This plugin integrates the JMH micro-benchmarking framework with Gradle.
Build script snippet for use in all Gradle versions:
buildscript {
repositories {
jcenter()
maven {
url "https://plugins.gradle.org/m2/"
}
}
dependencies {
classpath "me.champeau.gradle:jmh-gradle-plugin:0.3.1"
}
}
apply plugin: "me.champeau.gradle.jmh"
Build script snippet for new, incubating, plugin mechanism introduced in Gradle 2.1:
plugins {
id "me.champeau.gradle.jmh" version "0.3.1"
}
The plugin makes it easy to integrate into an existing project thanks to a specific configuration. In particular,
benchmark source files are expected to be found in the src/jmh
directory:
src/jmh |- java : java sources for benchmarks |- resources : resources for benchmarks
The plugin creates a jmh
configuration that you should use if your benchmark files depend on a 3rd party library.
For example, if you want to use commons-io
, you can add the dependency like this:
dependencies {
jmh 'commons-io:commons-io:2.4'
}
The plugin uses JMH 0.9. You can upgrade the version just by changing the version in the dependencies
block:
dependencies {
jmh 'org.openjdk.jmh:jmh-core:0.9'
jmh 'org.openjdk.jmh:jmh-generator-annprocess:0.9'
}
The project will add several tasks:
-
jmhClasses
: compiles raw benchmark code -
jmhRunBytecodeGenerator
: runs bytecode generator over raw benchmark code and generates actual benchmarks -
jmhCompileGeneratedClasses
: compiles generated benchmarks -
jmhJar
: builds the JMH jar containing the JHM runtime and your compiled benchmark classes -
jmh
: executes the benchmarks
The jmh
task is the main task and depends on the others so it is in general sufficient to execute this task:
gradle jmh
By default, all benchmarks will be executed, and the results will be generated into $buildDir/reports/jmh
. But you
can change various options thanks to the jmh
configuration block. All configurations variables apart from include
are unset, implying that they fall back to the default JMH values:
jmh {
include = ['some regular expression'] // include pattern (regular expression) for benchmarks to be executed
exclude = ['some regular expression'] // exclude pattern (regular expression) for benchmarks to be executed
iterations = 10 // Number of measurement iterations to do.
benchmarkMode = ['thrpt','ss'] // Benchmark mode. Available modes are: [Throughput/thrpt, AverageTime/avgt, SampleTime/sample, SingleShotTime/ss, All/all]
batchSize = 1 // Batch size: number of benchmark method calls per operation. (some benchmark modes can ignore this setting)
fork = 2 // How many times to forks a single benchmark. Use 0 to disable forking altogether
failOnError = false // Should JMH fail immediately if any benchmark had experienced the unrecoverable error?
forceGC = false // Should JMH force GC between iterations?
jvm = 'myjvm' // Custom JVM to use when forking.
jvmArgs = ['Custom JVM args to use when forking.']
jvmArgsAppend = ['Custom JVM args to use when forking (append these)']
jvmArgsPrepend =[ 'Custom JVM args to use when forking (prepend these)']
humanOutputFile = project.file("${project.buildDir}/reports/jmh/human.txt") // human-readable output file
resultsFile = project.file("${project.buildDir}/reports/jmh/results.txt") // results file
operationsPerInvocation = 10 // Operations per invocation.
benchmarkParameters = [:] // Benchmark parameters.
profilers = [] // Use profilers to collect additional data. Supported profilers: [cl, comp, gc, stack, perf, perfnorm, perfasm, xperf, xperfasm, hs_cl, hs_comp, hs_gc, hs_rt, hs_thr]
timeOnIteration = '1s' // Time to spend at each measurement iteration.
resultFormat = 'CSV' // Result format type (one of CSV, JSON, NONE, SCSV, TEXT)
synchronizeIterations = false // Synchronize iterations?
threads = 4 // Number of worker threads to run with.
threadGroups = [2,3,4] //Override thread group distribution for asymmetric benchmarks.
timeout = '1s' // Timeout for benchmark iteration.
timeUnit = 'ms' // Output time unit. Available time units are: [m, s, ms, us, ns].
verbosity = 'NORMAL' // Verbosity mode. Available modes are: [SILENT, NORMAL, EXTRA]
warmup = '1s' // Time to spend at each warmup iteration.
warmupBatchSize = 10 // Warmup batch size: number of benchmark method calls per operation.
warmupForks = 0 // How many warmup forks to make for a single benchmark. 0 to disable warmup forks.
warmupIterations = 1 // Number of warmup iterations to do.
warmupMode = 'INDI' // Warmup mode for warming up selected benchmarks. Warmup modes are: [INDI, BULK, BULK_INDI].
warmupBenchmarks = ['.*Warmup'] // Warmup benchmarks to include in the run in addition to already selected. JMH will not measure these benchmarks, but only use them for the warmup.
zip64 = true // Use ZIP64 format for bigger archives
jmhVersion = '1.15' // Specifies JMH version
includeTests = true // Allows to include test sources into generate JMH jar, i.e. use it when benchmarks depend on the test classes.
duplicateClassesStrategy = 'fail' // Strategy to apply when encountring duplicate classes during creation of the fat jar (i.e. while executing jmhJar task)
}
The following table describes the mappings between JMH’s command line options and the plugin’s extension properties.
JMH Option | Extension Property |
---|---|
-bm <mode> |
benchmarkMode |
-bs <int> |
batchSize |
-e <regexp+> |
exclude |
-f <int> |
fork |
-foe <bool> |
failOnError |
-gc <bool> |
forceGC |
-i <int> |
iterations |
-jvm <string> |
jvm |
-jvmArgs <string> |
jvmArgs |
-jvmArgsAppend <string> |
jvmArgsAppend |
-jvmArgsPrepend <string> |
jvmArgsPrepend |
-o <filename> |
humanOutputFile |
-opi <int> |
operationsPerInvocation |
-p <param={v,}*> |
benchmarkParameters? |
-prof <profiler> |
profilers |
-r <time> |
timeOnIteration |
-rf <type> |
resultFormat |
-rff <filename> |
resultsFile |
-si <bool> |
synchronizeIterations |
-t <int> |
threads |
-tg <int+> |
threadGroups |
-to <time> |
timeout |
-tu <TU> |
timeUnit |
-v <mode> |
verbosity |
-w <time> |
warmup |
-wbs <int> |
warmupBatchSize |
-wf <int> |
warmupForks |
-wi <int> |
warmupIterations |
-wm <mode> |
warmupMode |
-wmb <regexp+> |
warmupBenchmarks |
The jmh
plugin makes it easy to test existing sources without having to create a separate project for this. This is
the reason why you must put your benchmark source files into src/jmh/java
instead of src/main/java
. This means that
by default, the jmh
(benchmarks) task depends on your main
(production) source set.
It is possible a dependency on the test
source set by setting property includeTests
to true inside jmh
block.
Optionally it is possible to use Shadow Plugin to do actual JMH jar
creation. The configuration of Shadow Plugin for JMH jar is done via jmhJar
block.
For example:
jmhJar {
append('META-INF/spring.handlers')
append('META-INF/spring.schemas')
exclude 'LICENSE'
}
This plugin will merge all dependencies that are defined as part of jmh
, runtime
and optionally testRuntime
configurations into a single set from which fat jar will be created when executing jmhJar
task. This is done to ensure
that no duplicate dependencies will be added the generated jar.
In addition plugin applies DuplicatesStrategy
defined via duplicateClassesStrategy
extension property to every class while creating fat jar. By default this
property is set to DuplicatesStrategy.FAIL
which means that upon detection of
duplicate classes the task will fail.
It is possible to change this behavior by configuring duplicateClassesStrategy
property via jmh
block, e.g.:
jmh {
duplicateClassesStrategy = 'warn'
}
However if you do encounter problem with defaut value it means that the classpath or sources in your project do contain duplicate classes which means that it is not possible to predict which one will be used when fat jar will generated.
To deal with duplicate files other than classes use Shadow Plugin capabilities, see Using JMH Gradle Plugin with Shadow Plugin.