JAVA(1)

📦 version 17.0.20.1, in opensuse-leap 16.0

💖 JAVA(1) 💖

okay bestie, so this is the manual page for java — yes, that java — straight out of the JDK Commands section of the man pages. she's iconic, she's the main character of the whole JDK, and she's literally how you launch a Java application. let's get into it 🫶

✨ NAME ✨

java — launch a Java application. that's it. that's the whole name section. short king 💅

🎀 SYNOPSIS 🎀

okay so there's like four whole ways to call her, and honestly? each one has its own vibe. here they all are, verbatim, no cap:

to launch a class file:

java [options] mainclass [args ...]

to launch the main class in a JAR file:

java [options] -jar jarfile [args ...]

to launch the main class in a module:

java [options] -m module[/mainclass] [args ...]

or, same energy, spelled out:

java [options] --module module[/mainclass] [args ...]

to launch a single source-file program:

java [options] source-file [args ...]

and now the little glossary of everything in those brackets, bestie, don't skip it 👇

  • options — Optional: means command-line options separated by spaces. see Overview of Java Options for a description of the available options (that's a whole other section, iykyk).
  • mainclass — the name of the class to be launched. any command-line entries that follow classname are the arguments for the main method.
  • -jar jarfile — executes a program encapsulated in a JAR file. the jarfile argument is the name of a JAR file whose manifest contains a line in the form Main-Class:classname, and that line defines the class with the public static void main(String[] args) method — aka your application's starting point 🔥. and listen: when you use -jar, the specified JAR file is the source of all user classes, and other class path settings are ignored. literally ignored. if you're using JAR files, then see jar.
  • -m or --module module[/mainclass] — executes the main class in a module, specified by mainclass if it's given, or, if it isn't given, the value in the module. in other words, mainclass can be used when it is not specified by the module, or to override the value when it is specified. flexible queen 💅 see Standard Options for Java.
  • source-file — only used to launch a single source-file program. specifies the source file that contains the main class when using source-file mode. see Using Source-File Mode to Launch Single-File Source-Code Programs (we get there, keep reading 🌸).
  • args ... — Optional: arguments following mainclass, source-file, -jar jarfile, and -m or --module module/mainclass are passed as arguments to the main class.

🌈 DESCRIPTION 🌈

so. the java command starts a Java application. the way it pulls this off is: it starts the Java Virtual Machine (the JVM, she's famous), loads the specified class, and calls that class's main() method. and there are RULES for that method, bestie, it's not a free-for-all:

  • it must be declared public and static
  • it must not return any value
  • it must accept a String array as a parameter

the method declaration looks exactly like this:

public static void main(String[] args)

in source-file mode, the java command can launch a class declared in a source file. see Using Source-File Mode to Launch Single-File Source-Code Programs for a description of using the source-file mode. (yes, that section. the big one. it's coming 💜)

Note: you can use the JDK_JAVA_OPTIONS launcher environment variable to prepend its content to the actual command line of the java launcher. see Using the JDK_JAVA_OPTIONS Launcher Environment Variable ✨

by default, the first argument that isn't an option of the java command is the fully qualified name of the class to be called. if -jar is specified, then its argument is the name of the JAR file containing the class and resource files for the application. the startup class must be indicated by the Main-Class manifest header in its manifest file.

arguments after the class file name or the JAR file name get passed to the main() method. simple as that 🦋

🪟 javaw

Windows: the javaw command is identical to java, except that with javaw there's no associated console window. use javaw when you don't want a command prompt window to appear. the javaw launcher will, however, display a dialog box with error information if a launch fails. so like, it's java but without the chatty terminal window — errors still show up in a lil popup tho 🙈

🧚 USING SOURCE-FILE MODE TO LAUNCH SINGLE-FILE SOURCE-CODE PROGRAMS 🧚

okay this is the part everyone's like "wait, you can do that?" — and yes. to launch a class declared in a source file, you run the java launcher in source-file mode. entering source-file mode is determined by two items on the java command line:

  • the first item on the command line that is not an option or part of an option. in other words, the item in the command line that would otherwise be the main class name.
  • the --source version option, if present.

if the class identifies an existing file that has a .java extension, or if the --source option is specified, then source-file mode is selected. the source file is then compiled and run. the --source option can be used to specify the source version or N of the source code. this determines the API that can be used. when you set --source N, you can only use the public API that was defined in JDK N.

Note: the valid values of N change for each release, with new values added and old values removed. you'll get an error message if you use a value of N that is no longer supported. the supported values of N are the current Java SE release (17) and a limited number of previous releases, detailed in the command-line help for javac, under the --source and --release options. so don't come crying when your ancient N stops working 😭

if the file does not have the .java extension, the --source option must be used to tell the java command to use the source-file mode. the --source option is used for cases when the source file is a "script" to be executed and the name of the source file does not follow the normal naming conventions for Java source files.

in source-file mode, the effect is as though the source file is compiled into memory, and the first class found in the source file is executed. any arguments placed after the name of the source file in the original command line are passed to the compiled class when it is executed.

for example, if a file were named HelloWorld.java and contained a class named hello.World, then the source-file mode command to launch the class would be:

java HelloWorld.java

the example illustrates that the class can be in a named package, and does not need to be in the unnamed package. this use of source-file mode is informally equivalent to using the following two commands, where hello.World is the name of the class in the package:

javac -d <memory> HelloWorld.java
java -cp <memory> hello.World

in source-file mode, any additional command-line options are processed as follows:

  • the launcher scans the options specified before the source file for any that are relevant in order to compile the source file.

this includes: --class-path, --module-path, --add-exports, --add-modules, --limit-modules, --patch-module, --upgrade-module-path, and any variant forms of those options. it also includes the new --enable-preview option, described in JEP 12.

  • no provision is made to pass any additional options to the compiler, such as -processor or -Werror. soz, not happening 💀
  • command-line argument files (@-files) may be used in the standard way. long lists of arguments for either the VM or the program being invoked may be placed in files specified on the command-line by prefixing the filename with an @ character.

in source-file mode, compilation proceeds as follows:

  • any command-line options that are relevant to the compilation environment are taken into account.
  • no other source files are found and compiled, as if the source path is set to an empty value.
  • annotation processing is disabled, as if -proc:none is in effect.
  • if a version is specified, via the --source option, the value is used as the argument for an implicit --release option for the compilation. this sets both the source version accepted by compiler and the system API that may be used by the code in the source file.
  • the source file is compiled in the context of an unnamed module.
  • the source file should contain one or more top-level classes, the first of which is taken as the class to be executed.
  • the compiler does not enforce the optional restriction defined at the end of JLS §7.6, that a type in a named package should exist in a file whose name is composed from the type name followed by the .java extension.
  • if the source file contains errors, appropriate error messages are written to the standard error stream, and the launcher exits with a non-zero exit code. (non-zero exit code = not slay 😭)

in source-file mode, execution proceeds as follows:

  • the class to be executed is the first top-level class found in the source file. it must contain a declaration of the standard public static void main(String[]) method.
  • the compiled classes are loaded by a custom class loader, that delegates to the application class loader. this implies that classes appearing on the application class path cannot refer to any classes declared in the source file. like, they just can't 💅
  • the compiled classes are executed in the context of an unnamed module, as though --add-modules=ALL-DEFAULT is in effect. this is in addition to any other --add-module options that may have been specified on the command line.
  • any arguments appearing after the name of the file on the command line are passed to the standard main method in the obvious way.
  • it is an error if there is a class on the application class path whose name is the same as that of the class to be executed. name collision = instant L 🍓

see JEP 330: Launch Single-File Source-Code Programs [http://openjdk.java.net/jeps/330] for complete details.

🚀 USING THE JDK_JAVA_OPTIONS LAUNCHER ENVIRONMENT VARIABLE 🚀

Okay so JDK_JAVA_OPTIONS basically prepends its content to the options parsed from the command line. The content of the JDK_JAVA_OPTIONS environment variable is a list of arguments separated by white-space characters (as determined by isspace()). These are prepended to the command line arguments passed to java launcher. The encoding requirement for the environment variable is the same as the java command line on the system. JDK_JAVA_OPTIONS environment variable content is treated in the same manner as that specified in the command line.

Single (') or double (") quotes can be used to enclose arguments that contain whitespace characters. All content between the open quote and the first matching close quote are preserved by simply removing the pair of quotes. In case a matching quote is not found, the launcher will abort with an error message. @-files are supported as they are specified in the command line. However, as in @-files, use of a wildcard is not supported. In order to mitigate potential misuse of JDK_JAVA_OPTIONS behavior, options that specify the main class (such as -jar) or cause the java launcher to exit without executing the main class (such as -h) are disallowed in the environment variable. If any of these options appear in the environment variable, the launcher will abort with an error message. When JDK_JAVA_OPTIONS is set, the launcher prints a message to stderr as a reminder.

Example:

$ export JDK_JAVA_OPTIONS='-g @file1 -Dprop=value @file2 -Dws.prop="white spaces"'
$ java -Xint @file3

is equivalent to the command line:

java -g @file1 -Dprop=value @file2 -Dws.prop="white spaces" -Xint @file3

🌸 OVERVIEW OF JAVA OPTIONS 🌸

The java command supports a wide range of options in the following categories:

  • Standard Options for Java: Options guaranteed to be supported by all implementations of the Java Virtual Machine (JVM). They're used for common actions, such as checking the version of the JRE, setting the class path, enabling verbose output, and so on.
  • Extra Options for Java: General purpose options that are specific to the Java HotSpot Virtual Machine. They aren't guaranteed to be supported by all JVM implementations, and are subject to change. These options start with -X.

The advanced options aren't recommended for casual use. These are developer options used for tuning specific areas of the Java HotSpot Virtual Machine operation that often have specific system requirements and may require privileged access to system configuration parameters. Several examples of performance tuning are provided in Performance Tuning Examples. These options aren't guaranteed to be supported by all JVM implementations and are subject to change. Advanced options start with -XX.

  • Advanced Runtime Options for Java: Control the runtime behavior of the Java HotSpot VM.
  • Advanced JIT Compiler Options for java: Control the dynamic just-in-time (JIT) compilation performed by the Java HotSpot VM.
  • Advanced Serviceability Options for Java: Enable gathering system information and performing extensive debugging.
  • Advanced Garbage Collection Options for Java: Control how garbage collection (GC) is performed by the Java HotSpot

Boolean options are used to either enable a feature that's disabled by default or disable a feature that's enabled by default. Such options don't require a parameter. Boolean -XX options are enabled using the plus sign (-XX:+OptionName) and disabled using the minus sign (-XX:-OptionName).

For options that require an argument, the argument may be separated from the option name by a space, a colon (:), or an equal sign (=), or the argument may directly follow the option (the exact syntax differs for each option). If you're expected to specify the size in bytes, then you can use no suffix, or use the suffix k or K for kilobytes (KB), m or M for megabytes (MB), or g or G for gigabytes (GB). For example, to set the size to 8 GB, you can specify either 8g, 8192m, 8388608k, or 8589934592 as the argument. If you are expected to specify the percentage, then use a number from 0 to 1. For example, specify 0.25 for 25%.

The following sections describe the options that are obsolete, deprecated, and removed:

  • Deprecated Java Options: Accepted and acted upon --- a warning is issued when they're used.
  • Obsolete Java Options: Accepted but ignored --- a warning is issued when they're used.
  • Removed Java Options: Removed --- using them results in an error.

🎀 STANDARD OPTIONS FOR JAVA 🎀

These are the most commonly used options supported by all implementations of the JVM.

Note: To specify an argument for a long option, you can use either --name=value or --name value.
  • -agentlib:libname[=options] — Loads the specified native agent library. After the library name, a comma-separated list of options specific to the library can be used.
    • Linux and macOS: If the option -agentlib:foo is specified, then the JVM attempts to load the library named libfoo.so in the location specified by the LD_LIBRARY_PATH system variable (on macOS this variable is DYLD_LIBRARY_PATH).
    • Windows: If the option -agentlib:foo is specified, then the JVM attempts to load the library named foo.dll in the location specified by the PATH system variable.

The following example shows how to load the Java Debug Wire Protocol (JDWP) library and listen for the socket connection on port 8000, suspending the JVM before the main class loads:

-agentlib:jdwp=transport=dt_socket,server=y,address=8000
  • -agentpath:pathname[=options] — Loads the native agent library specified by the absolute path name. This option is equivalent to -agentlib but uses the full path and file name of the library.
  • --class-path classpath, -classpath classpath, or -cp classpath — A semicolon (;) separated list of directories, JAR archives, and ZIP archives to search for class files.

Specifying classpath overrides any setting of the CLASSPATH environment variable. If the class path option isn't used and classpath isn't set, then the user class path consists of the current directory (.).

As a special convenience, a class path element that contains a base name of an asterisk (*) is considered equivalent to specifying a list of all the files in the directory with the extension .jar or .JAR. A Java program can't tell the difference between the two invocations. For example, if the directory mydir contains a.jar and b.JAR, then the class path element mydir/* is expanded to A.jar:b.JAR, except that the order of JAR files is unspecified. All .jar files in the specified directory, even hidden ones, are included in the list. A class path entry consisting of an asterisk (*) expands to a list of all the jar files in the current directory. The CLASSPATH environment variable, where defined, is similarly expanded. Any class path wildcard expansion that occurs before the Java VM is started. Java programs never see wildcards that aren't expanded except by querying the environment, such as by calling System.getenv("CLASSPATH").

  • --disable-@files — Can be used anywhere on the command line, including in an argument file, to prevent further @filename expansion. This option stops expanding @-argfiles after the option.
  • --enable-preview — Allows classes to depend on preview features https://docs.oracle.com/en/java/javase/12/language/index.html#JS-LAN-GUID-5A82FE0E-0CA4-4F1F-B075-564874FE2823 of the release.
  • --module-path modulepath... or -p modulepath — A semicolon (;) separated list of directories in which each directory is a directory of modules.
  • --upgrade-module-path modulepath... — A semicolon (;) separated list of directories in which each directory is a directory of modules that replace upgradeable modules in the runtime image.
  • --add-modules module[,module...] — Specifies the root modules to resolve in addition to the initial module. module also can be ALL-DEFAULT, ALL-SYSTEM, and ALL-MODULE-PATH.
  • --list-modules — Lists the observable modules and then exits.
  • -d module_name or --describe-module module_name — Describes a specified module and then exits.
  • --dry-run — Creates the VM but doesn't execute the main method. This --dry-run option might be useful for validating the command-line options such as the module system configuration.
  • --validate-modules — Validates all modules and exit. This option is helpful for finding conflicts and other errors with modules on the module path.
  • -Dproperty=value — Sets a system property value. The -Dproperty flag names the property: property is a string with no spaces that represents the name of the property. value is a string that represents the value of the property. If value is a string with spaces, then enclose it in quotation marks (for example -Dfoo="foo bar").
  • -disableassertions[:[packagename]...|:classname] or -da[:[packagename]...|:classname] — Disables assertions. By default, assertions are disabled in all packages and classes. With no arguments, -disableassertions (-da) disables assertions in all packages and classes. With the packagename argument ending in ..., the switch disables assertions in the specified package and any subpackages. If the argument is simply ..., then the switch disables assertions in the unnamed package in the current working directory. With the classname argument, the switch disables assertions in the specified class.

So, picking up right where we left off with -disableassertions (-da): it applies to all class loaders and to system classes, which lowkey don't have a class loader. There's exactly one exception, bestie: if you pass the option with no arguments, then it does not apply to system classes. That's the whole point — it makes it easy to disable assertions in every class except system classes. The -disablesystemassertions option is your separate switch for disabling assertions in all system classes. And when you want to explicitly enable assertions in specific packages or classes, that's what -enableassertions (-ea) is for. Both options can be used at the same time, fr fr. For example, to run the MyClass application with assertions enabled in the package com.wombat.fruitbat (and any subpackages) but disabled in the class com.wombat.fruitbat.Brickbat, use this command:

java -ea:com.wombat.fruitbat... -da:com.wombat.fruitbat.Brickbat MyClass
  • -disablesystemassertions or -dsa — disables assertions in all system classes. Clean, simple, bye 👋
  • -enableassertions[:[packagename]...|:classname] or -ea[:[packagename]...|:classname] — enables assertions. By default, assertions are disabled in all packages and classes. With no arguments, -enableassertions (-ea) enables assertions in all packages and classes. With the packagename argument ending in ..., the switch enables assertions in the specified package and any subpackages. If the argument is simply ..., then the switch enables assertions in the unnamed package in the current working directory. With the classname argument, the switch enables assertions in the specified class.

Okay so the -enableassertions (-ea) option applies to all class loaders and to system classes (which don't have a class loader). There's one exception to this rule: if the option is provided with no arguments, then it doesn't apply to system classes. This makes it easy to enable assertions in all classes except for system classes. The -enablesystemassertions option provides a separate switch to enable assertions in all system classes. To explicitly disable assertions in specific packages or classes, use the -disableassertions (-da) option. If a single command contains multiple instances of these switches, then they're processed in order, before loading any classes. For example, to run the MyClass application with assertions enabled only in the package com.wombat.fruitbat (and any subpackages) but disabled in the class com.wombat.fruitbat.Brickbat, use this command:

java -ea:com.wombat.fruitbat... -da:com.wombat.fruitbat.Brickbat MyClass
  • -enablesystemassertions or -esa — enables assertions in all system classes. System classes get the love too 💖
  • -help, -h, or -? — prints the help message to the error stream.
  • --help — prints the help message to the output stream.
  • -javaagent:jarpath[=options] — loads the specified Java programming language agent. See java.lang.instrument.
  • --show-version — prints the product version to the output stream and continues.
  • -showversion — prints the product version to the error stream and continues.
  • --show-module-resolution — shows module resolution output during startup.
  • -splash:imagepath — shows the splash screen with the image specified by imagepath. HiDPI scaled images are automatically supported and used if available. The unscaled image file name, such as image.ext, should always be passed as the argument to the -splash option. The most appropriate scaled image provided is picked up automatically.

For example, to show the splash.gif file from the images directory when starting your application, use this option:

-splash:images/splash.gif

See the SplashScreen API documentation for more information.

  • -verbose:class — displays information about each loaded class.
  • -verbose:gc — displays information about each garbage collection (GC) event.
  • -verbose:jni — displays information about the use of native methods and other Java Native Interface (JNI) activity.
  • -verbose:module — displays information about the modules in use.
  • --version — prints product version to the output stream and exits.
  • -version — prints product version to the error stream and exits.
  • -X — prints the help on extra options to the error stream.
  • --help-extra — prints the help on extra options to the output stream.
  • @argfile — specifies one or more argument files prefixed by @ used by the java command. Ngl, it isn't uncommon for the java command line to be very long because of the .jar files needed in the classpath. The @argfile option overcomes command-line length limitations by enabling the launcher to expand the contents of argument files after shell expansion, but before argument processing. Contents in the argument files are expanded because otherwise, they would be specified on the command line until the --disable-@files option was encountered.

The argument files can also contain the main class name and all options. If an argument file contains all of the options required by the java command, then the command line could simply be:

java @argfile

See java Command-Line Argument Files for a description and examples of using @-argfiles.

🔥 EXTRA OPTIONS FOR JAVA 🔥

The following java options are general purpose options that are specific to the Java HotSpot Virtual Machine. Extra-options era, let's go 🫶

  • -Xbatch — disables background compilation. By default, the JVM compiles the method as a background task, running the method in interpreter mode until the background compilation is finished. The -Xbatch flag disables background compilation so that compilation of all methods proceeds as a foreground task until completed. This option is equivalent to -XX:-BackgroundCompilation.
  • -Xbootclasspath/a:directories|zip|JAR-files — specifies a list of directories, JAR files, and ZIP archives to append to the end of the default bootstrap class path.

Linux and macOS: colons (:) separate entities in this list.

Windows: semicolons (;) separate entities in this list.

  • -Xcheck:jni — performs additional checks for Java Native Interface (JNI) functions.

The following checks are considered indicative of significant problems with the native code, and the JVM terminates with an irrecoverable error in such cases:

  • The thread doing the call is not attached to the JVM.
  • The thread doing the call is using the JNIEnv belonging to another thread.
  • A parameter validation check fails:
    • A jfieldID, or jmethodID, is detected as being invalid. For example:
      • Of the wrong type
      • Associated with the wrong class
    • A parameter of the wrong type is detected.
    • An invalid parameter value is detected. For example:
      • NULL where not permitted
      • An out-of-bounds array index, or frame capacity
      • A non-UTF-8 string
      • An invalid JNI reference
      • An attempt to use a ReleaseXXX function on a parameter not produced by the corresponding GetXXX function

The following checks only result in warnings being printed:

  • A JNI call was made without checking for a pending exception from a previous JNI call, and the current call is not safe when an exception may be pending.
  • A class descriptor is in decorated format (Lname;) when it should not be.
  • A NULL parameter is allowed, but its use is questionable.
  • Calling other JNI functions in the scope of Get/ReleasePrimitiveArrayCritical or Get/ReleaseStringCritical

Expect a performance degradation when this option is used. That's the tradeoff, bestie 😭

  • -Xdebug — does nothing. Provided for backward compatibility.
  • -Xdiag — shows additional diagnostic messages.
  • -Xint — runs the application in interpreted-only mode. Compilation to native code is disabled, and all bytecode is executed by the interpreter. The performance benefits offered by the just-in-time (JIT) compiler aren't present in this mode.
  • -Xinternalversion — displays more detailed JVM version information than the -version option, and then exits.
  • -Xlog:option — configure or enable logging with the Java Virtual Machine (JVM) unified logging framework. See Enable Logging with the JVM Unified Logging Framework.
  • -Xmixed — executes all bytecode by the interpreter except for hot methods, which are compiled to native code. On by default. Use -Xint to switch off.
  • -Xmn size — sets the initial and maximum size (in bytes) of the heap for the young generation (nursery) in the generational collectors. Append the letter k or K to indicate kilobytes, m or M to indicate megabytes, or g or G to indicate gigabytes. The young generation region of the heap is used for new objects. GC is performed in this region more often than in other regions. If the size for the young generation is too small, then a lot of minor garbage collections are performed. If the size is too large, then only full garbage collections are performed, which can take a long time to complete. It is recommended that you do not set the size for the young generation for the G1 collector, and keep the size for the young generation greater than 25% and less than 50% of the overall heap size for other collectors. The following examples show how to set the initial and maximum size of young generation to 256 MB using various units:
-Xmn256m
-Xmn262144k
-Xmn268435456

Instead of the -Xmn option to set both the initial and maximum size of the heap for the young generation, you can use -XX:NewSize to set the initial size and -XX:MaxNewSize to set the maximum size.

  • -Xms size — sets the minimum and the initial size of the heap, in bytes 💖. Rules, bestie: this value has to be a multiple of 1024 and greater than 1 MB. Append k or K for kilobytes, m or M for megabytes, g or G for gigabytes. Here's the same 6 MB in three different units, because units are a vibe:
-Xms6291456
-Xms6144k
-Xms6m

So instead of using -Xms to set both the minimum and initial size at once, you can split it: -XX:MinHeapSize handles the minimum, -XX:InitialHeapSize handles the initial. Two flags, one job each, very main character energy ✨.

If you don't set this option, the initial size is set as the sum of the sizes allocated for the old generation and the young generation. And the initial size of the heap for the young generation can be set with the -Xmn option or the -XX:NewSize option. The more you know 🎀.

  • -Xmx size — sets the maximum size of the heap, in bytes. Same deal as before: must be a multiple of 1024 and greater than 2 MB (note: 2 MB here, not 1, don't get got 💀). Append k or K for kilobytes, m or M for megabytes, or g or G for gigabytes 🌈.

The default value is chosen at runtime based on system configuration. For server deployments, -Xms and -Xmx are often set to the same value — matching set, cute 🫶. Here's 80 MB of maximum allowed allocated memory in three units:

-Xmx83886080
-Xmx81920k
-Xmx80m

-Xmx is literally equivalent to -XX:MaxHeapSize. Same flag, different name, iykyk ⭐.

  • -Xnoclassgc — disables garbage collection (GC) of classes. Lowkey a speed thing: it can save some GC time, which shortens interruptions during the application run 🦋.

When you specify -Xnoclassgc at startup, the class objects in the application are left untouched during GC and are always considered live. The catch 😭: this can result in more memory being permanently occupied, which, if not used carefully, throws an out-of-memory exception. So like… use it on purpose, not by accident, bestie 💅.

  • -Xrs — reduces the JVM's use of operating system signals.

Quick context: shutdown hooks enable the orderly shutdown of a Java application by running user cleanup code (such as closing database connections) at shutdown, even if the JVM terminates abruptly. They're the responsible ones 🧚.

Linux and macOS 🍓:

  • The JVM catches signals to implement shutdown hooks for unexpected termination. It uses SIGHUP, SIGINT, and SIGTERM to initiate the running of shutdown hooks.
  • Applications embedding the JVM frequently need to trap signals such as SIGINT or SIGTERM, which can lead to interference with the JVM signal handlers. The -Xrs option is available to address this issue. When -Xrs is used, the signal masks for SIGINT, SIGTERM, SIGHUP, and SIGQUIT aren't changed by the JVM, and signal handlers for these signals aren't installed.

Windows 🪟:

  • The JVM watches for console control events to implement shutdown hooks for unexpected termination. Specifically, the JVM registers a console control handler that begins shutdown-hook processing and returns TRUE for CTRL_C_EVENT, CTRL_CLOSE_EVENT, CTRL_LOGOFF_EVENT, and CTRL_SHUTDOWN_EVENT.
  • The JVM uses a similar mechanism to implement the feature of dumping thread stacks for debugging purposes. The JVM uses CTRL_BREAK_EVENT to perform thread dumps.
  • If the JVM is run as a service (for example, as a servlet engine for a web server), then it can receive CTRL_LOGOFF_EVENT but shouldn't initiate shutdown, because the operating system doesn't actually terminate the process. To avoid possible interference such as this, the -Xrs option can be used. When the -Xrs option is used, the JVM doesn't install a console control handler, implying that it doesn't watch for or process CTRL_C_EVENT, CTRL_CLOSE_EVENT, CTRL_LOGOFF_EVENT, or CTRL_SHUTDOWN_EVENT.

There are two consequences of specifying -Xrs 😬:

  • Linux and macOS: SIGQUIT thread dumps aren't available.
  • Windows: Ctrl + Break thread dumps aren't available.

User code is responsible for causing shutdown hooks to run — for example, by calling the System.exit() when the JVM is to be terminated 👑.

  • -Xshare: mode — sets the class data sharing (CDS) mode. Possible mode arguments for this option include the following:
    • auto — use shared class data if possible (default) ✨.
    • on — require using shared class data, otherwise fail.
Note 📌: the -Xshare:on option is used for testing purposes only. It may cause the VM to unexpectedly exit during start-up when the CDS archive cannot be used (for example, when certain VM parameters are changed, or when a different JDK is used). This option should not be used in production environments 💀.
  • off — do not attempt to use shared class data.
  • -XshowSettings — shows all settings and then continues. That's it, that's the whole flag 🎀.
  • -XshowSettings: category — shows settings and continues. Possible category arguments for this option include the following:
    • all — shows all categories of settings. This is the default value.
    • locale — shows settings related to locale.
    • properties — shows settings related to system properties.
    • vm — shows the settings of the JVM.
    • system — Linux: shows host system or container configuration and continues 🐧.
  • -Xss size — sets the thread stack size (in bytes). Append k or K for KB, m or M for MB, or g or G for GB. The default value depends on the platform:
    • Linux/x64 (64-bit): 1024 KB
    • macOS (64-bit): 1024 KB
    • Windows: the default value depends on virtual memory 🙈

Here's the thread stack size set to 1024 KB in different units:

-Xss1m
-Xss1024k
-Xss1048576

This option is similar to -XX:ThreadStackSize 💜.

  • --add-reads module=target-module(,target-module)* — updates *module* to read the *target-module*, regardless of the module declaration. The *target-module* can be all unnamed` to read all unnamed modules 🌸.
  • --add-exports module/package=target-module(,target-module)* — updates *module* to export *package* to *target-module*, regardless of module declaration. The *target-module* can be all unnamed` to export to all unnamed modules 🦋.
  • --add-opens module/package=target-module(,target-module`)* — updates module to open package to target-module, regardless of module declaration. Opening the doors, politely, on purpose 💅.
  • --limit-modules module[,module...] — specifies the limit of the universe of observable modules. Cosmic 🌌.
  • --patch-module module=file(;file)* — overrides or augments a module with classes and resources in JAR files or directories 🧩.
  • --source version — sets the version of the source in source-file mode.

🍎 EXTRA OPTIONS FOR MACOS 🍎

The following extra options are macOS specific. Only the Apple girlies get these, no cap 🫶.

  • -XstartOnFirstThread — runs the main() method on the first (AppKit) thread 🎀.
  • -Xdock:name= application_name — overrides the default application name displayed in dock.
  • -Xdock:icon= path_to_icon_file — overrides the default icon displayed in dock. Cute icon override, literally iconic ✨.

🔓 ADVANCED OPTIONS FOR JAVA 🔓

These java options can be used to enable other advanced options. The unlock keys, basically 🗝️.

  • -XX:+UnlockDiagnosticVMOptions — unlocks the options intended for diagnosing the JVM. By default, this option is disabled and diagnostic options aren't available.
Real talk ⚠️: command line options that are enabled with the use of this option are not supported. If you encounter issues while using any of these options, it is very likely that you will be required to reproduce the problem without using any of these unsupported options before Oracle Support can assist with an investigation. It is also possible that any of these options may be removed or their behavior changed without any warning. So like… be brave, but know the risk 💀.
  • -XX:+UnlockExperimentalVMOptions — unlocks the options that provide experimental features in the JVM. By default, this option is disabled and experimental features aren't available. Welcome to the lab 🧪✨.

⚙️ ADVANCED RUNTIME OPTIONS FOR JAVA ⚙️

These java options control the runtime behavior of the Java HotSpot VM. The engine-room flags 🔥.

  • -XX:ActiveProcessorCount= x — overrides the number of CPUs that the VM will use to calculate the size of thread pools it will use for various operations such as Garbage Collection and ForkJoinPool.

The VM normally determines the number of available processors from the operating system. This flag can be useful for partitioning CPU resources when running multiple Java processes in docker containers 🐳. This flag is honored even if UseContainerSupport is not enabled. See -XX:-UseContainerSupport for a description of enabling and disabling container support 👑.

  • -XX:AllocateHeapAt= path — takes a path to the file system and uses memory mapping to allocate the object heap on the memory device. Using this option enables the HotSpot VM to allocate the Java object heap on an alternative memory device, such as an NV-DIMM, specified by the user 🌈.

Alternative memory devices that have the same semantics as DRAM — including the semantics of atomic operations — can be used instead of DRAM for the object heap without changing the existing application code. Everything else (all other memory structures, such as the code heap, metaspace, and thread stacks) continues to reside in DRAM. Know the difference, bestie 🎀.

Some operating systems expose non-DRAM memory through the file system. Memory-mapped files in these file systems bypass the page cache and provide a direct mapping of virtual memory to the physical memory on the device ✨. The existing heap related flags (such as -Xmx and -Xms) and garbage-collection related flags continue to work as before. No cap, everything just keeps working 🫶.

  • -XX:-CompactStrings — turns the Compact Strings feature off. By default this option is enabled (i.e. Compact Strings is on). When it's enabled, Java Strings that contain only single-byte characters are internally represented and stored as single-byte-per-character Strings using ISO-8859-1 / Latin-1 encoding — which honestly? slashes the space needed for those Strings by 50% ✨. For Java Strings that contain at least one multibyte character: those get represented and stored as 2 bytes per character using UTF-16 encoding. Disabling Compact Strings forces the use of UTF-16 encoding as the internal representation for all Java Strings.

Cases where it may be beneficial to disable Compact Strings include the following:

  • When it's known that an application overwhelmingly will be allocating multibyte character Strings
  • In the unexpected event where a performance regression is observed in migrating from Java SE 8 to Java SE 9 and an analysis shows that Compact Strings introduces the regression

In both of these scenarios, disabling Compact Strings makes sense. like, do the math bestie 💅

  • -XX:ErrorFile=filename — tells the JVM where to write the error data when an irrecoverable error happens 💀. By default, this file is created in the current working directory and named hs_err_pidpid.log where pid is the identifier of the process that encountered the error.

The following example shows how to set the default log file (note that the identifier of the process is specified as %p):

-XX:ErrorFile=./hs_err_pid%p.log
  • Linux and macOS: The following example shows how to set the error log to /var/log/java/java_error.log:
-XX:ErrorFile=/var/log/java/java_error.log
  • Windows: The following example shows how to set the error log file to C:/log/java/java_error.log:
-XX:ErrorFile=C:/log/java/java_error.log

If the file exists, and is writeable, then it will be overwritten. Otherwise, if the file can't be created in the specified directory (due to insufficient space, permission problem, or another issue), then the file is created in the temporary directory for the operating system:

  • Linux and macOS: The temporary directory is /tmp.
  • Windows: The temporary directory is specified by the value of the TMP environment variable; if that environment variable isn't defined, then the value of the TEMP environment variable is used.
  • -XX:+ExtensiveErrorReports — enables the reporting of more extensive error information in the ErrorFile 🔥. This option can be turned on in environments where maximal information is desired — even if the resulting logs may be quite large and/or contain information that might be considered sensitive 🙈. The information can vary from release to release, and across different platforms. By default this option is disabled.
  • -XX:FlightRecorderOptions=parameter=value (or) -XX:FlightRecorderOptions:parameter=value — sets the parameters that control the behavior of JFR. so many knobs, so little time 🎀

The following list contains the available JFR parameter=value entries:

  • globalbuffersize=size — Specifies the total amount of primary memory used for data retention. The default value is based on the value specified for memorysize. Change the memorysize parameter to alter the size of global buffers.
  • maxchunksize=size — Specifies the maximum size (in bytes) of the data chunks in a recording. Append m or M to specify the size in megabytes (MB), or g or G to specify the size in gigabytes (GB). By default, the maximum size of data chunks is set to 12 MB. The minimum allowed is 1 MB.
  • memorysize=size — Determines how much buffer memory should be used, and sets the globalbuffersize and numglobalbuffers parameters based on the size specified. Append m or M to specify the size in megabytes (MB), or g or G to specify the size in gigabytes (GB). By default, the memory size is set to 10 MB.
  • numglobalbuffers — Specifies the number of global buffers used. The default value is based on the memory size specified. Change the memorysize parameter to alter the number of global buffers.
  • old-object-queue-size=number-of-objects — Maximum number of old objects to track. By default, the number of objects is set to 256.
  • repository=path — Specifies the repository (a directory) for temporary disk storage. By default, the system's temporary directory is used.
  • retransform={true|false} — Specifies whether event classes should be retransformed using JVMTI. If false, instrumentation is added when event classes are loaded. By default, this parameter is enabled.
  • samplethreads={true|false} — Specifies whether thread sampling is enabled. Thread sampling occurs only if the sampling event is enabled along with this parameter. By default, this parameter is enabled.
  • stackdepth=depth — Stack depth for stack traces. By default, the depth is set to 64 method calls. The maximum is 2048. Values greater than 64 could create significant overhead and reduce performance.
  • threadbuffersize=size — Specifies the per-thread local buffer size (in bytes). By default, the local buffer size is set to 8 kilobytes, with a minimum value of 4 kilobytes. Overriding this parameter could reduce performance and is not recommended.

You can specify values for multiple parameters by separating them with a comma.

  • -XX:LargePageSizeInBytes=size — Sets the maximum large page size (in bytes) used by the JVM. The size argument must be a valid page size supported by the environment to have any effect. Append the letter k or K to indicate kilobytes, m or M to indicate megabytes, or g or G to indicate gigabytes. By default, the size is set to 0, meaning that the JVM will use the default large page size for the environment as the maximum size for large pages. See Large Pages.

The following example describes how to set the large page size to 1 gigabyte (GB):

-XX:LargePageSizeInBytes=1g
  • -XX:MaxDirectMemorySize=size — Sets the maximum total size (in bytes) of the java.nio package, direct-buffer allocations. Append the letter k or K to indicate kilobytes, m or M to indicate megabytes, or g or G to indicate gigabytes. By default, the size is set to 0, meaning that the JVM chooses the size for NIO direct-buffer allocations automatically. auto-pilot mode fr fr 🦋

The following examples illustrate how to set the NIO size to 1024 KB in different units:

-XX:MaxDirectMemorySize=1m
-XX:MaxDirectMemorySize=1024k
-XX:MaxDirectMemorySize=1048576
  • -XX:-MaxFDLimit — Disables the attempt to set the soft limit for the number of open file descriptors to the hard limit. By default, this option is enabled on all platforms, but is ignored on Windows. The only time that you may need to disable this is on Mac OS, where its use imposes a maximum of 10240, which is lower than the actual system maximum 🍓.
  • -XX:NativeMemoryTracking=mode — Specifies the mode for tracking JVM native memory usage. Possible mode arguments for this option include the following:
    • off — Instructs not to track JVM native memory usage. This is the default behavior if you don't specify the -XX:NativeMemoryTracking option.
    • summary — Tracks memory usage only by JVM subsystems, such as Java heap, class, code, and thread.
    • detail — In addition to tracking memory usage by JVM subsystems, track memory usage by individual CallSite, individual virtual memory region and its committed regions.
  • -XX:TrimNativeHeapInterval=millis — Interval, in ms, at which the JVM will trim the native heap. Lower values will reclaim memory more eagerly at the cost of higher overhead. A value of 0 (default) disables native heap trimming. Native heap trimming is performed in a dedicated thread 🧚

This option is only supported on Linux with GNU C Library (glibc).

  • -XX:ObjectAlignmentInBytes=alignment — Sets the memory alignment of Java objects (in bytes). By default, the value is set to 8 bytes. The specified value should be a power of 2, and must be within the range of 8 and 256 (inclusive). This option makes it possible to use compressed pointers with large Java heap sizes.

The heap size limit in bytes is calculated as:

4GB * ObjectAlignmentInBytes
Note: As the alignment value increases, the unused space between objects also increases. As a result, you may not realize any benefits from using compressed pointers with large Java heap sizes. ngl that's the catch 😭
  • -XX:OnError=string — Sets a custom command or a series of semicolon-separated commands to run when an irrecoverable error occurs. If the string contains spaces, then it must be enclosed in quotation marks.
    • Linux and macOS: The following example shows how the -XX:OnError option can be used to run the gcore command to create a core image, and start the gdb debugger to attach to the process in case of an irrecoverable error (the %p designates the current process identifier):
-XX:OnError="gcore %p;gdb -p %p"
  • 🪟 Windows: The following example shows how the -XX:OnError option can be used to run the userdump.exe utility to obtain a crash dump in case of an irrecoverable error (the %p designates the current process identifier). This example assumes that the path to the userdump.exe utility is specified in the PATH environment variable:
-XX:OnError="userdump.exe %p"
  • 🌸 -XX:OnOutOfMemoryError=string — Sets a custom command or a series of semicolon-separated commands to run when an OutOfMemoryError exception is first thrown. If the string contains spaces, then it must be enclosed in quotation marks. For an example of a command string, see the description of the -XX:OnError option.
  • ✨ -XX:+PrintCommandLineFlags — Enables printing of ergonomically selected JVM flags that appeared on the command line. It can be useful to know the ergonomic values set by the JVM, such as the heap space size and the selected garbage collector. By default, this option is disabled and flags aren't printed.
  • 🦋 -XX:+PreserveFramePointer — Selects between using the RBP register as a general purpose register (-XX:-PreserveFramePointer) and using the RBP register to hold the frame pointer of the currently executing method (-XX:+PreserveFramePointer). If the frame pointer is available, then external profiling tools (for example, Linux perf) can construct more accurate stack traces. 🔥
  • 💜 -XX:+PrintNMTStatistics — Enables printing of collected native memory tracking data at JVM exit when native memory tracking is enabled (see -XX:NativeMemoryTracking). By default, this option is disabled and native memory tracking data isn't printed.
  • 🎀 -XX:SharedArchiveFile=path — Specifies the path and name of the class data sharing (CDS) archive file.

See Application Class Data Sharing.

  • 🔐 -XX:+VerifySharedSpaces — If this option is specified, the JVM will load a CDS archive file only if it passes an integrity check based on CRC32 checksums. The purpose of this flag is to check for unintentional damage to CDS archive files in transmission or storage.
💀 Security check, bestie: To guarantee the security and proper operation of CDS, the user must ensure that the CDS archive files used by Java applications cannot be modified without proper authorization.
  • 🌈 -XX:SharedArchiveConfigFile=shared_config_file — Specifies additional shared data added to the archive file.
  • 🍓 -XX:SharedClassListFile=file_name — Specifies the text file that contains the names of the classes to store in the class data sharing (CDS) archive. This file contains the full name of one class per line, except slashes (/) replace dots (.). For example, to specify the classes java.lang.Object and hello.Main, create a text file that contains the following two lines:
java/lang/Object
hello/Main

The classes that you specify in this text file should include the classes that are commonly used by the application. They may include any classes from the application, extension, or bootstrap class paths.

See Application Class Data Sharing.

  • 👑 -XX:+ShowCodeDetailsInExceptionMessages — Enables printing of improved NullPointerException messages. When an application throws a NullPointerException, the option enables the JVM to analyze the program's bytecode instructions to determine precisely which reference is null, and describes the source with a null-detail message. The null-detail message is calculated and returned by NullPointerException.getMessage(), and will be printed as the exception message along with the method, filename, and line number. By default, this option is enabled.
  • 💅 -XX:+ShowMessageBoxOnError — Enables the display of a dialog box when the JVM experiences an irrecoverable error. This prevents the JVM from exiting and keeps the process active so that you can attach a debugger to it to investigate the cause of the error. By default, this option is disabled.
  • 🧚 -XX:StartFlightRecording=parameter=value — Starts a JFR recording for the Java application. This option is equivalent to the JFR.start diagnostic command that starts a recording during runtime. You can set the following parameter=value entries when starting a JFR recording:
    • delay=time — Specifies the delay between the Java application launch time and the start of the recording. Append s to specify the time in seconds, m for minutes, h for hours, or d for days (for example, specifying 10m means 10 minutes). By default, there's no delay, and this parameter is set to 0.
    • disk={true|false} — Specifies whether to write data to disk while recording. By default, this parameter is enabled.
    • dumponexit={true|false} — Specifies if the running recording is dumped when the JVM shuts down. If enabled and a filename is not entered, the recording is written to a file in the directory where the process was started. The file name is a system-generated name that contains the process ID, recording ID, and current timestamp, similar to hotspot-pid-47496-id-1-2018_01_25_19_10_41.jfr. By default, this parameter is disabled.
    • duration=time — Specifies the duration of the recording. Append s to specify the time in seconds, m for minutes, h for hours, or d for days (for example, specifying 5h means 5 hours). By default, the duration isn't limited, and this parameter is set to 0.
    • filename=path — Specifies the path and name of the file to which the recording is written when the recording is stopped, for example:
      • recording.jfr
      • /home/user/recordings/recording.jfr
      • c:\recordings\recording.jfr
    • name=identifier — Takes both the name and the identifier of a recording.
    • maxage=time — Specifies the maximum age of disk data to keep for the recording. This parameter is valid only when the disk parameter is set to true. Append s to specify the time in seconds, m for minutes, h for hours, or d for days (for example, specifying 30s means 30 seconds). By default, the maximum age isn't limited, and this parameter is set to 0s.
    • maxsize=size — Specifies the maximum size (in bytes) of disk data to keep for the recording. This parameter is valid only when the disk parameter is set to true. The value must not be less than the value for the maxchunksize parameter set with -XX:FlightRecorderOptions. Append m or M to specify the size in megabytes, or g or G to specify the size in gigabytes. By default, the maximum size of disk data isn't limited, and this parameter is set to 0.
    • path-to-gc-roots={true|false} — Specifies whether to collect the path to garbage collection (GC) roots at the end of a recording. By default, this parameter is disabled.

The path to GC roots is useful for finding memory leaks, but collecting it is time-consuming. Enable this option only when you start a recording for an application that you suspect has a memory leak. If the settings parameter is set to profile, the stack trace from where the potential leaking object was allocated is included in the information collected.

  • settings=path — Specifies the path and name of the event settings file (of type JFC). By default, the default.jfc file is used, which is located in JAVA_HOME/lib/jfr. This default settings file collects a predefined set of information with low overhead, so it has minimal impact on performance and can be used with recordings that run continuously.

A second settings file is also provided, profile.jfc, which provides more data than the default configuration, but can have more overhead and impact performance. Use this configuration for short periods of time when more information is needed.

You can specify values for multiple parameters by separating them with a comma. Event settings and .jfc options can be specified using the following syntax:

  • option=value — Specifies the option value to modify. To list available options, use the JAVA_HOME/bin/jfr tool.
  • event-setting=value — Specifies the event setting value to modify. Use the form: #=. To add a new event setting, prefix the event name with '+'.

You can specify values for multiple event settings and .jfc options by separating them with a comma. In case of a conflict between a parameter and a .jfc option, the parameter will take precedence. The whitespace delimiter can be omitted for timespan values, i.e. 20ms. For more information about the settings syntax, see Javadoc of the jdk.jfr package.

  • 🫶 -XX:ThreadStackSize=size — Sets the Java thread stack size (in kilobytes). Use of a scaling suffix, such as k, results in the scaling of the kilobytes value so that -XX:ThreadStackSize=1k sets the Java thread stack size to 1024*1024 bytes or 1 megabyte. The default value depends on the platform:
    • Linux/x64 (64-bit): 1024 KB
    • macOS (64-bit): 1024 KB
    • Windows: The default value depends on virtual memory

The following examples show how to set the thread stack size to 1 megabyte in different units:

-XX:ThreadStackSize=1k
-XX:ThreadStackSize=1024

This option is similar to -Xss. 💖

  • -XX:-UseCompressedOops — turns off compressed pointers. By default, this option is enabled, and compressed pointers are used. This will automatically limit the maximum ergonomically determined Java heap size to the maximum amount of memory that can be covered by compressed pointers. By default this range is 32 GB.

With compressed oops enabled, object references are represented as 32-bit offsets instead of 64-bit pointers, which typically increases performance when running the application with Java heap sizes smaller than the compressed oops pointer range. This option works only for 64-bit JVMs.

It's possible to use compressed pointers with Java heap sizes greater than 32 GB. See the -XX:ObjectAlignmentInBytes option.

  • -XX:-UseContainerSupport — the VM now provides automatic container detection support, which allows the VM to determine the amount of memory and number of processors that are available to a Java process running in docker containers. It uses this information to allocate system resources. This support is only available on Linux x64 platforms. If supported, the default for this flag is true, and container support is enabled by default. It can be disabled with -XX:-UseContainerSupport.

Unified Logging is available to help to diagnose issues related to this support.

Use -Xlog:os+container=trace for maximum logging of container information. See Enable Logging with the JVM Unified Logging Framework for a description of using Unified Logging.

  • -XX:+UseHugeTLBFS — Linux only: this option is the equivalent of specifying -XX:+UseLargePages. This option is disabled by default. This option pre-allocates all large pages up-front, when memory is reserved; consequently the JVM can't dynamically grow or shrink large pages memory areas; see -XX:UseTransparentHugePages if you want this behavior.

See Large Pages.

  • -XX:+UseLargePages — enables the use of large page memory. By default, this option is disabled and large page memory isn't used.

See Large Pages.

  • -XX:+UseTransparentHugePages — Linux only: enables the use of large pages that can dynamically grow or shrink. This option is disabled by default. You may encounter performance problems with transparent huge pages as the OS moves other pages around to create huge pages; this option is made available for experimentation.
  • -XX:+AllowUserSignalHandlers — enables installation of signal handlers by the application. By default, this option is disabled and the application isn't allowed to install signal handlers.
  • -XX:VMOptionsFile=filename — allows user to specify VM options in a file, for example, java -XX:VMOptionsFile=/var/my_vm_options HelloWorld.

🔥 ADVANCED JIT COMPILER OPTIONS FOR JAVA 🔥

These java options control the dynamic just-in-time (JIT) compilation performed by the Java HotSpot VM. this is the compiler-coded girlypop zone fr fr 💅

  • -XX:AllocateInstancePrefetchLines=lines — sets the number of lines to prefetch ahead of the instance allocation pointer. By default, the number of lines to prefetch is set to 1:
-XX:AllocateInstancePrefetchLines=1
  • -XX:AllocatePrefetchDistance=size — sets the size (in bytes) of the prefetch distance for object allocation. Memory about to be written with the value of new objects is prefetched up to this distance starting from the address of the last allocated object. Each Java thread has its own allocation point.

Negative values denote that prefetch distance is chosen based on the platform. Positive values are bytes to prefetch. Append the letter k or K to indicate kilobytes, m or M to indicate megabytes, or g or G to indicate gigabytes. The default value is set to -1.

The following example shows how to set the prefetch distance to 1024 bytes:

-XX:AllocatePrefetchDistance=1024
  • -XX:AllocatePrefetchInstr=instruction — sets the prefetch instruction to prefetch ahead of the allocation pointer. Possible values are from 0 to 3. The actual instructions behind the values depend on the platform. By default, the prefetch instruction is set to 0:
-XX:AllocatePrefetchInstr=0
  • -XX:AllocatePrefetchLines=lines — sets the number of cache lines to load after the last object allocation by using the prefetch instructions generated in compiled code. The default value is 1 if the last allocated object was an instance, and 3 if it was an array.

The following example shows how to set the number of loaded cache lines to 5:

-XX:AllocatePrefetchLines=5
  • -XX:AllocatePrefetchStepSize=size — sets the step size (in bytes) for sequential prefetch instructions. Append the letter k or K to indicate kilobytes, m or M to indicate megabytes, g or G to indicate gigabytes. By default, the step size is set to 16 bytes:
-XX:AllocatePrefetchStepSize=16
  • -XX:AllocatePrefetchStyle=style — sets the generated code style for prefetch instructions. The style argument is an integer from 0 to 3:
    • 0 — don't generate prefetch instructions.
    • 1 — execute prefetch instructions after each allocation. This is the default setting.
    • 2 — use the thread-local allocation block (TLAB) watermark pointer to determine when prefetch instructions are executed.
    • 3 — generate one prefetch instruction per cache line.
  • -XX:+BackgroundCompilation — enables background compilation. This option is enabled by default. To disable background compilation, specify -XX:-BackgroundCompilation (this is equivalent to specifying -Xbatch).
  • -XX:CICompilerCount=threads — sets the number of compiler threads to use for compilation. By default, the number of compiler threads is selected automatically depending on the number of CPUs and memory available for compiled code. The following example shows how to set the number of threads to 2:
-XX:CICompilerCount=2
  • -XX:+UseDynamicNumberOfCompilerThreads — dynamically create compiler thread up to the limit specified by -XX:CICompilerCount. This option is enabled by default.
  • -XX:CompileCommand=command,method[,option] — specifies a command to perform on a method. For example, to exclude the indexOf() method of the String class from being compiled, use the following:
-XX:CompileCommand=exclude,java/lang/String.indexOf

Note that the full class name is specified, including all packages and subpackages separated by a slash (/). For easier cut-and-paste operations, it's also possible to use the method name format produced by the -XX:+PrintCompilation and -XX:+LogCompilation options:

-XX:CompileCommand=exclude,java.lang.String::indexOf

If the method is specified without the signature, then the command is applied to all methods with the specified name. However, you can also specify the signature of the method in the class file format. In this case, you should enclose the arguments in quotation marks, because otherwise the shell treats the semicolon as a command end. For example, if you want to exclude only the indexOf(String) method of the String class from being compiled, use the following:

-XX:CompileCommand="exclude,java/lang/String.indexOf,(Ljava/lang/String;)I"

You can also use the asterisk (*) as a wildcard for class and method names. For example, to exclude all indexOf() methods in all classes from being compiled, use the following:

-XX:CompileCommand=exclude,*.indexOf

The commas and periods are aliases for spaces, making it easier to pass compiler commands through a shell. You can pass arguments to -XX:CompileCommand using spaces as separators by enclosing the argument in quotation marks:

-XX:CompileCommand="exclude java/lang/String indexOf"

Note that after parsing the commands passed on the command line using the -XX:CompileCommand options, the JIT compiler then reads commands from the .hotspot_compiler file. You can add commands to this file or specify a different file using the -XX:CompileCommandFile option.

To add several commands, either specify the -XX:CompileCommand option multiple times, or separate each argument with the new line separator (\n). The following commands are available:

  • break — sets a breakpoint when debugging the JVM to stop at the beginning of compilation of the specified method.
  • compileonly — excludes all methods from compilation except for the specified method. As an alternative, you can use the -XX:CompileOnly option, which lets you specify several methods.
  • dontinline — prevents inlining of the specified method.
  • exclude — excludes the specified method from compilation.
  • help — prints a help message for the -XX:CompileCommand option.
  • inline — attempts to inline the specified method.
  • log — excludes compilation logging (with the -XX:+LogCompilation option) for all methods except for the specified method. By default, logging is performed for all compiled methods.
  • option — passes a JIT compilation option to the specified method in place of the last argument (option). The compilation option is set at the end, after the method name. For example, to enable the BlockLayoutByFrequency option for the append() method of the StringBuffer class, use the following:
-XX:CompileCommand=option,java/lang/StringBuffer.append,BlockLayoutByFrequency

You can specify multiple compilation options, separated by commas or spaces.

  • print — prints generated assembler code after compilation of the specified method.
  • quiet — tells the JVM to not print the compile commands, bestie. By default the commands you pass with the -XX:CompileCommand option do get printed — like, if you exclude the indexOf() method of the String class from compilation, you get this on standard output:
CompilerOracle: exclude java/lang/String.indexOf

You can silence all that by specifying -XX:CompileCommand=quiet before your other -XX:CompileCommand options ✨

  • -XX:CompileCommandFile=filename — sets the file that JIT compiler commands get read from. Out of the box, the .hotspot_compiler file is the one used to store commands performed by the JIT compiler 🎀

Every line in the command file is a command, a class name, and a method name that the command applies to. For example, this line prints assembly code for the toString() method of the String class:

print java/lang/String toString

If you're using commands for the JIT compiler to perform on methods, then go look at the -XX:CompileCommand option 💅

Heads up: -XX:CompilerDirectivesFile has to be used together with the -XX:UnlockDiagnosticVMOptions option, the one that unlocks diagnostic JVM options 🔑

  • -XX:+CompilerDirectivesPrint — prints the directives stack when the program starts, or whenever a new directive gets added 🌸

Same deal as above: -XX:+CompilerDirectivesPrint has to be used together with the -XX:UnlockDiagnosticVMOptions option that unlocks diagnostic JVM options 🔑

  • -XX:CompileOnly=methods — sets the list of methods (comma-separated, bestie) that compilation should be restricted to. Only the methods you name get compiled. Write each one with its full class name, packages and subpackages included 💖

So like, to compile only the length() method of the String class and the size() method of the List class, you'd do:

-XX:CompileOnly=java/lang/String.length,java/util/List.size

Note that the full class name is specified, including all the packages and subpackages separated by a slash (/). And for easier cut-and-paste action, you can also use the method name format that the -XX:+PrintCompilation and -XX:+LogCompilation options produce:

-XX:CompileOnly=java.lang.String::length,java.util.List::size

Wildcards are not supported (sorry 😭), but you can specify just the class or package name to compile all methods in that class or package, and you can specify just the method to compile methods with that name in any class:

-XX:CompileOnly=java/lang/String
-XX:CompileOnly=java/lang
-XX:CompileOnly=.length
  • -XX:CompileThresholdScaling=scale — gives you unified control over first compilation ✨ This option controls when methods get first compiled, for both the tiered and the nontiered modes of operation. CompileThresholdScaling takes a floating point value between 0 and +Inf, and it scales the thresholds corresponding to the current mode of operation (both tiered and nontiered).

Setting CompileThresholdScaling to a value less than 1.0 means earlier compilation, while values greater than 1.0 delay compilation. And setting CompileThresholdScaling to 0 is equivalent to disabling compilation entirely 💀

  • -XX:+DoEscapeAnalysis — enables escape analysis. This one is enabled by default 🎀 To turn it off, specify -XX:-DoEscapeAnalysis.
  • -XX:InitialCodeCacheSize=size — sets the initial code cache size, in bytes. Append k or K for kilobytes, m or M for megabytes, or g or G for gigabytes 💜 The default value depends on the platform. Also, the initial code cache size shouldn't be less than the system's minimal memory page size. Here's how you'd set it to 32 KB:
-XX:InitialCodeCacheSize=32k
  • -XX:+Inline — enables method inlining. Enabled by default to boost performance 🔥 To disable method inlining, specify -XX:-Inline.
  • -XX:InlineSmallCode=size — sets the maximum code size, in bytes, for already compiled methods that may be inlined. This flag applies only to the C2 compiler 🦋 Append k or K for kilobytes, m or M for megabytes, or g or G for gigabytes. The default value depends on the platform and on whether tiered compilation is enabled. In this example it's set to 1000 bytes:
-XX:InlineSmallCode=1000
  • -XX:+LogCompilation — enables logging of compilation activity to a file named hotspot.log in the current working directory 📝 You can pick a different log file path and name with the -XX:LogFile option.

By default this option is disabled and compilation activity isn't logged. And -XX:+LogCompilation has to be used together with the -XX:UnlockDiagnosticVMOptions option that unlocks diagnostic JVM options 🔑

Oh, and you can get verbose diagnostic output with a message printed to the console every time a method gets compiled by using the -XX:+PrintCompilation option 👑

  • -XX:FreqInlineSize=size — sets the maximum bytecode size, in bytes, of a hot method to be inlined. C2 compiler only 🦋 Append k or K for kilobytes, m or M for megabytes, or g or G for gigabytes. The default value depends on the platform. In this example it's set to 325 bytes:
-XX:FreqInlineSize=325
  • -XX:MaxInlineSize=size — sets the maximum bytecode size, in bytes, of a cold method to be inlined. Again, C2 compiler only 🦋 Append k or K for kilobytes, m or M for megabytes, or g or G for gigabytes. By default the maximum bytecode size is set to 35 bytes:
-XX:MaxInlineSize=35
  • -XX:C1MaxInlineSize=size — sets the maximum bytecode size, in bytes, of a cold method to be inlined. This one applies only to the C1 compiler 🌈 Append k or K for kilobytes, m or M for megabytes, or g or G for gigabytes. By default the maximum bytecode size is set to 35 bytes:
-XX:MaxInlineSize=35
  • -XX:MaxTrivialSize=size — sets the maximum bytecode size, in bytes, of a trivial method to be inlined. C2 compiler only 🦋 Append k or K for kilobytes, m or M for megabytes, or g or G for gigabytes. By default the maximum bytecode size of a trivial method is set to 6 bytes:
-XX:MaxTrivialSize=6
  • -XX:C1MaxTrivialSize=size — sets the maximum bytecode size, in bytes, of a trivial method to be inlined. C1 compiler only 🌈 Append k or K for kilobytes, m or M for megabytes, or g or G for gigabytes. By default the maximum bytecode size of a trivial method is set to 6 bytes:
-XX:MaxTrivialSize=6
  • -XX:MaxNodeLimit=nodes — sets the maximum number of nodes used during compilation of a single method 🎯 By default the value depends on the features enabled. In this example the max number of nodes is set to 100,000:
-XX:MaxNodeLimit=100000
  • -XX:NonNMethodCodeHeapSize=size — sets the size in bytes of the code segment that contains nonmethod code 💜

A nonmethod code segment holds nonmethod code, like compiler buffers and the bytecode interpreter. This code type stays in the code cache forever (iconic, honestly). This flag is used only if -XX:SegmentedCodeCache is enabled ⭐

  • -XX:NonProfiledCodeHeapSize=size — sets the size in bytes of the code segment containing nonprofiled methods 🦋 This flag is used only if -XX:SegmentedCodeCache is enabled.
  • -XX:+OptimizeStringConcat — enables optimization of String concatenation operations ✨ Enabled by default. To turn it off, specify -XX:-OptimizeStringConcat.
  • -XX:+PrintAssembly — enables printing of assembly code for bytecoded and native methods, using the external hsdis-<arch>.so or .dll library 🔥 For 64-bit VM on Windows, it's hsdis-amd64.dll. This lets you actually see the generated code, which may help you diagnose performance issues 👑

By default this option is disabled and assembly code isn't printed. And -XX:+PrintAssembly has to be used together with the -XX:UnlockDiagnosticVMOptions option that unlocks diagnostic JVM options 🔑

  • -XX:ProfiledCodeHeapSize=size — sets the size in bytes of the code segment containing profiled methods 🦋 This flag is used only if -XX:SegmentedCodeCache is enabled.
  • -XX:+PrintCompilation — enables verbose diagnostic output from the JVM, printing a message to the console every single time a method is compiled 📢 This lets you see which methods actually get compiled. By default this option is disabled and no diagnostic output is printed.

You can also log compilation activity to a file with the -XX:+LogCompilation option 💖

  • -XX:+PrintInlining — enables printing of inlining decisions, so you can see which methods are getting inlined 👀

By default this option is disabled and inlining info isn't printed. And -XX:+PrintInlining has to be used together with the -XX:+UnlockDiagnosticVMOptions option that unlocks diagnostic JVM options 🔑

  • -XX:ReservedCodeCacheSize=size — sets the maximum code cache size (in bytes) for JIT-compiled code. Slap a k or K on the end for kilobytes, m or M for megabytes, or g or G for gigabytes 💖 The default max is 240 MB — but the second you disable tiered compilation with -XX:-TieredCompilation, the default drops to 48 MB. Heads up: this option has a hard limit of 2 GB, and if you go over, an error gets generated, no cap 🫶 Also, your max code cache size shouldn't be less than the initial code cache size — go peep -XX:InitialCodeCacheSize for that one.
  • -XX:RTMAbortRatio=abort_ratio — sets the RTM abort ratio as a percentage (%) of all executed RTM transactions. If the number of aborted transactions goes above this ratio, the compiled code gets deoptimized. This ratio is only used when -XX:+UseRTMDeopt is enabled. Default is 50, which literally means the compiled code gets deoptimized if 50% of all transactions are aborted 💀
  • -XX:RTMRetryCount=number_of_retries — how many times the RTM locking code gets retried when it's aborted or busy, before falling back to the normal locking mechanism. Default is 5. The -XX:UseRTMLocking option must be enabled for this, no exceptions ⭐
  • -XX:+SegmentedCodeCache — turns on segmentation of the code cache ✨ Without -XX:+SegmentedCodeCache, the code cache is just one big segment. With it, you get separate segments for non-method, profiled method, and nonprofiled method code. These segments aren't resized at runtime, so don't come for them. The feature is enabled by default if tiered compilation is enabled (-XX:+TieredCompilation) and -XX:ReservedCodeCacheSize >= 240 MB. The advantages are genuinely giving: better control of the memory footprint, reduced code fragmentation, and better iTLB/iCache behavior due to improved locality. iTLB/iCache is a CPU-specific term meaning Instruction Translation Lookaside Buffer (ITLB), and ICache is an instruction cache in the CPU 🧚 The implementation of the code cache lives in the file: /share/vm/code/codeCache.cpp.
  • -XX:StartAggressiveSweepingAt=percent — forces stack scanning of active methods to aggressively remove unused code, but only when the given percentage of the code cache is free. Default value is 10% 🍓
  • -XX:-TieredCompilation — disables the use of tiered compilation. By default this option is enabled (so yes, the minus is doing the work here, bestie).
  • -XX:UseSSE=version — enables the use of the SSE instruction set of a specified version. Defaults to the highest supported version available (x86 only) 💅
  • -XX:UseAVX=version — enables the use of the AVX instruction set of a specified version. Defaults to the highest supported version available (x86 only) 🔥
  • -XX:+UseAES — enables hardware-based AES intrinsics for hardware that supports it. This option is on by default on hardware that has the necessary instructions. -XX:+UseAES is used in conjunction with UseAESIntrinsics. Important lore: flags that control intrinsics now require the option -XX:+UnlockDiagnosticVMOptions 👑
  • -XX:+UseAESIntrinsics — enables AES intrinsics. Specifying -XX:+UseAESIntrinsics is equivalent to also enabling -XX:+UseAES. To disable hardware-based AES intrinsics, specify -XX:-UseAES -XX:-UseAESIntrinsics. For example, to enable hardware AES, use these flags:
-XX:+UseAES -XX:+UseAESIntrinsics

Flags that control intrinsics now require the option -XX:+UnlockDiagnosticVMOptions.

  • -XX:+UseAESCTRIntrinsics — analogous to -XX:+UseAESIntrinsics, it enables AES/CTR intrinsics 💜
  • -XX:+UseGHASHIntrinsics — controls the use of GHASH intrinsics. Enabled by default on platforms that support the corresponding instructions. Flags that control intrinsics now require the option -XX:+UnlockDiagnosticVMOptions 🌸
  • -XX:+UseBASE64Intrinsics — controls the use of accelerated BASE64 encoding routines for java.util.Base64. Enabled by default on platforms that support it. Flags that control intrinsics now require the option -XX:+UnlockDiagnosticVMOptions ✨
  • -XX:+UseAdler32Intrinsics — controls the use of the Adler32 checksum algorithm intrinsic for java.util.zip.Adler32. Enabled by default on platforms that support it. Flags that control intrinsics now require the option -XX:+UnlockDiagnosticVMOptions 🦋
  • -XX:+UseCRC32Intrinsics — controls the use of CRC32 intrinsics for java.util.zip.CRC32. Enabled by default on platforms that support it. Flags that control intrinsics now require the option -XX:+UnlockDiagnosticVMOptions 💖
  • -XX:+UseCRC32CIntrinsics — controls the use of CRC32C intrinsics for java.util.zip.CRC32C. Enabled by default on platforms that support it. Flags that control intrinsics now require the option -XX:+UnlockDiagnosticVMOptions 💖
  • -XX:+UseSHA — enables hardware-based intrinsics for SHA crypto hash functions on some hardware. The UseSHA option is used in conjunction with the UseSHA1Intrinsics, UseSHA256Intrinsics, and UseSHA512Intrinsics options.

The UseSHA and UseSHA*Intrinsics flags are enabled by default on machines that support the corresponding instructions.

This feature is applicable only when using the sun.security.provider.Sun provider for SHA operations. Flags that control intrinsics now require the option -XX:+UnlockDiagnosticVMOptions.

To disable all hardware-based SHA intrinsics, specify -XX:-UseSHA. To disable only a particular SHA intrinsic, use the appropriate corresponding option. For example: -XX:-UseSHA256Intrinsics 🔥

  • -XX:+UseSHA1Intrinsics — enables intrinsics for the SHA-1 crypto hash function. Flags that control intrinsics now require the option -XX:+UnlockDiagnosticVMOptions 🌟
  • -XX:+UseSHA256Intrinsics — enables intrinsics for the SHA-224 and SHA-256 crypto hash functions. Flags that control intrinsics now require the option -XX:+UnlockDiagnosticVMOptions 🌟
  • -XX:+UseSHA512Intrinsics — enables intrinsics for the SHA-384 and SHA-512 crypto hash functions. Flags that control intrinsics now require the option -XX:+UnlockDiagnosticVMOptions 🌟
  • -XX:+UseMathExactIntrinsics — enables intrinsification of various java.lang.Math.*Exact() functions. Enabled by default. Flags that control intrinsics now require the option -XX:+UnlockDiagnosticVMOptions 🎀
  • -XX:+UseMultiplyToLenIntrinsic — enables intrinsification of BigInteger.multiplyToLen(). Enabled by default on platforms that support it. Flags that control intrinsics now require the option -XX:+UnlockDiagnosticVMOptions 🎀
  • -XX:+UseSquareToLenIntrinsic — enables intrinsification of BigInteger.squareToLen(). Enabled by default on platforms that support it. Flags that control intrinsics now require the option -XX:+UnlockDiagnosticVMOptions 🎀
  • -XX:+UseMulAddIntrinsic — enables intrinsification of BigInteger.mulAdd(). Enabled by default on platforms that support it. Flags that control intrinsics now require the option -XX:+UnlockDiagnosticVMOptions 🎀
  • -XX:+UseMontgomeryMultiplyIntrinsic — enables intrinsification of BigInteger.montgomeryMultiply(). Enabled by default on platforms that support it. Flags that control intrinsics now require the option -XX:+UnlockDiagnosticVMOptions 🎀
  • -XX:+UseMontgomerySquareIntrinsic — enables intrinsification of BigInteger.montgomerySquare(). Enabled by default on platforms that support it. Flags that control intrinsics now require the option -XX:+UnlockDiagnosticVMOptions 🎀
  • -XX:+UseCMoveUnconditionally — generates CMove (scalar and vector) instructions regardless of profitability analysis. It's giving "we're doing it anyway" energy 💅
  • -XX:+UseCodeCacheFlushing — enables flushing of the code cache before shutting down the compiler. This option is enabled by default. To disable flushing of the code cache before shutting down the compiler, specify -XX:-UseCodeCacheFlushing 🌈
  • -XX:+UseCondCardMark — enables checking if the card is already marked before updating the card table. This option is disabled by default. It should be used only on machines with multiple sockets, where it increases the performance of Java applications that rely on concurrent operations 🫶
  • -XX:+UseCountedLoopSafepoints — keeps safepoints in counted loops. Its default value depends on whether the selected garbage collector requires low latency safepoints 👑
  • -XX:LoopStripMiningIter=number_of_iterations — controls the number of iterations in the inner strip mined loop. Strip mining transforms counted loops into two level nested loops. Safepoints are kept in the outer loop while the inner loop can execute at full speed. This option controls the maximum number of iterations in the inner loop. The default value is 1,000 ⭐
  • -XX:LoopStripMiningIterShortLoop=number_of_iterations — controls loop strip mining optimization. Loops with a number of iterations less than specified will not have safepoints in them. Default value is 1/10th of -XX:LoopStripMiningIter 🦋
  • -XX:+UseFMA — enables hardware-based FMA intrinsics for hardware where FMA instructions are available (such as Intel and ARM64). FMA intrinsics are generated for the java.lang.Math.fma(a, b, c) methods that calculate the value of ( a * b + c ) expressions 🔥
  • -XX:+UseRTMDeopt — autotunes RTM locking depending on the abort ratio. This ratio is specified by the -XX:RTMAbortRatio option. If the number of aborted transactions exceeds the abort ratio, then the method containing the lock is deoptimized and recompiled with all locks as normal locks. This option is disabled by default. The -XX:+UseRTMLocking option must be enabled 💀
  • -XX:+UseRTMLocking — generates Restricted Transactional Memory (RTM) locking code for all inflated locks, with the normal locking mechanism as the fallback handler. Disabled by default, bestie. And heads up: RTM options only exist on x86 CPUs that support Transactional Synchronization Extensions (TSX). 🧚

Okay so little lore drop: RTM is part of Intel's TSX, which is an x86 instruction set extension, and it helps you build multithreaded apps. RTM brings in the new instructions XBEGIN, XABORT, XEND, and XTEST. XBEGIN and XEND wrap a set of instructions to run as a transaction. If no conflict pops up while running the transaction, the memory and register modifications get committed together at the XEND instruction. XABORT can explicitly abort a transaction, and XTEST checks whether a set of instructions is being run in a transaction.

A lock on a transaction is inflated when another thread tries to access the same transaction — which blocks the thread that didn't originally request access to it. RTM requires that a fallback set of operations be specified in case a transaction aborts or fails. An RTM lock is a lock that's been delegated to the TSX's system.

RTM improves performance for highly contended locks with low conflict in a critical region (critical region = code that must not be accessed by more than one thread concurrently). RTM also improves the performance of coarse-grain locking, which typically doesn't perform well in multithreaded applications. (Quick vocab check: coarse-grain locking = holding locks for long periods to minimize the overhead of taking and releasing locks, while fine-grained locking = trying to achieve maximum parallelism by locking only when necessary and unlocking as soon as possible.) Also, for lightly contended locks used by different threads, RTM can reduce false cache line sharing — aka cache line ping-pong. That's when multiple threads from different processors access different resources, but the resources share the same cache line. So the processors keep invalidating each other's cache lines, forcing them to read from main memory instead of their cache. 💀

  • -XX:+UseSuperWord — enables the transformation of scalar operations into superword operations. Superword is a vectorization optimization, and this one is enabled by default, periodt. Want to turn it off? Specify -XX:-UseSuperWord. ✨

🩺 ADVANCED SERVICEABILITY OPTIONS FOR JAVA 🩺

These java options give you the ability to gather system information and do extensive debugging. Detective mode: activated. 🔍

  • -XX:+DisableAttachMechanism — disables the mechanism that lets tools attach to the JVM. By default this option is disabled, meaning the attach mechanism is enabled and you can use diagnostics and troubleshooting tools like jcmd, jstack, jmap, and jinfo. 🫶
Note: the tools like jcmd, jinfo, jmap, and jstack shipped with the JDK aren't supported when you use tools from one JDK version to troubleshoot a different JDK version. Keep it matching, bestie. 🙈
  • -XX:+ExtendedDTraceProbes — Linux and macOS: enables additional dtrace tool probes that affect performance. By default this option is disabled and dtrace performs only standard probes.
  • -XX:+HeapDumpOnOutOfMemoryError — enables dumping of the Java heap to a file in the current directory using the heap profiler (HPROF) when a java.lang.OutOfMemoryError exception is thrown. You can explicitly set the heap dump file path and name with the -XX:HeapDumpPath option. By default this option is disabled and the heap isn't dumped when an OutOfMemoryError exception is thrown. 😭
  • -XX:HeapDumpPath=path — sets the path and file name for writing the heap dump provided by the heap profiler (HPROF) when the -XX:+HeapDumpOnOutOfMemoryError option is set. By default the file is created in the current working directory and it's named java_pid<pid>.hprof, where <pid> is the identifier of the process that caused the error. Here's how to set the default file explicitly (%p represents the current process identifier):
-XX:HeapDumpPath=./java_pid%p.hprof
  • Linux and macOS: here's how to set the heap dump file to /var/log/java/java_heapdump.hprof:
-XX:HeapDumpPath=/var/log/java/java_heapdump.hprof
  • Windows: here's how to set the heap dump file to C:/log/java/java_heapdump.log:
-XX:HeapDumpPath=C:/log/java/java_heapdump.log
  • -XX:LogFile=path — sets the path and file name where log data is written. By default the file is created in the current working directory and it's named hotspot.log.
    • Linux and macOS: here's how to set the log file to /var/log/java/hotspot.log:
-XX:LogFile=/var/log/java/hotspot.log
  • Windows: here's how to set the log file to C:/log/java/hotspot.log:
-XX:LogFile=C:/log/java/hotspot.log
  • -XX:+PrintClassHistogram — enables printing of a class instance histogram after one of these events:
    • Linux and macOS: Control+Break
    • Windows: Control+C (SIGTERM)

By default this option is disabled. Setting this option is equivalent to running the jmap -histo command, or the jcmd pid GC.class_histogram command, where pid is the current Java process identifier. 💅

  • -XX:+PrintConcurrentLocks — enables printing of java.util.concurrent locks after one of these events:
    • Linux and macOS: Control+Break
    • Windows: Control+C (SIGTERM)

By default this option is disabled. Setting this option is equivalent to running the jstack -l command or the jcmd pid Thread.print -l command, where pid is the current Java process identifier.

  • -XX:+PrintFlagsRanges — prints the range specified and allows automatic testing of the values. See Validate Java Virtual Machine Flag Arguments. 👑
  • -XX:+PerfDataSaveToFile — if enabled, saves jstat binary data when the Java application exits. This binary data gets saved in a file named hsperfdata_pid, where pid is the process identifier of the Java application you ran. Use the jstat command to display the performance data inside this file like so:
jstat -class file:///path/hsperfdata_pid

jstat -gc file:///path/hsperfdata_pid
  • -XX:+UsePerfData — enables the perfdata feature. This option is enabled by default to allow JVM monitoring and performance testing. Disabling it suppresses the creation of the hsperfdata_userid directories. To disable the perfdata feature, specify -XX:-UsePerfData. 🌸

🗑️ ADVANCED GARBAGE COLLECTION OPTIONS FOR JAVA 🗑️

These java options control how garbage collection (GC) is performed by the Java HotSpot VM. Cleanup crew, assemble. 🧹✨

  • -XX:+AggressiveHeap — enables Java heap optimization. This sets various parameters to be optimal for long-running jobs with intensive memory allocation, based on the configuration of the computer (RAM and CPU). By default the option is disabled and the heap sizes are configured less aggressively. 🔥
  • -XX:+AlwaysPreTouch — requests the VM to touch every page on the Java heap after requesting it from the operating system and before handing memory out to the application. By default this option is disabled and all pages are committed as the application uses the heap space.
  • -XX:ConcGCThreads=threads — sets the number of threads used for concurrent GC. Sets threads to approximately 1/4 of the number of parallel garbage collection threads. The default value depends on the number of CPUs available to the JVM.

For example, to set the number of threads for concurrent GC to 2, specify this option:

-XX:ConcGCThreads=2
  • -XX:+DisableExplicitGC — enables the option that disables processing of calls to the System.gc() method. This option is disabled by default, meaning calls to System.gc() are processed. If processing of calls to System.gc() is disabled, the JVM still performs GC when necessary. No cap. 💯
  • -XX:+ExplicitGCInvokesConcurrent — enables invoking of concurrent GC by using the System.gc() request. This option is disabled by default and can be enabled only with the -XX:+UseG1GC option.
  • -XX:G1AdaptiveIHOPNumInitialSamples=number — when -XX:UseAdaptiveIHOP is enabled, this option sets the number of completed marking cycles used to gather samples until G1 adaptively determines the optimum value of -XX:InitiatingHeapOccupancyPercent. Before that, G1 uses the value of -XX:InitiatingHeapOccupancyPercent directly for this purpose. The default value is 3. 🦋
  • -XX:G1HeapRegionSize=size — sets the size of the regions into which the Java heap is subdivided when using the garbage-first (G1) collector. The value is a power of 2 and can range from 1 MB to 32 MB. The default region size is determined ergonomically based on the heap size with a goal of approximately 2048 regions.

This example sets the size of the subdivisions to 16 MB:

-XX:G1HeapRegionSize=16m
  • -XX:G1HeapWastePercent=percent — sets the percentage of heap you're willing to waste. The Java HotSpot VM doesn't initiate the mixed garbage collection cycle when the reclaimable percentage is less than the heap waste percentage. The default is 5 percent. 🍓
  • -XX:G1MaxNewSizePercent=percent — sets the percentage of the heap size to use as the maximum for the young generation size. The default value is 60 percent of your Java heap.

This is an experimental flag. This setting replaces the -XX:DefaultMaxNewGenPercent setting.

  • -XX:G1MixedGCCountTarget=number — okay so this sets the target number of mixed garbage collections after a marking cycle to collect old regions with at most G1MixedGCLiveThresholdPercent live data. The default is 8 mixed garbage collections. The goal for mixed collections is to be within this target number. 🫶
  • -XX:G1MixedGCLiveThresholdPercent=percent — this sets the occupancy threshold for an old region to be included in a mixed garbage collection cycle. The default occupancy is 85 percent.

This is an experimental flag. This setting replaces the -XX:G1OldCSetRegionLiveThresholdPercent setting. ✨

  • -XX:G1NewSizePercent=percent — sets the percentage of the heap to use as the minimum for the young generation size. The default value is 5 percent of your Java heap.

This is an experimental flag. This setting replaces the -XX:DefaultMinNewGenPercent setting. 💖

  • -XX:G1OldCSetRegionThresholdPercent=percent — sets an upper limit on the number of old regions to be collected during a mixed garbage collection cycle. The default is 10 percent of the Java heap. 🌸
  • -XX:G1ReservePercent=percent — sets the percentage of the heap (0 to 50) that’s reserved as a false ceiling to reduce the possibility of promotion failure for the G1 collector. When you increase or decrease the percentage, ensure that you adjust the total Java heap by the same amount. By default, this option is set to 10%.

The following example sets the reserved heap to 20%:

-XX:G1ReservePercent=20
  • -XX:+G1UseAdaptiveIHOP — controls adaptive calculation of the old generation occupancy to start background work preparing for an old generation collection. If enabled, G1 uses -XX:InitiatingHeapOccupancyPercent for the first few times as specified by the value of -XX:G1AdaptiveIHOPNumInitialSamples, and after that adaptively calculates a new optimum value for the initiating occupancy automatically. Otherwise, the old generation collection process always starts at the old generation occupancy determined by -XX:InitiatingHeapOccupancyPercent.

The default is enabled. 👑

  • -XX:InitialHeapSize=size — sets the initial size (in bytes) of the memory allocation pool. This value must be either 0, or a multiple of 1024 and greater than 1 MB. Append the letter k or K to indicate kilobytes, m or M to indicate megabytes, or g or G to indicate gigabytes. The default value is selected at run time based on the system configuration.

The following examples show how to set the size of allocated memory to 6 MB using various units:

-XX:InitialHeapSize=6291456
-XX:InitialHeapSize=6144k
-XX:InitialHeapSize=6m

If you set this option to 0, then the initial size is set as the sum of the sizes allocated for the old generation and the young generation. The size of the heap for the young generation can be set using the -XX:NewSize option. 💅

  • -XX:InitialRAMPercentage=percent — sets the initial amount of memory that the JVM will use for the Java heap before applying ergonomics heuristics as a percentage of the maximum amount determined as described in the -XX:MaxRAM option. The default value is 1.5625 percent.

The following example shows how to set the percentage of the initial amount of memory used for the Java heap:

-XX:InitialRAMPercentage=5
  • -XX:InitialSurvivorRatio=ratio — sets the initial survivor space ratio used by the throughput garbage collector (which is enabled by the -XX:+UseParallelGC option). Adaptive sizing is enabled by default with the throughput garbage collector by using the -XX:+UseParallelGC option, and the survivor space is resized according to the application behavior, starting with the initial value. If adaptive sizing is disabled (using the -XX:-UseAdaptiveSizePolicy option), then the -XX:SurvivorRatio option should be used to set the size of the survivor space for the entire execution of the application.

The following formula can be used to calculate the initial size of survivor space (S) based on the size of the young generation (Y), and the initial survivor space ratio (R):

S=Y/(R+2)

The 2 in the equation denotes two survivor spaces. The larger the value specified as the initial survivor space ratio, the smaller the initial survivor space size.

By default, the initial survivor space ratio is set to 8. If the default value for the young generation space size is used (2 MB), then the initial size of the survivor space is 0.2 MB.

The following example shows how to set the initial survivor space ratio to 4:

-XX:InitialSurvivorRatio=4
  • -XX:InitiatingHeapOccupancyPercent=percent — sets the percentage of the old generation occupancy (0 to 100) at which to start the first few concurrent marking cycles for the G1 garbage collector.

By default, the initiating value is set to 45%. A value of 0 implies nonstop concurrent GC cycles from the beginning until G1 adaptively sets this value.

See also the -XX:G1UseAdaptiveIHOP and -XX:G1AdaptiveIHOPNumInitialSamples options.

The following example shows how to set the initiating heap occupancy to 75%:

-XX:InitiatingHeapOccupancyPercent=75
  • -XX:MaxGCPauseMillis=time — sets a target for the maximum GC pause time (in milliseconds). This is a soft goal, and the JVM will make its best effort to achieve it. The specified value doesn’t adapt to your heap size. By default, for G1 the maximum pause time target is 200 milliseconds. The other generational collectors do not use a pause time goal by default.

The following example shows how to set the maximum target pause time to 500 ms:

-XX:MaxGCPauseMillis=500
  • -XX:MaxHeapSize=size — sets the maximum size (in bytes) of the memory allocation pool. This value must be a multiple of 1024 and greater than 2 MB. Append the letter k or K to indicate kilobytes, m or M to indicate megabytes, or g or G to indicate gigabytes. The default value is selected at run time based on the system configuration. For server deployments, the options -XX:InitialHeapSize and -XX:MaxHeapSize are often set to the same value.

The following examples show how to set the maximum allowed size of allocated memory to 80 MB using various units:

-XX:MaxHeapSize=83886080
-XX:MaxHeapSize=81920k
-XX:MaxHeapSize=80m

The -XX:MaxHeapSize option is equivalent to -Xmx. 🔥

  • -XX:MaxHeapFreeRatio=percent — sets the maximum allowed percentage of free heap space (0 to 100) after a GC event. If free heap space expands above this value, then the heap is shrunk. By default, this value is set to 70%.

Minimize the Java heap size by lowering the values of the parameters MaxHeapFreeRatio (default value is 70%) and MinHeapFreeRatio (default value is 40%) with the command-line options -XX:MaxHeapFreeRatio and -XX:MinHeapFreeRatio. Lowering MaxHeapFreeRatio to as low as 10% and MinHeapFreeRatio to 5% has successfully reduced the heap size without too much performance regression; however, results may vary greatly depending on your application. Try different values for these parameters until they’re as low as possible yet still retain acceptable performance.

-XX:MaxHeapFreeRatio=10 -XX:MinHeapFreeRatio=5

Customers trying to keep the heap small should also add the option -XX:-ShrinkHeapInSteps. See Performance Tuning Examples for a description of using this option to keep the Java heap small by reducing the dynamic footprint for embedded applications. 💜

  • -XX:MaxMetaspaceSize=size — sets the maximum amount of native memory that can be allocated for class metadata. By default, the size isn’t limited. The amount of metadata for an application depends on the application itself, other running applications, and the amount of memory available on the system.

The following example shows how to set the maximum class metadata size to 256 MB:

-XX:MaxMetaspaceSize=256m
  • -XX:MaxNewSize=size — sets the maximum size (in bytes) of the heap for the young generation (nursery). The default value is set ergonomically. 🦋
  • -XX:MaxRAM=size — sets the maximum amount of memory that the JVM may use for the Java heap before applying ergonomics heuristics. The default value is the maximum amount of available memory to the JVM process or 128 GB, whichever is lower.

The maximum amount of available memory to the JVM process is the minimum of the machine’s physical memory and any constraints set by the environment (e.g. container).

Specifying this option disables automatic use of compressed oops if the combined result of this and other options influencing the maximum amount of memory is larger than the range of memory addressable by compressed oops. See -XX:UseCompressedOops for further information about compressed oops.

The following example shows how to set the maximum amount of available memory for sizing the Java heap to 2 GB:

-XX:MaxRAM=2G
  • -XX:MaxRAMPercentage=percent — sets the maximum amount of memory that the JVM may use for the Java heap before applying ergonomics heuristics as a percentage of the maximum amount determined as described in the -XX:MaxRAM option. The default value is 25 percent.

Specifying this option disables automatic use of compressed oops if the combined result of this and other options influencing the maximum amount of memory is larger than the range of memory addressable by compressed oops. See -XX:UseCompressedOops for further information about compressed oops. 🌈

So like, the following example shows how to set the percentage of the maximum amount of memory used for the Java heap:

-XX:MaxRAMPercentage=75
  • -XX:MinRAMPercentage=percent — sets the maximum amount of memory the JVM can use for the Java heap before it applies its ergonomics heuristics, as a percentage of the maximum amount determined the way the -XX:MaxRAM option describes it, but for small heaps. And bestie, "small heap" here literally means a heap of approximately 125 MB. Default? 50 percent. 💅

The following example shows how to set the percentage of the maximum amount of memory used for the Java heap for small heaps:

-XX:MinRAMPercentage=75
  • -XX:MaxTenuringThreshold=threshold — sets the max tenuring threshold for adaptive GC sizing. Biggest value you can use is 15. Default is 15 for the parallel (throughput) collector. 🎀

The following example shows how to set the maximum tenuring threshold to 10:

-XX:MaxTenuringThreshold=10
  • -XX:MetaspaceSize=size — sets the size of the allocated class metadata space that triggers a garbage collection the first time it gets exceeded. That GC threshold then goes up or down depending on how much metadata is actually being used. The default size is platform-dependent, no cap. 🦋
  • -XX:MinHeapFreeRatio=percent — sets the minimum allowed percentage of free heap space (0 to 100) after a GC event. If free heap space drops below this, the heap expands. By default this is 40%.

Wanna minimize your Java heap size? Lower the values of the parameters MaxHeapFreeRatio (default is 70%) and MinHeapFreeRatio (default is 40%) with the command-line options -XX:MaxHeapFreeRatio and -XX:MinHeapFreeRatio. Dropping MaxHeapFreeRatio down to as low as 10% and MinHeapFreeRatio to 5% has successfully shrunk the heap size without too much performance regression — but fr fr, results vary a lot depending on your app. So try different values for these parameters until they're as low as possible while still keeping acceptable performance. ✨

-XX:MaxHeapFreeRatio=10 -XX:MinHeapFreeRatio=5

Customers trying to keep the heap small should also add the option -XX:-ShrinkHeapInSteps. See Performance Tuning Examples for a description of using this option to keep the Java heap small by reducing the dynamic footprint for embedded applications. 👑

  • -XX:MinHeapSize=size — sets the minimum size (in bytes) of the memory allocation pool. This value has to be either 0, or a multiple of 1024 and greater than 1 MB. Slap a k or K on the end for kilobytes, m or M for megabytes, g or G for gigabytes. Default value gets picked at run time based on your system config. 🌸

The following examples show how to set the mimimum size of allocated memory to 6 MB using various units:

-XX:MinHeapSize=6291456
-XX:MinHeapSize=6144k
-XX:MinHeapSize=6m

If you set this option to 0, then the minimum size gets set to the same value as the initial size. 🫶

  • -XX:NewRatio=ratio — sets the ratio between young and old generation sizes. Default is 2. The following example shows how to set the young-to-old ratio to 1:
-XX:NewRatio=1
  • -XX:NewSize=size — sets the initial size (in bytes) of the heap for the young generation (nursery). Add k or K for kilobytes, m or M for megabytes, g or G for gigabytes. 🧚

The young generation region of the heap is where new objects go. GC happens in this region more often than in other regions. If the young generation size is too low, you get a lot of minor GCs. If it's too high, then only full GCs happen, and those can take forever to finish 😭. Recommendation: keep the young generation size greater than 25% and less than 50% of the overall heap size.

The following examples show how to set the initial size of the young generation to 256 MB using various units:

-XX:NewSize=256m
-XX:NewSize=262144k
-XX:NewSize=268435456

The -XX:NewSize option is equivalent to -Xmn. ⭐

  • -XX:ParallelGCThreads=threads — sets the number of the stop-the-world (STW) worker threads. Default value depends on how many CPUs the JVM has available and which garbage collector you picked. 🍓

For example, to set the number of threads for G1 GC to 2, specify the following option:

-XX:ParallelGCThreads=2
  • -XX:+ParallelRefProcEnabled — enables parallel reference processing. By default, this option is disabled. 💀
  • -XX:+PrintAdaptiveSizePolicy — enables printing of information about adaptive-generation sizing. By default, this option is disabled.
  • -XX:+ScavengeBeforeFullGC — enables GC of the young generation before each full GC. This option is enabled by default. Please don't disable it 🙈 — scavenging the young generation before a full GC can reduce the number of objects reachable from the old generation space into the young generation space. To disable GC of the young generation before each full GC, specify the option -XX:-ScavengeBeforeFullGC.
  • -XX:SoftRefLRUPolicyMSPerMB=time — sets the amount of time (in milliseconds) a softly reachable object stays active on the heap after the last time it was referenced. Default value is one second of lifetime per free megabyte in the heap. The -XX:SoftRefLRUPolicyMSPerMB option accepts integer values representing milliseconds per one megabyte of the current heap size (for Java HotSpot Client VM) or the maximum possible heap size (for Java HotSpot Server VM). This difference means the Client VM tends to flush soft references rather than grow the heap, whereas the Server VM tends to grow the heap rather than flush soft references. In that latter case, the value of the -Xmx option has a significant effect on how quickly soft references get garbage collected. 🔥

The following example shows how to set the value to 2.5 seconds:

-XX:SoftRefLRUPolicyMSPerMB=2500
ngl, no backticks on that example in the original — it's just sitting there raw. Respect the lore. 💅
  • -XX:-ShrinkHeapInSteps — incrementally reduces the Java heap to the target size, specified by the option -XX:MaxHeapFreeRatio. This option is enabled by default. If disabled, it immediately reduces the Java heap to the target size instead of requiring multiple garbage collection cycles. Disable this option if you want to minimize the Java heap size. Heads up though: you'll likely hit performance degradation when this option is disabled. 😭

See Performance Tuning Examples for a description of using the MaxHeapFreeRatio option to keep the Java heap small by reducing the dynamic footprint for embedded applications.

  • -XX:StringDeduplicationAgeThreshold=threshold — identifies String objects reaching the specified age that are considered candidates for deduplication. An object's age is a measure of how many times it has survived garbage collection. This is sometimes called tenuring. 🎀
Note: String objects that are promoted to an old heap region before this age has been reached are always considered candidates for deduplication. The default value for this option is 3. See the -XX:+UseStringDeduplication option.
  • -XX:SurvivorRatio=ratio — sets the ratio between eden space size and survivor space size. Default is 8. The following example shows how to set the eden/survivor space ratio to 4:
-XX:SurvivorRatio=4
  • -XX:TargetSurvivorRatio=percent — sets the desired percentage of survivor space (0 to 100) used after young garbage collection. Default is 50%. 🌈

The following example shows how to set the target survivor space ratio to 30%:

-XX:TargetSurvivorRatio=30
  • -XX:TLABSize=size — sets the initial size (in bytes) of a thread-local allocation buffer (TLAB). Append k or K for kilobytes, m or M for megabytes, g or G for gigabytes. If this option is set to 0, then the JVM picks the initial size automatically. 🦋

The following example shows how to set the initial TLAB size to 512 KB:

-XX:TLABSize=512k
  • -XX:+UseAdaptiveSizePolicy — enables the use of adaptive generation sizing. This option is enabled by default. To disable adaptive generation sizing, specify -XX:-UseAdaptiveSizePolicy and set the size of the memory allocation pool explicitly. See the -XX:SurvivorRatio option. ✨
  • -XX:+UseG1GC — enables the use of the garbage-first (G1) garbage collector. It's a server-style garbage collector, targeted for multiprocessor machines with a large amount of RAM. This option meets GC pause time goals with high probability while maintaining good throughput. G1 is recommended for applications requiring large heaps (sizes of around 6 GB or larger) with limited GC latency requirements (a stable and predictable pause time below 0.5 seconds). By default, this option is enabled and G1 is used as the default garbage collector. 👑💜
  • -XX:+UseGCOverheadLimit — enables the use of a policy that limits the proportion of time the JVM spends on GC before an OutOfMemoryError exception gets thrown. This option is enabled by default, and the parallel GC will throw an OutOfMemoryError if more than 98% of the total time is spent on garbage collection and less than 2% of the heap is recovered. When the heap is small, this feature can be used to stop applications from running for long periods of time with little or no progress. To disable this option, specify the option -XX:-UseGCOverheadLimit. 🔥
  • -XX:+UseNUMA — turns on performance optimization for apps running on a machine with nonuniform memory architecture (NUMA) by making the app use more of the low-latency memory. Lowkey free performance if you're on the right hardware 🧚 By default this is disabled and no NUMA optimization happens. Heads up tho: it's only available when you're using the parallel garbage collector (-XX:+UseParallelGC).
  • -XX:+UseParallelGC — enables the parallel scavenge garbage collector, aka the throughput collector, so your app can leverage multiple processors and just go faster. By default this option is disabled and the default collector is used.
  • -XX:+UseSerialGC — enables the serial garbage collector. Honestly? Usually the best choice for small and simple applications that don't need any special functionality from garbage collection. By default this option is disabled and the default collector is used.
  • -XX:+UseSHM — Linux only! Lets the JVM use shared memory to set up large pages. See the Large Pages section for how to set those up ✨
  • -XX:+UseStringDeduplication — enables string deduplication. By default this is disabled. To use it you must enable the garbage-first (G1) garbage collector, no exceptions bestie.

So like, what does it even do? It reduces the memory footprint of String objects on the Java heap by exploiting the fact that sooo many String objects are literally identical. Instead of every String object pointing to its own character array, identical String objects can point to and share the same character array. Iconic memory efficiency behavior 💅

  • -XX:+UseTLAB — enables the use of thread-local allocation blocks (TLABs) in the young generation space. This one is enabled by default 🎀 If you wanna turn TLABs off, you specify -XX:-UseTLAB.
  • -XX:+UseZGC — enables the Z garbage collector (ZGC) 👑 This is a low latency collector, giving you max pause times of a few milliseconds, at some throughput cost. The pause times are independent of what heap size is used. Supports heap sizes from 8MB to 16TB — yes, terabytes, we're so back.
  • *-XX:ZAllocationSpikeTolerance=factor* — sets the allocation spike tolerance for ZGC. By default this is set to 2.0. What does the factor even mean? It describes the level of allocation spikes you expect. Example: a factor of 3.0 means the current allocation rate can be expected to triple at any time. Plan accordingly 🔥
  • *-XX:ZCollectionInterval=seconds* — sets the maximum interval (in seconds) between two GC cycles when using ZGC. By default it's set to 0 (aka disabled).
  • *-XX:ZFragmentationLimit=percent* — sets the maximum acceptable heap fragmentation (in percent) for ZGC. By default it's set to 25. Using a lower value causes the heap to be compacted more aggressively, so you reclaim more memory, but it costs more CPU time. Tradeoffs, always tradeoffs 😭
  • -XX:+ZProactive — enables proactive GC cycles when using ZGC. Enabled by default. ZGC will start a proactive GC cycle if doing so is expected to have minimal impact on the running application. Super useful if your app is mostly idle or allocates very few objects, but you still want to keep the heap size down and let reference processing happen even when there's a lot of free space on the heap. Vibes-based housekeeping fr fr.
  • -XX:+ZUncommit — enables uncommitting of unused heap memory when using ZGC. Enabled by default. Uncommitting that unused heap memory lowers the memory footprint of the JVM and makes that memory available for other processes to use. Sharing is caring 🫶
  • *-XX:ZUncommitDelay=seconds* — sets the amount of time (in seconds) that heap memory must have been unused before it gets uncommitted. By default it's set to 300 (5 minutes). Committing and uncommitting memory are relatively expensive operations, so a lower value means heap memory gets uncommitted earlier — at the risk of soon having to commit it again. Pick your fighter 💀

💀 DEPRECATED JAVA OPTIONS 💀

Okay so, these java options are deprecated and might be removed in a future JDK release. They're still accepted and acted upon, but a warning gets issued when you use them. Not gone, just on notice 🚨

  • -Xfuture — enables strict class-file format checks that enforce close conformance to the class-file format specification. Developers should use this flag when developing new code, because stricter checks may become the default in future releases. Get ahead of it bestie.
  • *-Xloggc:filename* — sets the file that verbose GC events info should be redirected to for logging. The -Xloggc option overrides -verbose:gc if both are given with the same java command. Note that -Xloggc:filename is replaced by -Xlog:gc:filename. See Enable Logging with the JVM Unified Logging Framework.

Example:

-Xlog:gc:garbage-collection.log
  • -XX:+FlightRecorder — enables the use of Java Flight Recorder (JFR) during the runtime of the application. Since JDK 8u40 this option has not been required to use JFR. Ancient lore 🫠
  • *-XX:InitialRAMFraction=ratio* — sets the initial amount of memory that the JVM may use for the Java heap before applying ergonomics heuristics, as a ratio of the maximum amount determined as described in the -XX:MaxRAM option. The default value is 64.

Use the option -XX:InitialRAMPercentage instead ✨

  • *-XX:MaxRAMFraction=ratio* — sets the maximum amount of memory that the JVM may use for the Java heap before applying ergonomics heuristics, as a fraction of the maximum amount determined as described in the -XX:MaxRAM option. The default value is 4.

Careful: specifying this option disables automatic use of compressed oops if the combined result of this and other options influencing the maximum amount of memory is larger than the range of memory addressable by compressed oops. See -XX:UseCompressedOops for further info about compressed oops.

Use the option -XX:MaxRAMPercentage instead ✨

  • *-XX:MinRAMFraction=ratio* — sets the maximum amount of memory that the JVM may use for the Java heap before applying ergonomics heuristics, as a fraction of the maximum amount determined as described in the -XX:MaxRAM option for small heaps. A small heap is a heap of approximately 125 MB. The default value is 2.

Use the option -XX:MinRAMPercentage instead ✨

  • -XX:+UseBiasedLocking — enables the use of biased locking. Some applications with significant amounts of uncontended synchronization may attain significant speedups with this flag enabled, but applications with certain patterns of locking may see slowdowns 😬 By default this option is disabled.

👻 OBSOLETE JAVA OPTIONS 👻

These java options are still accepted but ignored, and a warning is issued when they're used. They're literally just there for the vibes now, doing absolutely nothing 💀

🗑️ REMOVED JAVA OPTIONS 🗑️

Okay so heads up bestie — these bad boys got yeeted in JDK 17 💀 and if you try to use one, the JVM hits you with:

Unrecognized VM option option-name

That's it. That's the whole vibe check. No graceful fallback, no warning, just unrecognized. Periodt.

  • -XX:+UseMembar — this one made the VM issue membars on thread-state transitions. It was disabled by default on every platform except ARM servers, where it was turned on. RIP 🪦
  • -XX:MaxPermSize=size — set the max permanent generation space size (in bytes). Deprecated back in JDK 8, and the glow-up replacement is -XX:MaxMetaspaceSize. She's been superseded, no cap.
  • -XX:PermSize=size — set the space (in bytes) handed to the permanent generation, which triggered a garbage collection if you exceeded it. Also deprecated in JDK 8, and superseded by -XX:MetaspaceSize. Byeeee ✨
  • -XX:+TraceClassLoading — traced classes as they load. Default was disabled, meaning classes weren't traced. The replacement Unified Logging syntax is -Xlog:class+load=level — see Enable Logging with the JVM Unified Logging Framework. Use level=info for the regular tea ☕, or level=debug if you want the extra lore. And in Unified Logging syntax, -verbose:class equals -Xlog:class+load=info,class+unload=info. Iykyk.
  • -XX:+TraceClassLoadingPreorder — traced all loaded classes in the order they were referenced. Disabled by default, so no tracing happened. Replacement Unified Logging syntax is -Xlog:class+preorder=debug — see Enable Logging with the JVM Unified Logging Framework.
  • -XX:+TraceClassResolution — traced constant pool resolutions. Disabled by default, nothing got traced. Replacement Unified Logging syntax is -Xlog:class+resolve=debug — see Enable Logging with the JVM Unified Logging Framework.
  • -XX:+TraceLoaderConstraints — traced the loader constraints recording. Disabled by default, that recording wasn't traced. Replacement Unified Logging syntax is -Xlog:class+loader+constraints=info — see Enable Logging with the JVM Unified Logging Framework.

If you're feeling nostalgic (or you're debugging something ancient 👵), the full lists and descriptions of options removed in previous releases live in the Removed Java Options section of:

🎀 JAVA COMMAND-LINE ARGUMENT FILES 🎀

So here's the slay: you can shorten or simplify the java command by using @ argument files to point at one or more text files stuffed with arguments — options, class names, whatever — which then get passed to the java command. This is the moment where you can build java commands of literally any length on any operating system. No more 4,000-character one-liner that your terminal ate for breakfast 🫶

In the command line, use the at sign (@) prefix to identify an argument file that contains java options and class names. When the java command encounters a file beginning with @, it expands that file's contents into an argument list exactly as if they'd been typed out on the command line. Iconic behavior, honestly.

The java launcher keeps expanding argument file contents until it runs into the --disable-@files option. And you can use --disable-@files anywhere on the command line — including inside an argument file — to stop @ argument file expansion dead in its tracks. 🛑

Here's the syntax tea for java argument files:

  • The argument file must contain only ASCII characters, or characters in the system default encoding that's ASCII friendly, such as UTF-8.
  • The argument file size must not exceed MAXINT (2,147,483,647) bytes.
  • The launcher doesn't expand wildcards that are present within an argument file.
  • Use white space or new line characters to separate the arguments included in the file.
  • White space includes a white space character, \t, \n, \r, and \f.
    • For example, it's possible to have a path with a space, such as c:\Program Files, that can be specified as either "c:\\Program Files" or, to avoid an escape, c:\Program" "Files.
  • Any option that contains spaces — like a path component — must be within quotation marks using quotation ('"') characters in its entirety.
  • A string within quotation marks may contain the characters \n, \r, \t, and \f. They are converted to their respective ASCII codes.
  • If a file name contains embedded spaces, then put the whole file name in double quotation marks.
  • File names in an argument file are relative to the current directory, not to the location of the argument file. (Yes this trips everyone up, no you're not dumb 💖)
  • Use the number sign # in the argument file to identify comments. All characters following the # are ignored until the end of line.
  • Additional at sign @ prefixes to @ prefixed options act as an escape: the first @ is removed and the rest of the arguments are presented to the launcher literally.
  • Lines may be continued using the continuation character (\) at the end-of-line. The two lines are concatenated with the leading white spaces trimmed. To prevent trimming the leading white spaces, a continuation character (\) may be placed at the first column.
  • Because backslash (\) is an escape character, a backslash character must be escaped with another backslash character.
  • Partial quote is allowed and is closed by an end-of-file.
  • An open quote stops at end-of-line unless \ is the last character, which then joins the next line by removing all leading white space characters.
  • Wildcards (*) aren't allowed in these lists (such as specifying *.java).
  • Use of the at sign (@) to recursively interpret files isn't supported.

🌸 Example of Open or Partial Quotes in an Argument File 🌸

In the argument file,

-cp "lib/
cool/
app/
jars

this is interpreted as:

-cp lib/cool/app/jars

🌸 Example of a Backslash Character Escaped with Another Backslash Character in an Argument File 🌸

To output the following:

-cp c:\Program Files (x86)\Java\jre\lib\ext;c:\Program Files\Java\jre9\lib\ext

the backslash character must be specified in the argument file as:

-cp "c:\\Program Files (x86)\\Java\\jre\\lib\\ext;c:\\Program Files\\Java\\jre9\\lib\\ext"

🌸 Example of an EOL Escape Used to Force Concatenation of Lines in an Argument File 🌸

In the argument file,

-cp "/lib/cool app/jars:\
    /lib/another app/jars"

This is interpreted as:

-cp /lib/cool app/jars:/lib/another app/jars

🌸 Example of Line Continuation with Leading Spaces in an Argument File 🌸

In the argument file,

-cp "/lib/cool\
\app/jars???

This is interpreted as:

-cp /lib/cool app/jars

🌸 Examples of Using Single Argument File 🌸

You can use a single argument file — like myargumentfile in the example below — to hold all the java arguments you need:

java @myargumentfile

🌸 Examples of Using Argument Files with Paths 🌸

You can include relative paths in argument files, but plot twist: they're relative to the current working directory, not to the paths of the argument files themselves. In the following example, path1/options and path2/options represent argument files with different paths. Any relative paths they contain are relative to the current working directory, not to the argument files:

java @path1/options @path2/classes

🔍 CODE HEAP STATE ANALYTICS 🔍

Overview

Sometimes you just need insight into the current state of the JVM code heap, because you've got questions like:

  • Why was the JIT turned off and then on again and again?
  • Where has all the code heap space gone?
  • Why is the method sweeper not working effectively?

To give you that insight, a code heap state analytics feature exists, and it enables on-the-fly analysis of the code heap. The analytics process is split into two parts 🫶 The first part examines the entire code heap and aggregates all the information that's believed to be useful or important. The second part is several independent steps that print out the collected information, each one emphasizing a different aspect of the data. Data collection and printing both happen on an "on request" basis — no surprise work happening in the background, we love a low-maintenance queen ✨

Syntax

Requests for real-time, on-the-fly analysis get issued with this command:

jcmd pid Compiler.CodeHeap_Analytics [function] [granularity]

If you're only curious about how the code heap looks after running a sample workload, use the command line option:

-Xlog:codecache=Trace

And to see the code heap state when a "CodeCache full" condition exists, start the VM with the command line option:

-Xlog:codecache=Debug

For a detailed description of the code heap state analytics feature, the supported functions, and the granularity options, see CodeHeap State Analytics (OpenJDK): https://bugs.openjdk.java.net/secure/attachment/75649/JVM_CodeHeap_StateAnalytics_V2.pdf 📄

✨ ENABLE LOGGING WITH THE JVM UNIFIED LOGGING FRAMEWORK ✨

You use the -Xlog option to configure or enable logging with the Java Virtual Machine (JVM) unified logging framework. This is the one, bestie — the main character of JVM logging 👑

Synopsis

-Xlog[:[what][:[output][:[decorators][:output-options[,...]]]]]
  • -Xlog:directive — the big one, bestie. Here's what all the little pieces inside a directive actually mean:
    • what — picks the combo of tags and levels, written like tag1[+tag2...][*][=level][,...]. Real talk: unless you slap the wildcard (*) on there, only log messages tagged with exactly the tags you listed get matched. Peep -Xlog Tags and Levels below. ✨
    • output — sets where the logs go. Omit it and it just defaults to stdout. See -Xlog Output. 🎀
    • decorators — lets you build a custom set of decorators for that output. Omit them and you get uptime, level, and tags by default. See Decorations. 🌸
    • output-options — sets the -Xlog logging output options. 🫶
    • directive — a global option or subcommand: help, disable, async. 👑

💜 Description

Okay so the Java Virtual Machine (JVM) unified logging framework is basically one common logging system for every single component of the JVM. GC logging for the JVM got moved over to this new logging framework — the mapping of old GC flags to the matching new Xlog config lives in Convert GC Logging Flags to Xlog. And runtime logging got the same glow-up too: the mapping of legacy runtime logging flags to the corresponding new Xlog config is in Convert Runtime Logging Flags to Xlog. no cap, everything's unified now 😭💅

Here's your quick reference to the -Xlog command and the syntax for its options:

  • -Xlog — turns on JVM logging at info level. iconic default behavior.
  • -Xlog:help — prints -Xlog usage syntax plus the available tags, levels, and decorators, along with example command lines and explanations. basically the built-in tutorial bestie 🎓
  • -Xlog:disable — turns off all logging and clears all configuration of the logging framework, including the default configuration for warnings and errors. scorched earth, respectfully 💀
  • -Xlog[:option] — applies multiple arguments in the order they appear on the command line. Multiple -Xlog arguments for the same output override each other in their given order. (so like, last one wins for that output, slay)

The option is set as:

[tag-selection][:[output][:[decorators][:output-options]]]

Omitting the tag-selection defaults to a tag-set of all and a level of info. 💖

tag[+...] all

The all tag is a meta tag consisting of all tag-sets available. The asterisk * in a tag set definition denotes a wildcard tag match. Matching with a wildcard selects all tag sets that contain at least the specified tags. Without the wildcard, only exact matches of the specified tag sets get selected. precision queen 👑

output-options is:

filecount=file-count filesize=file size with optional K, M or G suffix

🎀 Default Configuration

When you pass -Xlog and nothing else on the command line, the default configuration kicks in. That default logs all messages with a level matching either warning or error, regardless of what tags the message is associated with. It's equivalent to typing this on the command line:

-Xlog:all=warning:stdout:uptime,level,tags

🦋 Controlling Logging at Runtime

Plot twist: you can also control logging at run time through Diagnostic Commands (with the jcmd utility). Literally everything you can specify on the command line can also be specified dynamically with the VM.log command. And because the diagnostic commands are automatically exposed as MBeans, you can use JMX to change the logging configuration at run time. main character flexibility fr fr ✨

🌈 -Xlog Tags and Levels

Every log message has a level and a tag set attached to it. The level corresponds to how detailed the message is, and the tag set corresponds to what the message contains or which JVM component it involves (things like, gc, jit, or os). Mapping GC flags to the Xlog configuration is described in Convert GC Logging Flags to Xlog. Mapping legacy runtime logging flags to the corresponding Xlog configuration is described in Convert Runtime Logging Flags to Xlog. 📚

Available log levels, from most off to most alarming:

  • off
  • trace
  • debug
  • info
  • warning
  • error

Available log tags:

There are literally dozens of log tags, and in the right combinations they'll enable a whole range of logging output. The full set of available log tags can be seen using -Xlog:help. Specifying all instead of a tag combination matches all tag combinations. 🌟

🍓 -Xlog Output

The -Xlog option supports these types of outputs:

  • stdout --- sends output to stdout
  • stderr --- sends output to stderr
  • file=filename --- sends output to text file(s).

When you use file=filename, specifying %p and/or %t in the file name expands to the JVM's PID and startup timestamp, respectively. You can also configure text files to handle file rotation based on file size and a number of files to rotate. For example, to rotate the log file every 10 MB and keep 5 files in rotation, specify the options filesize=10M, filecount=5. 🌀

Heads up bestie: the target size of the files isn't guaranteed to be exact — it's just an approximate value. Files are rotated by default with up to 5 rotated files of target size 20 MB, unless configured otherwise. Specifying filecount=0 means the log file shouldn't be rotated. And there's a possibility of the pre-existing log file getting overwritten. 💀

🧚 -Xlog Output Mode

By default, logging messages are output synchronously — each log message is written to the designated output when the logging call is made. But you can flip to asynchronous logging mode instead by specifying:

  • -Xlog:async — write all logging asynchronously. 🚀

In asynchronous logging mode, log sites enqueue all logging messages to an intermediate buffer, and a standalone thread is responsible for flushing them to the corresponding outputs. The intermediate buffer is bounded, and on buffer exhaustion the enqueuing message is discarded. Log entry write operations are guaranteed non-blocking. (yes, you read that right — messages can get dropped, that's the tradeoff 🙈)

The option -XX:AsyncLogBufferSize=N specifies the memory budget in bytes for the intermediate buffer. The default value should be big enough to cater for most cases. Users can provide a custom value to trade memory overhead for log accuracy if they need to. ⚖️

🌸 Decorations

Logging messages get decorated with info about the message. You can configure each output to use a custom set of decorators. The order of the output is always the same as listed below. You can configure the decorations to be used at run time. Decorations are prepended to the log message. For example:

[6.567s][info][gc,old] Old collection complete

Omitting decorators defaults to uptime, level, and tags. The none decorator is special — it's the one used to turn off all decorations.

The time (t), utctime (utc), uptime (u), timemillis (tm), uptimemillis (um), timenanos (tn), uptimenanos (un), hostname (hn), pid (p), tid (ti), level (l), and tags (tg) decorators can also be specified as none for no decoration. 🎀

  • time or t — current time and date in ISO-8601 format.
  • utctime or utc — Universal Time Coordinated or Coordinated Universal Time.
  • uptime or u — time since the start of the JVM in seconds and milliseconds. For example, 6.567s.
  • timemillis or tm — the same value as generated by System.currentTimeMillis()
  • uptimemillis or um — milliseconds since the JVM started.
  • timenanos or tn — the same value generated by System.nanoTime().
  • uptimenanos or un — nanoseconds since the JVM started.
  • hostname or hn — the host name.
  • pid or p — the process identifier.
  • tid or ti — the thread identifier.
  • level or l — the level associated with the log message.
  • tags or tg — the tag-set associated with the log message.

Convert GC Logging Flags to Xlog 🎀

Okay bestie, so you've got legacy GC logging flags just lying around? Relatable. Here's the glow-up chart: legacy flag on the left, its -Xlog era on the right, and the tea in between. 💅

  • G1PrintHeapRegions → -Xlog:gc+region=trace — Not Applicable
  • GCLogFileSize → No configuration available — Log rotation is handled by the framework.
  • NumberOfGCLogFiles → Not Applicable — Log rotation is handled by the framework.
  • PrintAdaptiveSizePolicy → -Xlog:gc+ergo*=level — Use a level of debug for most of the information, or a level of trace for all of what was logged for PrintAdaptiveSizePolicy.
  • PrintGC → -Xlog:gc — Not Applicable
  • PrintGCApplicationConcurrentTime → -Xlog:safepoint — Note that PrintGCApplicationConcurrentTime and PrintGCApplicationStoppedTime are logged on the same tag and aren't separated in the new logging.
  • PrintGCApplicationStoppedTime → -Xlog:safepoint — Note that PrintGCApplicationConcurrentTime and PrintGCApplicationStoppedTime are logged on the same tag and not separated in the new logging.
  • PrintGCCause → Not Applicable — GC cause is now always logged.
  • PrintGCDateStamps → Not Applicable — Date stamps are logged by the framework.
  • PrintGCDetails → -Xlog:gc* — Not Applicable
  • PrintGCID → Not Applicable — GC ID is now always logged.
  • PrintGCTaskTimeStamps → -Xlog:gc+task*=debug — Not Applicable
  • PrintGCTimeStamps → Not Applicable — Time stamps are logged by the framework.
  • PrintHeapAtGC → -Xlog:gc+heap=trace — Not Applicable
  • PrintReferenceGC → -Xlog:gc+ref*=debug — Note that in the old logging, PrintReferenceGC had an effect only if PrintGCDetails was also enabled.
  • PrintStringDeduplicationStatistics → -Xlog:gc+stringdedup*=debug — Not Applicable
  • PrintTenuringDistribution → -Xlog:gc+age*=level — Use a level of debug for the most relevant information, or a level of trace for all of what was logged for PrintTenuringDistribution.
  • UseGCLogFileRotation → Not Applicable — What was logged for PrintTenuringDistribution.

Convert Runtime Logging Flags to Xlog ✨

Heads up bestie 🚨 — these legacy flags are no longer recognized and will straight-up throw an error if you use them directly. So use their unified logging equivalent instead, mkay?

  • TraceExceptions → -Xlog:exceptions=info — Not Applicable
  • TraceClassLoading → -Xlog:class+load=level — Use level=info for regular information, or level=debug for additional information. In Unified Logging syntax, -verbose:class equals -Xlog:class+load=info,class+unload=info.
  • TraceClassLoadingPreorder → -Xlog:class+preorder=debug — Not Applicable
  • TraceClassUnloading → -Xlog:class+unload=level — Use level=info for regular information, or level=trace for additional information. In Unified Logging syntax, -verbose:class equals -Xlog:class+load=info,class+unload=info.
  • VerboseVerification → -Xlog:verification=info — Not Applicable
  • TraceClassPaths → -Xlog:class+path=info — Not Applicable
  • TraceClassResolution → -Xlog:class+resolve=debug — Not Applicable
  • TraceClassInitialization → -Xlog:class+init=info — Not Applicable
  • TraceLoaderConstraints → -Xlog:class+loader+constraints=info — Not Applicable
  • TraceClassLoaderData → -Xlog:class+loader+data=level — Use level=debug for regular information or level=trace for additional information.
  • TraceSafepointCleanupTime → -Xlog:safepoint+cleanup=info — Not Applicable
  • TraceSafepoint → -Xlog:safepoint=debug — Not Applicable
  • TraceMonitorInflation → -Xlog:monitorinflation=debug — Not Applicable
  • TraceBiasedLocking → -Xlog:biasedlocking=level — Use level=info for regular information, or level=trace for additional information.
  • TraceRedefineClasses → -Xlog:redefine+class*=level — level=info, debug, and trace provide increasing amounts of information.

-Xlog Usage Examples 🌸

The following are -Xlog examples. Get comfy, we're going through all of them. 🫶

-Xlog — Logs all messages by using the info level to stdout with uptime, levels, and tags decorations. This is equivalent to using:

-Xlog:all=info:stdout:uptime,levels,tags

-Xlog:gc — Logs messages tagged with the gc tag using info level to stdout. The default configuration for all other messages at level warning is in effect.

-Xlog:gc,safepoint — Logs messages tagged either with the gc or safepoint tags, both using the info level, to stdout, with default decorations. Messages tagged with both gc and safepoint won't be logged.

-Xlog:gc+ref=debug — Logs messages tagged with both gc and ref tags, using the debug level to stdout, with default decorations. Messages tagged only with one of the two tags won't be logged.

-Xlog:gc=debug:file=gc.txt:none — Logs messages tagged with the gc tag using the debug level to a file called gc.txt with no decorations. The default configuration for all other messages at level warning is still in effect.

-Xlog:gc=trace:file=gctrace.txt:uptimemillis,pids:filecount=5,file-size=1024 — Logs messages tagged with the gc tag using the trace level to a rotating file set with 5 files with size 1 MB with the base name gctrace.txt and uses decorations uptimemillis and pid.

The default configuration for all other messages at level warning is still in effect.

-Xlog:gc::uptime,tid — Logs messages tagged with the gc tag using the default 'info' level to default the output stdout and uses decorations uptime and tid. The default configuration for all other messages at level warning is still in effect.

-Xlog:gc*=info,safepoint*=off — Logs messages tagged with at least gc using the info level, but turns off logging of messages tagged with safepoint. Messages tagged with both gc and safepoint won't be logged.

-Xlog:disable -Xlog:safepoint=trace:safepointtrace.txt — Turns off all logging, including warnings and errors, and then enables messages tagged with safepoint using trace level to the file safepointtrace.txt. The default configuration doesn't apply, because the command line started with -Xlog:disable.

Complex -Xlog Usage Examples 👑

The following describes a few complex examples of using the -Xlog option. These are the main-character energy ones. 💜

-Xlog:gc+class*=debug — Logs messages tagged with at least gc and class tags using the debug level to stdout. The default configuration for all other messages at the level warning is still in effect

-Xlog:gc+meta*=trace,class*=off:file=gcmetatrace.txt — Logs messages tagged with at least the gc and meta tags using the trace level to the file metatrace.txt but turns off all messages tagged with class. Messages tagged with gc, meta, and class aren't be logged as class* is set to off. The default configuration for all other messages at level warning is in effect except for those that include class.

-Xlog:gc+meta=trace — Logs messages tagged with exactly the gc and meta tags using the trace level to stdout. The default configuration for all other messages at level warning is still be in effect.

-Xlog:gc+class+heap*=debug,meta*=warning,threads*=off — Logs messages tagged with at least gc, class, and heap tags using the trace level to stdout but only log messages tagged with meta with level. The default configuration for all other messages at the level warning is in effect except for those that include threads.

✨ VALIDATE JAVA VIRTUAL MACHINE FLAG ARGUMENTS ✨

Okay bestie, so every value you hand to any JVM command-line flag gets a whole vibe check. If the input value is invalid or out-of-range, you get an appropriate error message. No silent chaos, no mystery nonsense.

Whether they’re set ergonomically, in a command line, by an input tool, or through the APIs (for example, classes contained in the package java.lang.management), the values provided to all Java Virtual Machine (JVM) command-line flags are validated. Ergonomics are described in Java Platform, Standard Edition HotSpot Virtual Machine Garbage Collection Tuning Guide.

Range and constraints are validated either when all flags have their values set during JVM initialization or a flag’s value is changed during runtime (for example using the jcmd tool). The JVM is terminated if a value violates either the range or constraint check, and an appropriate error message is printed on the error stream. Like, it will absolutely shut the whole thing down. 💅

For example, if a flag violates a range or a constraint check, then the JVM exits with an error:

java -XX:AllocatePrefetchStyle=5 -version
intx AllocatePrefetchStyle=5 is outside the allowed range [ 0 ... 3 ]
Improperly specified VM option 'AllocatePrefetchStyle=5'
Error: Could not create the Java Virtual Machine.
Error: A fatal exception has occurred. Program will exit.

The flag -XX:+PrintFlagsRanges prints the range of all the flags. This flag allows automatic testing of the flags by the values provided by the ranges. For the flags that have the ranges specified, the type, name, and the actual range is printed in the output.

For example,

intx   ThreadStackSize [ 0 ... 9007199254740987 ] {pd product}

For the flags that don’t have the range specified, the values aren’t displayed in the print out. For example:

size_t NewSize         [   ...                  ] {product}

This helps to identify the flags that need to be implemented. The automatic testing framework can skip those flags that don’t have values and aren’t implemented. Very “we’re aware, bestie” energy. 🫶

🦋 LARGE PAGES 🦋

You use large pages, also known as huge pages, as memory pages that are significantly larger than the standard memory page size (which varies depending on the processor and operating system). Large pages optimize processor Translation-Lookaside Buffers. Iconic memory behavior, no cap. ✨

A Translation-Lookaside Buffer (TLB) is a page translation cache that holds the most-recently used virtual-to-physical address translations. A TLB is a scarce system resource. A TLB miss can be costly because the processor must then read from the hierarchical page table, which may require multiple memory accesses. By using a larger memory page size, a single TLB entry can represent a larger memory range. This results in less pressure on a TLB, and memory-intensive applications may have better performance.

However, using large pages can negatively affect system performance. This is the part where we stay serious for a sec. For example, when a large amount of memory is pinned by an application, it may create a shortage of regular memory and cause excessive paging in other applications and slow down the entire system. Also, a system that has been up for a long time could produce excessive fragmentation, which could make it impossible to reserve enough large page memory. When this happens, either the OS or JVM reverts to using regular pages.

Linux and Windows support large pages.

🌸 Large Pages Support for Linux 🌸

Linux supports large pages since version 2.6. To check if your environment supports large pages, try the following:

# cat /proc/meminfo | grep Huge
HugePages_Total: 0
HugePages_Free: 0
...
Hugepagesize: 2048 kB

If the output contains items prefixed with "Huge", then your system supports large pages. The values may vary depending on environment. The Hugepagesize field shows the default large page size in your environment, and the other fields show details for large pages of this size. Newer kernels have support for multiple large page sizes. To list the supported page sizes, run this:

# ls /sys/kernel/mm/hugepages/
hugepages-1048576kB  hugepages-2048kB

The above environment supports 2 MB and 1 GB large pages, but they need to be configured so that the JVM can use them. When using large pages and not enabling transparent huge pages (option -XX:+UseTransparentHugePages), the number of large pages must be pre-allocated. For example, to enable 8 GB of memory to be backed by 2 MB large pages, login as root and run:

# echo 4096 > /sys/kernel/mm/hugepages/hugepages-2048kB/nr_hugepages

It is always recommended to check the value of nr_hugepages after the request to make sure the kernel was able to allocate the requested number of large pages. Do the follow-up check, bestie. It’s giving responsible. 🎀

When using the option -XX:+UseSHM to enable large pages you also need to make sure the SHMMAX parameter is configured to allow large enough shared memory segments to be allocated. To allow a maximum shared segment of 8 GB, login as root and run:

# echo 8589934592 > /proc/sys/kernel/shmmax

In some environments this is not needed since the default value is large enough, but it is important to make sure the value is large enough to fit the amount of memory intended to be backed by large pages.

Note: The values contained in /proc and /sys reset after you reboot your system, so you may want to set them in an initialization script (for example, rc.local or sysctl.conf). Reboot amnesia is real. 😭

If you configure the OS kernel parameters to enable use of large pages, the Java processes may allocate large pages for the Java heap as well as other internal areas, for example:

  • Code cache
  • Marking bitmaps

Consequently, if you configure the nr_hugepages parameter to the size of the Java heap, then the JVM can still fail to allocate the heap using large pages because other areas such as the code cache might already have used some of the configured large pages. Plot twist: the code cache ate some of your pages. 💀

🎀 Large Pages Support for Windows 🎀

To use large pages support on Windows, the administrator must first assign additional privileges to the user who is running the application:

  1. Select Control Panel, Administrative Tools, and then Local Security Policy.
  2. Select Local Policies and then User Rights Assignment.
  3. Double-click Lock pages in memory, then add users and/or groups.
  4. Reboot your system.

Note that these steps are required even if it’s the administrator who’s running the application, because administrators by default don’t have the privilege to lock pages in memory. Yes, even the admin. Especially the admin. Periodt. 🫶

💖 APPLICATION CLASS DATA SHARING 💖

okay so bestie, Application Class Data Sharing (AppCDS) is the whole vibe of storing the classes your apps use inside an archive file. Since these classes live in a format that can be loaded very quickly compared to classes stored in a JAR file, AppCDS can improve your app’s start-up time. It can also reduce runtime memory footprint by sharing parts of these classes across multiple processes. slay 💅

Classes in the CDS archive are stored in an optimized format that’s about 2 to 5 times larger than classes stored in JAR files or the JDK runtime image. So it’s a good idea to archive only the classes your app actually uses — usually that’s just a small portion of all available classes. Like, your app might only use a few APIs from a huge library. 🎀

✨ Using CDS Archives ✨

By default, in most JDK distributions, unless -Xshare:off is specified, the JVM starts up with a default CDS archive. That archive is usually in JAVA_HOME/lib/server/classes.jsa (or JAVA_HOME\bin\server\classes.jsa on Windows). It contains about 1300 core library classes that most applications use. 👑

To use CDS for the exact set of classes your app uses, you can use the -XX:SharedArchiveFile option. Its general form is:

-XX:SharedArchiveFile=<static_archive>:<dynamic_archive>
  • The <static_archive> overrides the default CDS archive.
  • The <dynamic_archive> provides additional classes that can be loaded on top of those in the <static_archive>.
  • On Windows, that path delimiter : should be replaced with ;.

The names "static" and "dyanmic" (yes, that spelling) are used for historical reasons — the only significance is that the "static" archive is loaded first and the "dynamic" archive is loaded second. 💀

The JVM can use up to two archives. To use only a single <static_archive>, you can omit the <dynamic_archive> portion:

-XX:SharedArchiveFile=<static_archive>

For convenience, the <dynamic_archive> records the location of the <static_archive>. So you can omit the <static_archive> and say only:

-XX:SharedArchiveFile=<dynamic_archive>

🦋 Creating CDS Archives 🦋

CDS archives can be created with several methods:

  • -Xshare:dump
  • -XX:ArchiveClassesAtExit
  • jcmd VM.cds

One common operation in all these methods is a “trial run”, where you run the application once to determine the classes that should be stored in the archive. literally the vibe check first 😌

🌸 Creating a Static CDS Archive File with -Xshare:dump 🌸

The following steps create a static CDS archive file that contains all the classes used by the test.Hello application.

  1. Create a list of all classes used by the test.Hello application. The following command creates a file named hello.classlist that contains a list of all classes used by this application:
java -Xshare:off -XX:DumpLoadedClassList=hello.classlist -cp hello.jar test.Hello

The classpath specified by the -cp parameter must contain only JAR files.

  1. Create a static archive, named hello.jsa, that contains all the classes in hello.classlist:
java -Xshare:dump -XX:SharedArchiveFile=hello.jsa -XX:SharedClassListFile=hello.classlist -cp hello.jar
  1. Run the application test.Hello with the archive hello.jsa:
java -XX:SharedArchiveFile=hello.jsa -cp hello.jar test.Hello
  1. Optional: Verify that the test.Hello application is using the class contained in the hello.jsa shared archive:
java -XX:SharedArchiveFile=hello.jsa -cp hello.jar -Xlog:class+load test.Hello

The output of this command should contain the following text:

[info][class,load] test.Hello source: shared objects file

🧚 Creating a Dynamic CDS Archive File with -XX:SharedArchiveFile 🧚

Advantages of dynamic CDS archives are:

  • They usually use less disk space, since they don’t need to store the classes that are already in the static archive.
  • They are created with one fewer step than the comparable static archive.

The following steps create a dynamic CDS archive file that contains the classes used by the test.Hello application, excluding those already in the default CDS archive. 🫶

  1. Create a dynamic CDS archive, named hello.jsa, that contains all the classes in hello.jar loaded by the application test.Hello:
java -XX:ArchiveClassesAtExit=hello.jsa -cp hello.jar Hello
  1. Run the application test.Hello with the shared archive hello.jsa:
java -XX:SharedArchiveFile=hello.jsa -cp hello.jar test.Hello
  1. Optional: Repeat step 4 of the previous section to verify that the test.Hello application is using the class contained in the hello.jsa shared archive.

It’s also possible to create a dynamic CDS archive with a non-default static CDS archive. E.g.:

java -XX:SharedArchiveFile=base.jsa -XX:ArchiveClassesAtExit=hello.jsa -cp hello.jar Hello

To run the application using this dynamic CDS archive:

java -XX:SharedArchiveFile=base.jsa:hello.jsa -cp hello.jar Hello

(On Windows, the above path delimiter : should be replaced with ;.)

As mentioned above, the name of the static archive can be skipped:

java -XX:SharedArchiveFile=hello.jsa -cp hello.jar Hello

💜 Creating CDS Archive Files with jcmd 💜

The previous two sections require you to modify the application’s start-up script in order to create a CDS archive. Sometimes that could be difficult — for example, if the application’s class path is set up by complex routines. 😭

The jcmd VM.cds command provides a less intrusive way to create a CDS archive by connecting to a running JVM process. You can create either a static:

jcmd <pid> VM.cds static_dump my_static_archive.jsa

or a dynamic archive:

jcmd <pid> VM.cds dynamic_dump my_dynamic_archive.jsa

To use the resulting archive file in a subsequent run of the application without modifying the application’s start-up script, you can use the following technique:

env JAVA_TOOL_OPTIONS=-XX:SharedArchiveFile=my_static_archive.jsa bash app_start.sh
Note: to use jcmd <pid> VM.cds dynamic_dump, the JVM process identified by <pid> must be started with -XX:+RecordDynamicDumpInfo, which can also be passed to the application start-up script with the same technique:
env JAVA_TOOL_OPTIONS=-XX:+RecordDynamicDumpInfo bash app_start.sh

🎀 Restrictions on Class Path and Module Path 🎀

  • Neither the class path (-classpath and -Xbootclasspath/a) nor the module path (--module-path) can contain non-empty directories.
  • Only modular JAR files are supported in --module-path. Exploded modules are not supported.
  • The class path used at archive creation time must be the same as (or a prefix of) the class path used at run time. (There’s no such requirement for the module path.)
  • The CDS archive cannot be loaded if any JAR files in the class path or module path are modified after the archive is generated.
  • If any of the VM options --upgrade-module-path, --patch-module or --limit-modules are specified, CDS is disabled. This means that the JVM will execute without loading any CDS archives. In addition, if you try to create a CDS archive with any of these 3 options specified, the JVM will report an error.

🔥 PERFORMANCE TUNING EXAMPLES 🔥

You can use the Java advanced runtime options to optimize the performance of your applications. performance queen behavior 👑

🌈 Tuning for Higher Throughput 🌈

Use the following commands and advanced options to achieve higher throughput performance for your application:

java -server -XX:+UseParallelGC -XX:+UseLargePages -Xmn10g  -Xms26g -Xmx26g

⭐ Tuning for Lower Response Time ⭐

Use the following commands and advanced options to achieve lower response times for your application:

java -XX:+UseG1GC -XX:MaxGCPauseMillis=100

🍓 Keeping the Java Heap Small and Reducing the Dynamic Footprint of Embedded Applications 🍓

Use the following advanced runtime options to keep the Java heap small and reduce the dynamic footprint of embedded applications:

-XX:MaxHeapFreeRatio=10 -XX:MinHeapFreeRatio=5
Note: The defaults for these two options are 70% and 40% respectively. Because performance sacrifices can occur when using these small settings, you should optimize for a small footprint by reducing these settings as much as possible without introducing unacceptable performance degradation.

✨ EXIT STATUS ✨

The following exit values are typically returned by the launcher when the launcher is called with the wrong arguments, serious errors, or exceptions thrown by the JVM. However, a Java application may choose to return any value by using the API call System.exit(exitValue). The values are:

  • 0 — Successful completion
  • >0 — An error occurred

This page is from JDK 17, dated 2021 — that’s the source/version/date vibe check, bestie. Go archive something iconic 💅✨🌸

📦 where this page comes from