This sets the maximum number of bits that can be stored to memory efficiently to bits. The valid bits are 128, 256 and 512.
• none — No extra limitations applied to GCC other than defined by the selected platform.
• 128 — Prefer 128-bit vector width for instructions.
• 256 — Prefer 256-bit vector width for instructions.
• 512 — Prefer 512-bit vector width for instructions.
• -mnoreturn-no-callee-saved-registers — This optimizes functions with "noreturn" attribute or "_Noreturn" specifier by not saving in the function prologue callee-saved registers which are used in the function (except for the "BP" register). But this can interfere with debugging of the caller of the "noreturn" function or any function further up in the call stack, so it is not enabled by default. 🚨
• -mcx16 — This lets GCC generate "CMPXCHG16B" instructions in 64-bit code to implement compare-and-exchange operations on 16-byte aligned 128-bit objects. Super useful for atomic updates of data structures exceeding one machine word in size. The compiler uses this instruction to implement __sync Builtins. However, for __atomic Builtins operating on 128-bit integers, a library call is always used.
• -msahf — This enables generation of "SAHF" instructions in 64-bit code. Early Intel Pentium 4 CPUs with Intel 64 support, prior to the introduction of Pentium 4 G1 step in December 2005, lacked the "LAHF" and "SAHF" instructions which are supported by AMD64. These are load and store instructions, respectively, for certain status flags. In 64-bit mode, the "SAHF" instruction is used to optimize "fmod", "drem", and "remainder" built-in functions; see Other Builtins for details.
• -mmovbe — This enables use of the "movbe" instruction to optimize byte swapping of four and eight byte entities. 🔄
• -mshstk — The -mshstk option enables shadow stack built-in functions from x86 Control-flow Enforcement Technology (CET). 🛡️
• -mcrc32 — This enables built-in functions "__builtin_ia32_crc32qi", "__builtin_ia32_crc32hi", "__builtin_ia32_crc32si" and "__builtin_ia32_crc32di" to generate the "crc32" machine instruction.
• -mmwait — This enables built-in functions "__builtin_ia32_monitor", and "__builtin_ia32_mwait" to generate the "monitor" and "mwait" machine instructions.
• -mrecip — This enables use of "RCPSS" and "RSQRTSS" instructions (and their vectorized variants "RCPPS" and "RSQRTPS") with an additional Newton-Raphson step to increase precision instead of "DIVSS" and "SQRTSS" (and their vectorized variants) for single-precision floating-point arguments. These instructions are generated only when -funsafe-math-optimizations is enabled together with -ffinite-math-only and -fno-trapping-math. Note that while the throughput of the sequence is higher than the throughput of the non-reciprocal instruction, the precision of the sequence can be decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994). 😬

Note that GCC implements 1.0f/sqrtf(x) in terms of "RSQRTSS" (or "RSQRTPS") already with -ffast-math (or the above option combination), and doesn’t need -mrecip.

Also note that GCC emits the above sequence with additional Newton-Raphson step for vectorized single-float division and vectorized sqrtf(x) already with -ffast-math (or the above option combination), and doesn’t need -mrecip.

-mrecip=opt 🎯

okay so this one is the boss of which reciprocal estimate instructions are allowed to show up. opt is a comma-separated list of options, and any one of them can be prefixed with a ! to invert it (i.e. turn that one off):

• all — enable all the estimate instructions. the whole menu, bestie.
• default — enable the default instructions, which is exactly the same vibe as plain -mrecip.
• none — disable all of them, equivalent to -mno-recip. nothing gets approximated, periodt.
• div — turn on the approximation for scalar division.
• vec-div — turn on the approximation for vectorized division.
• sqrt — turn on the approximation for scalar square root.
• vec-sqrt — turn on the approximation for vectorized square root.

so for example, -mrecip=all,!sqrt enables all of the reciprocal approximations except for square root. see how the ! flips just that one? iconic behaviour honestly 💅

---

-mveclibabi=type 📚

this one picks the ABI type used for vectorizing intrinsics via an external library. type can be:

• svml — for the Intel short vector math library
• aocl — for the math library (LibM) from AMD Optimizing CPU Libraries (AOCL)
• acml — for the end-of-life AMD core math library (and just so you know, AOCL-LibM is its successor 👑)

to actually use this option, both -ftree-vectorize and -funsafe-math-optimizations have to be enabled, and an SVML- or ACML-ABI-compatible library has to be specified at link time. no library, no party.

GCC currently emits calls to these when -mveclibabi=svml is used, for the corresponding function type:

"vmldExp2", "vmldLn2", "vmldLog102", "vmldPow2", "vmldTanh2", "vmldTan2",
"vmldAtan2", "vmldAtanh2", "vmldCbrt2", "vmldSinh2", "vmldSin2",
"vmldAsinh2", "vmldAsin2", "vmldCosh2", "vmldCos2", "vmldAcosh2",
"vmldAcos2", "vmlsExp4", "vmlsLn4", "vmlsLog104", "vmlsPow4", "vmlsTanh4",
"vmlsTan4", "vmlsAtan4", "vmlsAtanh4", "vmlsCbrt4", "vmlsSinh4",
"vmlsSin4", "vmlsAsinh4", "vmlsAsin4", "vmlsCosh4", "vmlsCos4",
"vmlsAcosh4" and "vmlsAcos4"

and these when -mveclibabi=aocl is used, for the corresponding function type:

"amd_vrs4_acosf", "amd_vrs16_acosf", "amd_vrd8_asin", "amd_vrs4_asinf",
"amd_vrs8_asinf", "amd_vrs16_asinf", "amd_vrd2_atan", "amd_vrd8_atan",
"amd_vrs4_atanf", "amd_vrs8_atanf", "amd_vrs16_atanf", "amd_vrd2_cos",
"amd_vrd4_cos", "amd_vrd8_cos", "amd_vrs4_cosf", "amd_vrs8_cosf",
"amd_vrs16_cosf", "amd_vrs4_coshf", "amd_vrs8_coshf", "amd_vrd2_erf",
"amd_vrd4_erf", "amd_vrd8_erf", "amd_vrs4_erff", "amd_vrs8_erff",
"amd_vrs16_erff", "amd_vrd2_exp", "amd_vrd4_exp", "amd_vrd8_exp",
"amd_vrs4_expf", "amd_vrs8_expf", "amd_vrs16_expf", "amd_vrd2_exp10",
"amd_vrs4_exp10f", "amd_vrd2_exp2", "amd_vrd4_exp2", "amd_vrd8_exp2",
"amd_vrs4_exp2f", "amd_vrs8_exp2f", "amd_vrs16_exp2f", "amd_vrs4_expm1f",
"amd_vrd2_log", "amd_vrd4_log", "amd_vrd8_log", "amd_vrs4_logf",
"amd_vrs8_logf", "amd_vrs16_logf", "amd_vrd2_log10", "amd_vrs4_log10f",
"amd_vrs8_log10f", "amd_vrs16_log10f", "amd_vrd2_log1p", "amd_vrs4_log1pf",
"amd_vrd2_log2", "amd_vrd4_log2", "amd_vrd8_log2", "amd_vrs4_log2f",
"amd_vrs8_log2f", "amd_vrs16_log2f", "amd_vrd2_pow", "amd_vrd4_pow",
"amd_vrd8_pow", "amd_vrs4_powf", "amd_vrs8_powf", "amd_vrs16_powf",
"amd_vrd2_sin", "amd_vrd4_sin", "amd_vrd8_sin", "amd_vrs4_sinf",
"amd_vrs8_sinf", "amd_vrs16_sinf", "amd_vrd2_tan", "amd_vrd4_tan",
"amd_vrd8_tan", "amd_vrs16_tanf", "amd_vrs4_tanhf", "amd_vrs8_tanhf",
"amd_vrs16_tanhf"

and finally these when -mveclibabi=acml is used, for the corresponding function type:

"__vrd2_sin", "__vrd2_cos", "__vrd2_exp", "__vrd2_log", "__vrd2_log2",
"__vrd2_log10", "__vrs4_sinf", "__vrs4_cosf", "__vrs4_expf",
"__vrs4_logf", "__vrs4_log2f", "__vrs4_log10f" and "__vrs4_powf"

like. that's a LOT of names but every single one of them is chilling there in the docs so we keep them all 🫶

---

-mabi=name 🔀

generate code for the specified calling convention. the permissible values are:

• sysv — the ABI used on GNU/Linux and other systems
• ms — the Microsoft ABI

the default is to use the Microsoft ABI when targeting Microsoft Windows, and the SysV ABI on all other systems. you can also control this behaviour for specific functions by using the function attributes ms_abi and sysv_abi. mix and match as needed, bestie ✨

---

-mforce-indirect-call ☎️

force all calls to functions to be indirect. this is useful when you're using Intel Processor Trace, where it generates more precise timing information for function calls. every call goes the long way round, on purpose 👑

---

-mmanual-endbr 🚪

insert an ENDBR instruction at function entry only via the cf_check function attribute. this is useful when used together with the option -fcf-protection=branch, so you can control ENDBR insertion at the function entry. no random ENDBRs, only the ones you asked for 💅

---

-mcet-switch 🎚️

by default, CET instrumentation is turned off on switch statements that use a jump table, and indirect branch track is disabled. since jump tables are stored in read-only memory, this does not result in a direct loss of hardening. but — if the jump table index is attacker-controlled, the indirect jump may not be constrained by CET 😭

this option turns on CET instrumentation to enable indirect branch track for switch statements with jump tables, which leads to the jump targets being reachable via any indirect jumps. safety first, fr fr.

---

-mcall-ms2sysv-xlogues 🌉

due to differences in 64-bit ABIs, any Microsoft ABI function that calls a System V ABI function must consider RSI, RDI and XMM6-15 as clobbered. by default, the code for saving and restoring these registers is emitted inline, which gives you fairly lengthy prologues and epilogues 🥴

using -mcall-ms2sysv-xlogues emits prologues and epilogues that use stubs in the static portion of libgcc to do those saves and restores instead — so you get reduced function size, at the cost of a few extra instructions. trade-off city.

---

-mtls-dialect=type 🧵

generate code to access thread-local storage using the gnu or gnu2 conventions.

• gnu — the conservative default
• gnu2 — more efficient, but it may add compile- and run-time requirements that cannot be satisfied on all systems

so like, gnu2 is the glow-up but it won't work everywhere. pick wisely 🌸

---

-mpush-args, -mno-push-args 📥

use PUSH operations to store outgoing parameters. this method is shorter and usually equally fast as the method using SUB/MOV operations, and it's enabled by default. in some cases disabling it may improve performance, because of improved scheduling and reduced dependencies. not me messing with the scheduler for fun 🙈

---

-maccumulate-outgoing-args 📦

if enabled, the maximum amount of space required for outgoing arguments is computed in the function prologue. this is faster on most modern CPUs, because of reduced dependencies, improved scheduling, and reduced stack usage when the preferred stack boundary is not equal to 2.

the drawback? a notable increase in code size 😭. also, this switch implies -mno-push-args. noted and filed.

---

-mthreads 🧷

support thread-safe exception handling on MinGW. programs that rely on thread-safe exception handling must compile and link all code with the -mthreads option.

when compiling, -mthreads defines -D_MT; when linking, it links in a special thread helper library -lmingwthrd, which cleans up per-thread exception-handling data. all or nothing, bestie 💜

---

-mms-bitfields, -mno-ms-bitfields 🧩

enable/disable bit-field layout compatible with the native Microsoft Windows compiler.

if packed is used on a structure, or if bit-fields are used, it may be that the Microsoft ABI lays out the structure differently than the way GCC normally does. particularly when moving packed data between functions compiled with GCC and the native Microsoft compiler (either via a function call or as data in a file), it may be necessary to access either format. 👀

this option is enabled by default for Microsoft Windows targets. this behaviour can also be controlled locally by use of variable or type attributes. for more information, see x86 Variable Attributes and x86 Type Attributes.

the Microsoft structure layout algorithm is fairly simple — except for the bit-field packing part 💀. the padding and alignment of members of structures, and whether a bit-field can straddle a storage-unit boundary, are determined by these rules:

1. Structure members are stored sequentially in the order in which they are declared: the first member has the lowest memory address and the last member the highest.
2. Every data object has an alignment requirement. The alignment requirement for all data except structures, unions, and arrays is either the size of the object or the current packing size (specified with either the aligned attribute or the pack pragma), whichever is less. For structures, unions, and arrays, the alignment requirement is the largest alignment requirement of its members. Every object is allocated an offset so that:

offset % alignment_requirement == 0

3. Adjacent bit-fields are packed into the same 1-, 2-, or 4-byte allocation unit if the integral types are the same size and if the next bit-field fits into the current allocation unit without crossing the boundary imposed by the common alignment requirements of the bit-fields.

MSVC interprets zero-length bit-fields in the following ways:

1. If a zero-length bit-field is inserted between two bit-fields that are normally coalesced, the bit-fields are not coalesced.

for example:

struct
 {
   unsigned long bf_1 : 12;
   unsigned long : 0;
   unsigned long bf_2 : 12;
 } t1;

the size of t1 is 8 bytes with the zero-length bit-field. if the zero-length bit-field were removed, t1's size would be 4 bytes. tiny field, big consequences 😭

2. If a zero-length bit-field is inserted after a bit-field, foo, and the alignment of the zero-length bit-field is greater than the member that follows it, bar, bar is aligned as the type of the zero-length bit-field.

for example:

struct
 {
   char foo : 4;
   short : 0;
   char bar;
 } t2;

struct
 {
   char foo : 4;
   short : 0;
   double bar;
 } t3;

for t2, bar is placed at offset 2, rather than offset 1. accordingly, the size of t2 is 4. for t3, the zero-length bit-field does not affect the alignment of bar or, as a result, the size of the structure.

taking this into account, it is important to note the following:

1. If a zero-length bit-field follows a normal bit-field, the type of that zero-length bit-field can mess with the alignment of the whole struct. Like, t2 is 4 bytes big, because the zero-length bit-field follows a normal bit-field and it's of type short. the vibes of t2 are large 💖
2. And even if a zero-length bit-field has no normal bit-field after it, it can still affect the alignment of the struct:

struct
 {
   char foo : 6;
   long : 0;
 } t4;

Here t4 takes up 4 bytes. it said "I'm built different" ✨

3. Zero-length bit-fields that follow non-bit-field members are just ignored:

struct
 {
   char foo;
   long : 0;
   char bar;
 } t5;

Here t5 takes up 2 bytes. tiny queen 🎀

---

• -mno-align-stringops — do NOT align the destination of inlined string ops. This one shrinks your code size and speeds things up when the destination is already aligned but GCC has no idea about it. gcc doesn't know bestie, we know 🙈
• -minline-all-stringops — by default GCC only inlines string operations when the destination is known to be aligned to at least a 4-byte boundary. This turns up the inlining and grows your code size, but it may make code faster when it depends on fast memcpy and memset for short lengths. This option also enables inline expansion of strlen for all pointer alignments. we inline everything here, no survivors 💅
• -minline-stringops-dynamically — for string operations of unknown size, use run-time checks with inline code for small blocks and a library call for large blocks. it's giving "let's decide at runtime" 🦋
• -mstringop-strategy=alg — override the internal decision heuristic for which algorithm gets used to inline string operations. The allowed values for alg are:
  • rep_byte
  • rep_4byte
  • rep_8byte

Expand using the i386 rep prefix of the specified size.

• byte_loop
• loop
• unrolled_loop

Expand into an inline loop.

• libcall

Always use a library call.

• -mmemcpy-strategy=strategy — override the internal decision heuristic that decides whether __builtin_memcpy should be inlined, and which inline algorithm to use when the expected size of the copy is known. strategy is a comma-separated list of alg:max_size:dest_align triplets. alg is specified in -mstringop-strategy, max_size specifies the max byte size for which inline algorithm alg is allowed. For the last triplet, the max_size must be -1. The max_size of the triplets in the list must be specified in increasing order. The minimal byte size for alg is 0 for the first triplet and "max_size + 1" of the preceding range. math homework but make it compilers 😭
• -mmemset-strategy=strategy — same idea as -mmemcpy-strategy= except this one controls __builtin_memset expansion. memset's turn to shine ✨
• -momit-leaf-frame-pointer — don't keep the frame pointer in a register for leaf functions. This dodges the instructions to save, set up, and restore frame pointers and frees up an extra register in leaf functions. Heads up: the option -fomit-leaf-frame-pointer removes the frame pointer for leaf functions, which might make debugging harder. debugging found crying in the corner 💀
• -mtls-direct-seg-refs
• -mno-tls-direct-seg-refs — controls whether TLS variables may be accessed with offsets from the TLS segment register (%gs for 32-bit, %fs for 64-bit), or whether the thread base pointer must be added instead. Whether this is even valid depends on the operating system, and on whether it maps the segment to cover the entire TLS area.

For systems that use the GNU C Library, the default is on. glibc stays winning 👑

• -msse2avx
• -mno-sse2avx — specifies that the assembler should encode SSE instructions with a VEX prefix. The option -mavx turns this on by default.
• -mfentry
• -mno-fentry — if profiling is active (-pg), put the profiling counter call before the prologue. Note: on x86 architectures the attribute ms_hook_prologue isn't possible at the moment for -mfentry and -pg. not possible bestie, not today 🙈
• -mrecord-mcount
• -mno-record-mcount — if profiling is active (-pg), generate a __mcount_loc section that contains pointers to each profiling call. This is useful for automatically patching and out calls.
• -mnop-mcount
• -mno-nop-mcount — if profiling is active (-pg), generate the calls to the profiling functions as NOPs. This is useful when they're supposed to be patched in later dynamically. This is likely only useful together with -mrecord-mcount. dynamic patching girlies rise 🌸
• -minstrument-return=type — instrument function exit in -pg -mfentry instrumented functions with a call to the specified function. This only instruments true returns ending with ret, but not sibling calls ending with jump. Valid types are none to not instrument, call to generate a call to __return__, or nop5 to generate a 5 byte nop.
• -mrecord-return
• -mno-record-return — generate a __return_loc section pointing to all return instrumentation code.
• -mfentry-name=name — set the name of the __fentry__ symbol called at function entry for -pg -mfentry functions.
• -mfentry-section=name — set the name of the section to record -mrecord-mcount calls (default __mcount_loc).
• -mskip-rax-setup
• -mno-skip-rax-setup — when generating code for the x86-64 architecture with SSE extensions disabled, -mskip-rax-setup can be used to skip setting up the RAX register when there are no variable arguments passed in vector registers.

Warning: since the RAX register is used to avoid unnecessarily saving vector registers on the stack when passing variable arguments, the impacts of this option are that callees may waste some stack space, misbehave or jump to a random location. GCC 4.4 or newer doesn't have those issues, regardless of the RAX register value. do not do this on old gcc, I am BEGGING 😭

• -m8bit-idiv
• -mno-8bit-idiv — on some processors, like Intel Atom, 8-bit unsigned integer divide is way faster than 32-bit/64-bit integer divide. This option generates a run-time check. If both the dividend and divisor are within the range of 0 to 255, an 8-bit unsigned integer divide gets used instead of the 32-bit/64-bit integer divide. small numbers, big speed 🔥
• -mavx256-split-unaligned-load
• -mavx256-split-unaligned-store — split 32-byte AVX unaligned load and store. snip snip ✂️
• -mstack-protector-guard=guard
• -mstack-protector-guard-reg=reg
• -mstack-protector-guard-offset=offset
• -mstack-protector-guard-symbol=symbol — generate stack protection code using a canary at guard. Supported locations are global for a global canary, or tls for a per-thread canary in the TLS block (the default). This option only has an effect when -fstack-protector or -fstack-protector-all is specified.

With the latter choice, the options -mstack-protector-guard-reg=reg and -mstack-protector-guard-offset=offset additionally specify which segment register (%fs or %gs) to use as the base register for reading the canary, and from what offset from that base register. The default for those is as specified in the relevant ABI.

-mstack-protector-guard-symbol=symbol overrides the offset with a symbol reference to a canary in the TLS block. canary but make it symbolic 🐤

• -mgeneral-regs-only — generate code that only uses the general-purpose registers. This stops the compiler from using floating-point, vector, mask and bound registers. no fancy registers allowed, only the OGs 💅
• -mrelax-cmpxchg-loop — when emitting a compare-and-swap loop for __sync Builtins and __atomic Builtins that lack a native instruction, optimize for the highly contended case by issuing an atomic load before the CMPXCHG instruction, and using the PAUSE instruction to save CPU power when restarting the loop. contended? pause. breathe. try again 🧘
• -mindirect-branch=choice — convert indirect call and jump with choice. The default is keep, which keeps indirect call and jump unmodified. thunk converts indirect call and jump to call and return thunk. thunk-inline converts indirect call and jump to inlined call and return thunk. thunk-extern converts indirect call and jump to an external call and return thunk provided in a separate object file. You can control this behavior for a specific function by using the function attribute indirect_branch.

Note that -mcmodel=large is incompatible with -mindirect-branch=thunk and -mindirect-branch=thunk-extern since the thunk function may not be reachable in the large code model. incompatible. do not. I said what I said.

Note that -mindirect-branch=thunk-extern is compatible with -fcf-protection=branch since the external thunk can be made to enable control-flow check.

• -mfunction-return=choice — convert function return with choice. The default is keep, which keeps function return unmodified. thunk converts function return to call and return thunk. thunk-inline converts function return to inlined call and return thunk. thunk-extern converts function return to an external call and return thunk provided in a separate object file. You can control this behavior for a specific function by using the function attribute function_return.

Note that -mindirect-return=thunk-extern is compatible with -fcf-protection=branch since the external thunk can be made to enable control-flow check.

Note that -mcmodel=large is incompatible with -mfunction-return=thunk and -mfunction-return=thunk-extern since the thunk function may not be reachable in the large code model.

• -mindirect-branch-register — force indirect call and jump via register. register route only, periodt
• -mharden-sls=choice — generate code to mitigate against straight line speculation (SLS) with choice. The default is none, which disables all SLS hardening. return enables SLS hardening for function returns. indirect-jmp enables SLS hardening for indirect jumps. all enables all SLS hardening. hardened and hot 🛡️
• -mindirect-branch-cs-prefix — add a CS prefix to call and jmp to indirect thunk with branch target in r8-r15 registers, so that the call and jmp instruction length is 6 bytes, letting them be replaced at run-time with lfence; call *%r8-r15 or lfence; jmp *%r8-r15. 6 bytes of pure opportunity ✨
• -mapx-inline-asm-use-gpr32 — for inline asm support with APX, by default the EGPR feature was disabled to prevent potential illegal instruction with EGPR from happening. To invoke egpr usage in inline asm, use the new compiler option -mapx-inline-asm-use-gpr32, and the user should ensure the instruction supports EGPR. you have to double-check that one yourself bestie, no cap 🫶
• -mevex512, -mno-evex512 — turns 512-bit vectors on or off. lowkey it's the default when AVX512F is enabled, no cap.

---

These -m switches are also supported on x86-64 processors when you're living that 64-bit environment life. ✨

• -m32, -m64, -mx32, -m16, -miamcu — generate code for a 16-bit, 32-bit or 64-bit environment.

So like, -m32 sets int, long and pointer types to 32 bits and spits out code that runs in 32-bit mode. iconic.

-m64 sets int to 32 bits and long and pointer types to 64 bits, and generates code for the x86-64 architecture. And bestie, Darwin-only detail: for Darwin, the -m64 option also turns off the -fno-pic and -mdynamic-no-pic options. 💅

-mx32 sets int, long, and pointer types to 32 bits, and generates code for the x86-64 architecture.

-m16 is literally -m32 but it also outputs the .code16gcc assembly directive at the beginning of the assembly output so the binary can run in 16-bit mode. cute.

-miamcu generates code which conforms to Intel MCU psABI. It requires -m32 to be turned on — no -m32, no deal. 🫶

• -mno-red-zone — do not use a so-called "red zone" for x86-64 code. The red zone is mandated by the x86-64 ABI: it's a 128-byte area beyond the location of the stack pointer that is not modified by signal or interrupt handlers, which means you can stash temporary data there without adjusting the stack pointer. -mno-red-zone disables this red zone. it's giving "please don't touch my little buffer" 🌸
• -mcmodel=small — generate code for the small code model: the program and its symbols must be linked in the lower 2 GB of the address space. Pointers are 64 bits. Programs can be statically or dynamically linked. This is the default code model, she's the main character. 👑
• -mcmodel=kernel — generate code for the kernel code model. The kernel runs in the negative 2 GB of the address space. This model has to be used for Linux kernel code. mandatory, fr fr.
• -mcmodel=medium — generate code for the medium model: the program is linked in the lower 2 GB of the address space. Small symbols are also placed there. Symbols with sizes larger than -mlarge-data-threshold get put into large data or BSS sections and can be located above 2GB. Programs can be statically or dynamically linked.
• -mcmodel=large — generate code for the large model. This model makes no assumptions about addresses and sizes of sections. zero assumptions, zero problems, highkey chaotic but valid. 💀
• -maddress-mode=long — generate code for long address mode. Only supported for 64-bit and x32 environments. It's the default address mode for 64-bit environments.
• -maddress-mode=short — generate code for short address mode. Only supported for 32-bit and x32 environments. It's the default address mode for 32-bit and x32 environments.
• -mneeded, -mno-needed — emit the GNU_PROPERTY_X86_ISA_1_NEEDED GNU property for Linux target to indicate the micro-architecture ISA level required to execute the binary. it's a lil receipt for what the binary needs 🔥
• -mno-direct-extern-access — without -fpic nor -fPIC, always use the GOT pointer to access external symbols. With -fpic or -fPIC, treat access to protected symbols as local symbols. The default is -mdirect-extern-access.

Warning: shared libraries compiled with -mno-direct-extern-access and executable compiled with -mdirect-extern-access may not be binary compatible if protected symbols are used in shared libraries and executable. bestie this is your sign to not mix and match. 😭

• -munroll-only-small-loops — controls conservative small loop unrolling. It's default enabled by O2, and unrolls loop with less than 4 insns by 1 time. Explicit -f[no-]unroll-[all-]loops would disable this flag to avoid any unintended unrolling behavior that user does not want. leave the loops alone unless you truly mean it. 🙈
• -mlam=choice — LAM (linear-address masking) allows special bits in the pointer to be used for metadata. The default is none. With u48, pointer bits in positions 62:48 can be used for metadata; with u57, pointer bits in positions 62:57 can be used for metadata. metadata girls stay winning. ✨

---

💻 x86 Windows Options 💻

(that's the heading, bestie — nothing else lives under it here 🎀)

---

🌈 Xstormy16 Options 🌈

These options are defined for Xstormy16:

• -msim — choose startup files and linker script suitable for the simulator. very demure, very simulator-coded. 🦋

---

🧚 Xtensa Options 🧚

These options are supported for Xtensa targets:

• -mconst16, -mno-const16 — enable or disable use of CONST16 instructions for loading constant values. The CONST16 instruction is currently not a standard option from Tensilica. When enabled, CONST16 instructions are always used in place of the standard L32R instructions. The use of CONST16 is enabled by default only if the L32R instruction is not available. conditional queen behavior, periodt.
• -mfused-madd, -mno-fused-madd — enable or disable use of fused multiply/add and multiply/subtract instructions in the floating-point option. This has no effect if the floating-point option is not also enabled. Disabling fused multiply/add and multiply/subtract instructions forces the compiler to use separate instructions for the multiply and add/subtract operations. This may be desirable in some cases where strict IEEE 754-compliant results are required: the fused multiply add/subtract instructions do not round the intermediate result, thereby producing results with more bits of precision than specified by the IEEE standard. Disabling fused multiply add/subtract instructions also ensures that the program output is not sensitive to the compiler's ability to combine multiply and add/subtract operations. precision girlies, this one's for you 💜
• -mserialize-volatile, -mno-serialize-volatile — when this option is enabled, GCC inserts MEMW instructions before volatile memory references to guarantee sequential consistency. The default is -mserialize-volatile. Use -mno-serialize-volatile to omit the MEMW instructions.
• -mforce-no-pic — for targets, like GNU/Linux, where all user-mode Xtensa code must be position-independent code (PIC), this option disables PIC for compiling kernel code. not me enabling chaos for kernel purposes 💀
• -mtext-section-literals, -mno-text-section-literals — these options control the treatment of literal pools. The default is -mno-text-section-literals, which places literals in a separate section in the output file. This allows the literal pool to be placed in a data RAM/ROM, and it also allows the linker to combine literal pools from separate object files to remove redundant literals and improve code size. With -mtext-section-literals, the literals are interspersed in the text section in order to keep them as close as possible to their references. This may be necessary for large assembly files. Literals for each function are placed right before that function. ✨
• -mauto-litpools, -mno-auto-litpools — these options control the treatment of literal pools. The default is -mno-auto-litpools, which places literals in a separate section in the output file unless -mtext-section-literals is used. With -mauto-litpools the literals are interspersed in the text section by the assembler. Compiler does not produce explicit .literal directives and loads literals into registers with MOVI instructions instead of L32R to let the assembler do relaxation and place literals as necessary. This option allows assembler to create several literal pools per function and assemble very big functions, which may not be possible with -mtext-section-literals. big-function energy only, iykyk. 🔥
• -mtarget-align, -mno-target-align — when this option is enabled, GCC instructs the assembler to automatically align instructions to reduce branch penalties at the expense of some code density. The assembler attempts to widen density instructions to align branch targets and the instructions following call instructions. If there are not enough preceding safe density instructions to align a target, no widening is performed. The default is -mtarget-align. These options do not affect the treatment of auto-aligned instructions like LOOP, which the assembler always aligns, either by widening density instructions or by inserting NOP instructions. LOOP does what it wants, always. 🎀
• -mlongcalls, -mno-longcalls — when this option is enabled, GCC instructs the assembler to translate direct calls to indirect calls unless it can determine that the target of a direct call is in the range allowed by the call instruction. This translation typically occurs for calls to functions in other source files. Specifically, the assembler translates a direct CALL instruction into an L32R followed by a CALLX instruction. The default is -mno-longcalls. This option should be used in programs where the call target can potentially be out of range. This option is implemented in the assembler, not the compiler, so the assembly code generated by GCC still shows direct call instructions — look at the disassembled object code to see the actual instructions. Note that the assembler uses an indirect call for every cross-file call, not just those that really are out of range. she's doing the most for cross-file calls ngl 😭
• -mabi=name — generate code for the specified ABI. Permissible values are: call0, windowed. Default ABI is chosen by the Xtensa core configuration.
• -mabi=call0 — when this option is enabled function parameters are passed in registers a2 through a7, registers a12 through a15 are caller-saved, and register a15 may be used as a frame pointer. When this version of the ABI is enabled, the C preprocessor symbol __XTENSA_CALL0_ABI__ is defined. 💖
• -mabi=windowed — when this option is enabled function parameters are passed in registers a10 through a15, and called function rotates register window by 8 registers on entry so that its arguments are found in registers a2 through a7. Register a7 may be used as a frame pointer. Register window is rotated 8 registers back upon return. When this version of the ABI is enabled, the C preprocessor symbol __XTENSA_WINDOWED_ABI__ is defined. rotating her window like a true main character ✨
• -mextra-l32r-costs=n — specify an extra cost of instruction RAM/ROM access for L32R instructions, in clock cycles. This affects, when optimizing for speed, whether loading a constant from literal pool using L32R or synthesizing the constant from a small one with a couple of arithmetic instructions. The default value is 0. 🌸
• -mstrict-align, -mno-strict-align — okay so this is a ✨vibe check✨ for your memory accesses: do you wanna avoid generating memory accesses that might not land on a natural object boundary as described in the architecture specification, or do you wanna allow it? -mstrict-align = avoid, -mno-strict-align = allow, bestie, it's giving boundaries 💅.

The default is -mno-strict-align for cores that support both unaligned loads and stores in hardware, and -mstrict-align for all other cores. So like, if your chip can already do the unaligned thing, you don't need the strictness, periodt 🔥.

---

🌸 zSeries Options 🌸

These are listed under 🦋

---

💖 ENVIRONMENT 💖

Okay so this whole section is about the environment variables that change how GCC does its little thing. Some of them work by specifying directories or prefixes to use when searching for various kinds of files. Some are used to specify other aspects of the compilation environment. Iconic behavior honestly 👑

Not me reminding you that you can also specify places to search using options such as -B, -I and -L. These take precedence over places specified using environment variables, which in turn take precedence over those specified by the configuration of GCC. So like: options > env vars > whatever GCC was configured with. Remember the order bestie, it matters fr fr 😭

🌈 Localization girlies

• LANG
• LC_CTYPE
• LC_MESSAGES
• LC_ALL

These environment variables control the way that GCC uses localization information, which is the fancy way of saying it lets GCC work with different national conventions 🫶. GCC inspects the locale categories LC_CTYPE and LC_MESSAGES if it has been configured to do so. These locale categories can be set to any value supported by your installation. A typical value is en_GB.UTF-8 for English in the United Kingdom encoded in UTF-8 ✨.

The LC_CTYPE environment variable specifies character classification. GCC uses it to determine the character boundaries in a string; this is needed for some multibyte encodings that contain quote and escape characters that are otherwise interpreted as a string end or escape. Lowkey crucial tbh.

The LC_MESSAGES environment variable specifies the language to use in diagnostic messages 🗣️.

If the LC_ALL environment variable is set, it overrides the value of LC_CTYPE and LC_MESSAGES; otherwise, LC_CTYPE and LC_MESSAGES default to the value of the LANG environment variable. If none of these variables are set, GCC defaults to traditional C English behavior 💀.

🍓 TMPDIR

If TMPDIR is set, it specifies the directory to use for temporary files. GCC uses temporary files to hold the output of one stage of compilation which is to be used as input to the next stage: for example, the output of the preprocessor, which is the input to the compiler proper. It's giving ✨pipeline✨ no cap.

🎀 GCC_COMPARE_DEBUG

Setting GCC_COMPARE_DEBUG is nearly equivalent to passing -fcompare-debug to the compiler driver. See the documentation of this option for more details 👀.

🧚 GCC_EXEC_PREFIX

If GCC_EXEC_PREFIX is set, it specifies a prefix to use in the names of the subprograms executed by the compiler. No slash is added when this prefix is combined with the name of a subprogram, but you can specify a prefix that ends with a slash if you wish 💅.

If GCC_EXEC_PREFIX is not set, GCC attempts to figure out an appropriate prefix to use based on the pathname it is invoked with.

If GCC cannot find the subprogram using the specified prefix, it tries looking in the usual places for the subprogram.

The default value of GCC_EXEC_PREFIX is prefix/lib/gcc/ where prefix is the prefix to the installed compiler. In many cases prefix is the value of "prefix" when you ran the configure script.

Other prefixes specified with -B take precedence over this prefix.

This prefix is also used for finding files such as crt0.o that are used for linking.

In addition, the prefix is used in an unusual way in finding the directories to search for header files. For each of the standard directories whose name normally begins with /usr/local/lib/gcc (more precisely, with the value of GCC_INCLUDE_DIR), GCC tries replacing that beginning with the specified prefix to produce an alternate directory name. Thus, with -Bfoo/, GCC searches foo/bar just before it searches the standard directory /usr/local/lib/bar. If a standard directory begins with the configured prefix then the value of prefix is replaced by GCC_EXEC_PREFIX when looking for header files. Wild right?? 😳

💜 COMPILER_PATH

The value of COMPILER_PATH is a colon-separated list of directories, much like PATH. GCC tries the directories thus specified when searching for subprograms, if it cannot find the subprograms using GCC_EXEC_PREFIX.

🌸 LIBRARY_PATH

The value of LIBRARY_PATH is a colon-separated list of directories, much like PATH. When configured as a native compiler, GCC tries the directories thus specified when searching for special linker files, if it cannot find them using GCC_EXEC_PREFIX. Linking using GCC also uses these directories when searching for ordinary libraries for the -l option (but directories specified with -L come first). Order girlies, order 😌

---

🌈 LANG (again, she's back)

This variable is used to pass locale information to the compiler. One way in which this information is used is to determine the character set to be used when character literals, string literals and comments are parsed in C and C++. When the compiler is configured to allow multibyte characters, the following values for LANG are recognized:

• C-JIS — Recognize JIS characters.
• C-SJIS — Recognize SJIS characters.
• C-EUCJP — Recognize EUCJP characters.

If LANG is not defined, or if it has some other value, then the compiler uses "mblen" and "mbtowc" as defined by the default locale to recognize and translate multibyte characters. Default behavior it is 🙈

---

⭐ GCC_EXTRA_DIAGNOSTIC_OUTPUT

If GCC_EXTRA_DIAGNOSTIC_OUTPUT is set to one of the following values, then additional text will be emitted to stderr when fix-it hints are emitted. -fdiagnostics-parseable-fixits and -fno-diagnostics-parseable-fixits take precedence over this environment variable 👑 (them over her, always).

• fixits-v1 — Emit parseable fix-it hints, equivalent to -fdiagnostics-parseable-fixits. In particular, columns are expressed as a count of bytes, starting at byte 1 for the initial column.
• fixits-v2 — As "fixits-v1", but columns are expressed as display columns, as per -fdiagnostics-column-unit=display.

---

Some additional environment variables affect the behavior of the preprocessor. Slay, more of them 💖

🦋 CPATH / C_INCLUDE_PATH / CPLUS_INCLUDE_PATH / OBJC_INCLUDE_PATH

• CPATH
• C_INCLUDE_PATH
• CPLUS_INCLUDE_PATH
• OBJC_INCLUDE_PATH

Each variable's value is a list of directories separated by a special character, much like PATH, in which to look for header files. The special character, "PATH_SEPARATOR", is target-dependent and determined at GCC build time. For Microsoft Windows-based targets it is a semicolon, and for almost all other targets it is a colon. iykyk 🫶

CPATH specifies a list of directories to be searched as if specified with -I, but after any paths given with -I options on the command line. This environment variable is used regardless of which language is being preprocessed.

The remaining environment variables apply only when preprocessing the particular language indicated. Each specifies a list of directories to be searched as if specified with -isystem, but after any paths given with -isystem options on the command line.

In all these variables, an empty element instructs the compiler to search its current working directory. Empty elements can appear at the beginning or end of a path. For instance, if the value of CPATH is :/special/include, that has the same effect as -I. -I/special/include. Sneaky little empty element, iconic 💅

---

🎀 DEPENDENCIES_OUTPUT

If this variable is set, its value specifies how to output dependencies for Make based on the non-system header files processed by the compiler. System header files are ignored in the dependency output. Not the system girls, they're not invited 💀

The value of DEPENDENCIES_OUTPUT can be just a file name, in which case the Make rules are written to that file, guessing the target name from the source file name. Or the value can have the form file target, in which case the rules are written to file file using target as the target name.

In other words, this environment variable is equivalent to combining the options -MM and -MF, with an optional -MT switch too. Combo move ✨

---

🍓 SUNPRO_DEPENDENCIES

This variable is the same as DEPENDENCIES_OUTPUT (see above), except that system header files are not ignored, so it implies -M rather than -MM. However, the dependence on the main input file is omitted. So close to her sister but with ✨main character differences✨

---

🌟 SOURCE_DATE_EPOCH

If this variable is set, its value specifies a UNIX timestamp to be used in replacement of the current date and time in the "__DATE__" and "__TIME__" macros, so that the embedded timestamps become reproducible. Reproducible builds girlies, we love to see it 🌈

The value of SOURCE_DATE_EPOCH must be a UNIX timestamp, defined as the number of seconds (excluding leap seconds) since 01 Jan 1970 00:00:00 represented in ASCII; identical to the output of "date +%s" on GNU/Linux and other systems that support the %s extension in the "date" command.

The value should be a known timestamp such as the last modification time of the source or package and it should be set by the build process. Don't just make one up bestie 😭

---

💀 BUGS 💀

For instructions on reporting bugs, see https://gcc.gnu.org/bugs/. Go tell them everything, they wanna hear it 🫶

---

🧚 FOOTNOTES 🧚

1. On some systems, gcc -shared needs to build supplementary stub code for constructors to work. On multi-libbed systems, gcc -shared must select the correct support libraries to link against. Failing to supply the correct flags may lead to subtle defects ⚠️. Supplying them in cases where they are not necessary is innocuous. -shared suppresses the addition of startup code to alter the floating-point environment as done with -ffast-math, -Ofast or -funsafe-math-optimizations on some targets. Okay so this one's highkey important, don't skip it 📌

---

🦋 SEE ALSO 🦋

gpl(7), gfdl(7), fsf-funding(7), cpp(1), gcov(1), as(1), ld(1), gdb(1) and the Info entries for gcc, cpp, as, ld, binutils and gdb. The whole friend group, periodt 👑

---

👑 AUTHOR 👑

See the Info entry for gcc, or https://gcc.gnu.org/onlinedocs/gcc/Contributors.html, for contributors to GCC. Thank you to the contributors, they ate and left no crumbs 💖✨

💖 COPYRIGHT 💖

okay bestie, legal era, but make it cute. here's the receipts:

Copyright (c) 1988-2025 Free Software Foundation, Inc.

and the vibe is: you're totally allowed to copy, distribute and/or modify this document 🫶 — under the terms of the GNU Free Documentation License, Version 1.3 or any later version published by the Free Software Foundation. no cap, that's the actual license.

but — and this is the part where you gotta pay attention, bestie — there are some conditions attached, and they are not optional:

• the Invariant Sections are "GNU General Public License" and "Funding Free Software"
• the Front-Cover texts are (a) (peep below 👇)
• the Back-Cover texts are (b) (also below 👇)

and if you want the whole license text, a copy of it is literally included in the gfdl(7) man page. go read it, it's giving ✨ thorough ✨.

---

🎀 (a) The FSF's Front-Cover Text 🎀

and it goes a lil something like this:

A GNU Manual

short. iconic. periodt.

---

🎀 (b) The FSF's Back-Cover Text 🎀

and the back cover said:

You have freedom to copy and modify this GNU Manual, like GNU
software.  Copies published by the Free Software Foundation raise
funds for GNU development.

so like... yes, you can copy and remix the manual, same energy as GNU software itself 🌈 — and every copy the FSF publishes is out here funding GNU development. that's the whole vibe. slay.

---

and that's a wrap on the whole entire page, bestie 😭💅 you read all 132 parts of it, which is genuinely main character behavior, no notes. this was gcc-15, dated 2026-09-15, and it lives in GCC(1). go compile something unhinged. mwah 🎀✨🦋