LLVM-MC(1)

📦 version 12.0.1, in ubuntu jammy

💖 LLVM-MC(1) 💖

okay bestie, welcome to the manual page for llvm-mc 12 — yes it says "User Commands" up top, that's the section, we move 💅

✨ NAME ✨

llvm-mc — the manual page for llvm-mc 12 🎀

🌈 DESCRIPTION 🌈

OVERVIEW: the LLVM machine code playground 🛝✨

USAGE:

llvm-mc [options] <input file>

OPTIONS: (there are a LOT, buckle up bestie, this is the whole menu) 🍓

🎨 Color Options 🎨

--color — use colors in the output, obviously, we're not animals 💅 (default is autodetect, so it figures out the vibe itself)

⚙️ General options ⚙️

-I=<directory> — directory where include files live 📁
--aarch64-neon-syntax=<value> — pick which style of NEON code the AArch64 backend should emit:
- =generic — emit generic NEON assembly
- =apple — emit Apple-style NEON assembly 🍎
--abort-on-max-devirt-iterations-reached — bail out (abort) once the max iterations for the devirtualization CGSCC repeat pass has been hit 💀
--amdgpu-bypass-slow-div — skip the 64-bit divide for dynamic 32-bit values
--amdgpu-disable-loop-alignment — don't align and prefetch loops, just leave them be 🙈
--amdgpu-disable-power-sched — turn off scheduling that exists to minimize mAI power bursts
--amdgpu-dpp-combine — turn on the DPP combiner
--amdgpu-dump-hsa-metadata — dump the AMDGPU HSA metadata
--amdgpu-enable-flat-scratch — use flat scratch instructions
--amdgpu-enable-merge-m0 — merge and hoist M0 initializations
--amdgpu-promote-alloca-to-vector-limit=<uint> — the maximum byte size to even consider for promote alloca to vector
--amdgpu-reserve-vgpr-for-sgpr-spill — allocates one VGPR for a future SGPR spill
--amdgpu-sdwa-peephole — turn on the SDWA peepholer 👀
--amdgpu-use-aa-in-codegen — enable using AA during codegen
--amdgpu-verify-hsa-metadata — verify the AMDGPU HSA metadata
--amdgpu-vgpr-index-mode — use GPR indexing mode instead of movrel for vector indexing
--arch=<string> — the target arch to assemble for, and yes you can check -version to see which targets are actually available 🎯
--arm-add-build-attributes — (no description in the source at all, it's giving mystery 🕵️♀️)
--arm-implicit-it=<value> — allow conditional instructions outside of an IT block:
- =always — accept in both ISAs, emit implicit ITs in Thumb
- =never — warn in ARM, reject in Thumb
- =arm — accept in ARM, reject in Thumb
- =thumb — warn in ARM, emit implicit ITs in Thumb
--asm-show-inst — emit the internal instruction representation into the assembly file 🫶

🎬 Action to perform:

--as-lex — lex tokens from a .s file
--assemble — assemble a .s file (this is the default, bestie, the main character 👑)
--disassemble — disassemble strings of hex bytes
--mdis — marked-up disassembly of strings of hex bytes ✨
--atomic-counter-update-promoted — do the counter update with atomic fetch add, but only for promoted counters
--atomic-first-counter — use atomic fetch add for the first counter in a function (which is usually the entry counter)
--bounds-checking-single-trap — use one trap block per function
--cfg-hide-deoptimize-paths — (again, no description given, just vibes 🦋)
--cfg-hide-unreachable-paths — (also undescribed, we simply don't know 🙈)
--compress-debug-sections=<value> — choose how DWARF debug sections get compressed:
- =none — no compression
- =zlib — use zlib compression
- =zlib-gnu — use zlib-gnu compression (deprecated, so like… she's retired 💀)
--cost-kind=<value> — which target cost kind you want:
- =throughput — reciprocal throughput
- =latency — instruction latency
- =code-size — code size
- =size-latency — code size and latency
--cvp-dont-add-nowrap-flags — (no description provided, it's mysterious and that's fine 🎀)
--debugify-level=<value> — what kind of debug info to add:
- =locations — locations only
- =location+variables — locations and variables
--debugify-quiet — suppress the verbose debugify output, shhh 🤫
--defsym=<string> — defines a symbol to be an integer constant
--disable-promote-alloca-to-lds — disable promote alloca to LDS
--disable-promote-alloca-to-vector — disable promote alloca to vector
--do-counter-promotion — do counter register promotion
--dot-cfg-mssa=<file name for generated dot file> — the file name for the generated dot file
--dwarf-version=<int> — the Dwarf version
--dwarf64 — generate debugging info in the 64-bit DWARF format
--emscripten-cxx-exceptions-allowed=<string> — the list of function names where Emscripten-style exception handling is enabled (peep the emscripten EMSCRIPTEN_CATCHING_ALLOWED options)
--enable-cse-in-irtranslator — should enable CSE in irtranslator
--enable-cse-in-legalizer — should enable CSE in Legalizer
--enable-emscripten-cxx-exceptions — WebAssembly Emscripten-style exception handling
--enable-emscripten-sjlj — WebAssembly Emscripten-style setjmp/longjmp handling
--enable-gvn-hoist — enable the GVN hoisting pass (default = off)
--enable-gvn-memdep — (no description, it's a secret 🤐)
--enable-gvn-sink — enable the GVN sinking pass (default = off)
--enable-load-in-loop-pre — (undescribed, love that for us 🦋)
--enable-load-pre — (also undescribed 💜)
--enable-loop-simplifycfg-term-folding — (no description here either, the mystery continues ✨)
--enable-name-compression — enable name/filename string compression
--enable-split-backedge-in-load-pre — (no description, just a name and a dream 🎀)
--fatal-warnings — treat warnings as errors, no mercy periodt 🔥
--fdebug-compilation-dir=<string> — specifies the debug info's compilation dir
--fdebug-prefix-map=<= separated key-value pairs> — map file source paths in debug info
--filetype=<value> — choose an output file type:
- =asm — emit an assembly (.s) file
- =null — don't emit anything at all (it's for timing purposes, bestie)
- =obj — emit a native object (.o) file
-g — generate dwarf debugging info for assembly source files 🐛
--gpsize=<uint> — Global Pointer Addressing Size. The default size is 8.
--hash-based-counter-split — rename the counter variable of a comdat function based on cfg hash
--hot-cold-split — enable the hot-cold splitting pass
--import-all-index — import all external functions in index
--incremental-linker-compatible — when used with filetype=obj, emit an object file that can be used with an incremental linker
--instcombine-code-sinking — enable code sinking
--instcombine-guard-widening-window=<uint> — how wide an instruction window to bypass while looking for another guard
--instcombine-max-iterations=<uint> — limit the maximum number of instruction combining iterations
--instcombine-max-num-phis=<uint> — the maximum number of phis to handle in intptr/ptrint folding
--instcombine-maxarray-size=<uint> — the maximum array size considered when doing a combine
--instcombine-negator-enabled — should we attempt to sink negations? 🤔
--instcombine-negator-max-depth=<uint> — the maximal lookup depth when trying to check whether negation sinking is viable
--instcombine-unsafe-select-transform — enable the poison-unsafe select to and/or transform
--instrprof-atomic-counter-update-all — make all profile counter updates atomic (for testing only, so don't go shipping that 🚨)
--internalize-public-api-file=<filename> — a file containing a list of symbol names to preserve
--internalize-public-api-list=<list> — a list of symbol names to preserve
--iterative-counter-promotion — allow counter promotion across the whole loop nest
--large-code-model — create cfi directives that assume the code might be more than 2gb away
--lto-embed-bitcode=<value> — embed LLVM bitcode in object files produced by LTO:
- =none — do not embed
- =optimized — embed after all optimization passes
- =post-merge-pre-opt — embed post merge, but before optimizations
--lto-pass-remarks-filter=<regex> — only record optimization remarks from passes whose names match the given regular expression
--lto-pass-remarks-format=<format> — the format used for serializing remarks (default: YAML)
--lto-pass-remarks-output=<filename> — output filename for pass remarks
--main-file-name=<string> — specifies the name we should consider the input file
--masm-hexfloats — enable MASM-style hex float initializers (3F800000r)
--masm-integers — enable binary and hex masm integers (0b110 and 0ABCh)
--matrix-default-layout=<value> — sets the default matrix layout:
- =column-major — use column-major layout
- =row-major — use row-major layout
--mattr=<a1,+a2,-a3,...> — target specific attributes (run -mattr=help for the details 🫶)
--max-counter-promotions=<int> — the max number of allowed counter promotions
--max-counter-promotions-per-loop=<uint> — the max number of counter promotions per loop, so you don't blow up register pressure too much

…and that's only part 1, bestie, there is MORE coming, stay with me 💖

--mc-relax-all — when you're using it together with filetype=obj, it relaxes all the fixups in the emitted object file. all of them. every last one ✨
--mcpu=<cpu-name> — target a specific cpu type, like, be intentional about it. and if you're nosy about the options, -mcpu=help gives you the details 🫖
--merror-missing-parenthesis — errors out when there's a missing parenthesis around predicate registers. no more freestyling, we're checking 😤
--merror-noncontigious-register — errors out for register names that aren't contigious. yes, that's the spelling, we're leaving it exactly as it is 💅
--mhvx — enable Hexagon Vector eXtensions 🔥
--mhvx=<value> — also enables Hexagon Vector eXtensions, but now you get to choose, main-character style:
- =v60 — build for HVX v60
- =v62 — build for HVX v62
- =v65 — build for HVX v65
- =v66 — build for HVX v66
- =v67 — build for HVX v67
--mips-compact-branches=<value> — MIPS specific: compact branch policy. pick your era:
- =never — do not use compact branches if possible.
- =optimal — use compact branches where appropriate. this is the default, she's the moment 💖
- =always — always use compact branches if possible.
--mips16-constant-islands — enable mips16 constant islands 🌸
--mips16-hard-float — enable mips16 hard float 🌸
--mir-strip-debugify-only — should mir-strip-debug only strip debug info from debugified modules by default? that's the question bestie
--mno-compound — disable looking for compound instructions for Hexagon
--mno-fixup — disable fixing up resolved relocations for Hexagon
--mno-ldc1-sdc1 — expand double precision loads and stores into their single precision counterparts
--mno-pairing — disable looking for duplex instructions for Hexagon
--mwarn-missing-parenthesis — warns (doesn't error, she's gentle) for missing parenthesis around predicate registers
--mwarn-noncontigious-register — warns for register names that arent contigious (again: their spelling, not mine, love that for them)
--mwarn-sign-mismatch — warns when a signed and unsigned value are mismatched
-n — don't assume the assembly file starts in the text section. assumptions? not today
--no-deprecated-warn — suppress all deprecated warnings
--no-discriminators — disable generation of discriminator information.
--no-exec-stack — the file doesn't need an exec stack
--no-warn — suppress all warnings. silence, iconic, sometimes a little scary 🫣
--nvptx-sched4reg — NVPTX specific: schedule for register pressue (their typo, we're keeping it, it's part of the lore)
-o=<filename> — output filename 📝
--output-asm-variant=<uint> — which syntax variant to use for output printing
--poison-checking-function-local — check that returns are non-poison. for testing purposes only, bestie, don't get any ideas
--position-independent — position independent ✨
--preserve-comments — preserve comments in the outputted assembly. your notes stay, periodt
--print-imm-hex — prefer hex format for immediate values
--r600-ir-structurize — use the StructurizeCFG IR pass
--rdf-dump — no description in the original. mysterious queen behaviour, she says nothing and we respect it 💀
--rdf-limit=<uint> — also no description. same energy
--relax-relocations — emit R_X86_64_GOTPCRELX instead of R_X86_64_GOTPCREL
--runtime-counter-relocation — enable relocating counters at runtime.
--safepoint-ir-verifier-print-only — again, no description. silent, powerful, unbothered
--sample-profile-check-record-coverage=<N> — emit a warning if less than N% of records in the input profile are matched to the IR.
--sample-profile-check-sample-coverage=<N> — emit a warning if less than N% of samples in the input profile are matched to the IR.
--sample-profile-max-propagate-iterations=<uint> — maximum number of iterations to go through when propagating sample block/edge weights through the CFG.
--save-temp-labels — don't discard temporary labels 🙏
--show-encoding — show instruction encodings
--show-inst — show the internal instruction representation
--show-inst-operands — show instruction operands as parsed
--skip-ret-exit-block — suppress counter promotion if exit blocks contain ret.
--speculative-counter-promotion-max-exiting=<uint> — the max number of exiting blocks of a loop to allow speculative counter promotion
--speculative-counter-promotion-to-loop — when this option is false, if the target block is in a loop, the promotion will be disallowed unless the promoted counter update can be further/iteratively promoted into an acyclic region.
--split-dwarf-file=<filename> — DWO output filename
--summary-file=<string> — the summary file to use for function importing.
--tail-predication=<value> — MVE tail-predication pass options. the menu:
- =disabled — don't tail-predicate loops
- =enabled-no-reductions — enable tail-predication, but not for reduction loops
- =enabled — enable tail-predication, including reduction loops
- =force-enabled-no-reductions — enable tail-predication, but not for reduction loops, and force this which might be unsafe ⚠️
- =force-enabled — enable tail-predication, including reduction loops, and force this which might be unsafe ⚠️
--thinlto-assume-merged — assume the input has already undergone ThinLTO function importing and the other pre-optimization pipeline changes.
--threads=<int> — no description given. she's booked and busy, no time to explain 💅
--triple=<string> — target triple to assemble for; peep -version for the available targets
--verify-region-info — verify region info. heads up: it's time consuming ⏳
--vp-counters-per-site=<number> — the average number of profile counters allocated per value profiling site.
--vp-static-alloc — do static counter allocation for the value profiler
--x86-align-branch=<string> — specify which types of branches to align. it's a plus-separated list of types, and the types are:
- jcc — indicates conditional jumps
- fused — indicates fused conditional jumps
- jmp — indicates direct unconditional jumps
- call — indicates direct and indirect calls
- ret — indicates rets
- indirect — indicates indirect unconditional jumps
--x86-align-branch-boundary=<uint> — controls how the assembler should align branches with NOP. if the boundary's size is not 0, it should be a power of 2 and no less than 32. branches get aligned so they don't end up across or against the boundary of that specified size. the default value 0 does not align branches.
--x86-branches-within-32B-boundaries — align selected instructions to mitigate the negative performance impact of Intel's micro code update for errata skx102.

⚠️ this one may break assumptions about labels corresponding to particular instructions, and it should be used with caution. bestie, read that twice.

--x86-pad-max-prefix-size=<uint> — maximum number of prefixes to use for padding

✨ Generic Options ✨

--help — display the available options (--help-hidden for more) 👑
--help-list — display the list of available options (--help-list-hidden for more) 👑
--version — display the version of this program

🦋 Polly Options 🦋 — Configure the polly loop optimizer

--polly — enable the polly optimizer. note that it only kicks in at -O3 ✨
--polly-2nd-level-tiling — enable a 2nd level loop of loop tiling
--polly-ast-print-accesses — print memory access functions
--polly-context=<isl parameter set> — provide additional constraints on the context parameters
--polly-dce-precise-steps=<int> — the number of precise steps between two approximating iterations. (a value of -1 schedules another approximation stage before the actual dead code elimination. — yes, that parenthesis never closes in the original, we're leaving it exactly as shipped 🙈
--polly-delicm-max-ops=<int> — maximum number of isl operations to invest for lifetime analysis; 0=no limit
--polly-detect-full-functions — allow the detection of full functions
--polly-dump-after — dump the module after Polly transformations into a file suffixed with "-after"
--polly-dump-after-file=<string> — dump the module after Polly transformations to the given file
--polly-dump-before — dump the module before Polly transformations into a file suffixed with "-before"
--polly-dump-before-file=<string> — dump the module before Polly transformations to the given file
--polly-enable-simplify — simplify SCoP after optimizations
--polly-ignore-func=<string> — ignore functions that match a regex. multiple regexes can be comma separated. Scop detection will ignore all functions that match ANY of the regexes provided.
--polly-isl-arg=<argument> — option passed to ISL
--polly-on-isl-error-abort — abort if an isl error is encountered 💀
--polly-only-func=<string> — only run on functions that match a regex. multiple regexes can be comma separated. Scop detection will run on all functions that match ANY of the regexes provided.
--polly-only-region=<identifier> — only run on certain regions, bestie. Heads up tho: that identifier you pass must appear in the name of the region's entry block, or it's simply not gonna find its people. 🫶
--polly-only-scop-detection — run scop detection and that's it. No other optimizations, none, zero, nada. Just the detection era. 🔍
--polly-optimized-scops — Polly dumps the polyhedral description of Scops that got optimized with the isl scheduling optimizer, and the set of post-scheduling transformations is applied on the schedule tree. It's giving very "show me the receipts". 📜
--polly-parallel — generate thread parallel code. (isl codegen only, so keep that in mind 👀)
--polly-parallel-force — force generation of thread parallel code, ignoring any cost model whatsoever. No thoughts, no cost model, just parallel. 😤
--polly-pattern-matching-based-opts — perform optimizations based on pattern matching. ✨
--polly-process-unprofitable — process scops that are unlikely to benefit from Polly optimizations. Yes, even the ones that probably won't slay. Equal opportunity optimization. 💖
--polly-register-tiling — enable register tiling. 🎀
--polly-report — print information about the activities of Polly. It's basically Polly's storytime. 📢
--polly-show — highlight the code regions that will be optimized in a (CFG BBs and LLVM-IR instructions). 🌈
--polly-show-only — highlight the code regions that will be optimized in a (CFG only BBs). 🌸
--polly-stmt-granularity=<value> — the algorithm to use for splitting basic blocks into multiple statements. Pick your fighter:
- =bb — one statement per basic block. 🧱
- =scalar-indep — the scalar independence heuristic. 🧮
- =store — store-level granularity. 🛒
--polly-target=<value> — the hardware to target. Your pick:
- =cpu — generate CPU code. 💻
--polly-tiling — enable loop tiling. 🔲
--polly-vectorizer=<value> — select the vectorization strategy. The lineup:
- =none — no Vectorization. Rest era. 😴
- =polly — Polly internal vectorizer. 👑
- =stripmine — strip-mine outer loops so the loop-vectorizer triggers. ⛏️

And that's the whole page, fr fr. This masterpiece is from llvm-mc 12, dated February 2022, and it lives on the LLVM-MC(1) man page. Periodt. 💅✨

📦 where this page comes from

ubuntu jammy: llvm-12 1:12.0.1-19ubuntu3 · APACHE-2-LLVM-EXCEPTIONS
💖 girlified by deepseek-ai/DeepSeek-V4.1-Flash on 2026-10-13: a language model's rewrite of the original, under its terms