This is the JMH fork that implements dynamic reconfiguration to stop warmup iterations and forks when the results are stable.

Dynamic reconfiguration is the prototype implementation proposed in the FSE'2020 paper "Dynamically Reconfiguring Software Microbenchmarks: Reducing Execution Time Without Sacrificing Result Quality" by Christoph Laaber, Stefan Würsten, Harald C. Gall, and Philipp Leitner [1]. For details, please check out the paper available at http:https://t.uzh.ch/13k.

Dynamic reconfiguration extends JMH 1.21 by an additional benchmark mode, three stoppage criteria and corresponding thresholds, minimum number of warmup iterations, warmup forks, and forks, and changes the semantic of warmup iterations and forks to be an upper bound.

The dynamic reconfiguration prototype adds the following additional parameters to JMH 1.21:

Java annotation:

@BenchmarkMode(Mode.Reconfigure)

Command line parameter:

-bm Reconfigure

or

-bm re

Dynamic reconfiguration supports three stoppage criteria (also called reconfigure modes) based on the coefficient of variatio, the Kullback-Leibler divergence, and bootstrap relative confidence interval widths.

Java annotation:

@Reconfigure(MODE)

where MODE is one of ReconfigureMode.COV, ReconfigureMode.CI, or ReconfigureMode.DIVERGENCE.

Command line argument:

-rm <mode>

Reconfigure mode. Available modes are: [COV/cov, CI/ci, DIVERGENCE/kld]. (default: DIVERGENCE)

Coefficient of variation threshold defaults to 0.01.

Java annotation:

@Reconfigure(ReconfigureMode.COV, covThreshold = <double>)

Command line argument:

-rcov <double>

Kullback-Leibler divergence threshold defaults to 0.99.

Java annotation:

@Reconfigure(ReconfigureMode.DIVERGENCE, kldThreshold = <double>)

Command line argument:

-rkld <double>

Relative confidence interval width threshold defaults to 0.03.

Java annotation:

@Reconfigure(ReconfigureMode.CI, kldThreshold = <double>)

Command line argument:

-rci <double>

Minimum number of warmup iterations. Warmup iterations are not counted towards the benchmark score. Default value is 5.

Java annotation:

@Warmup(minIterations = <int>)

Command line argument:

-mwi <int>

Minimum number of warmup forks defaults to 0.

Java annotation:

@Fork(minWarmups = <int>)

Command line argument:

-mwf <int>

Minimum number of forks defaults to 2.

Java annotation:

@Fork(minValue = <int>)

Command line argument:

-mf <int>

JMH with dynamic reconfiguration must be installed from source. To do so, run the following command from your command line:

mvn clean install -DskipTests

JMH with dynamic reconfiguration can then be used in your JVM projects, by adding the dependency to your build script. Set groupId to org.openjdk.jmh, artifactId to jmh-core, and version to 1.21-Reconfigure.

This protoype relies on the Go tool pa for checking the CI stoppage criteria. It already includes three pre-compiled pa versions for Windows, Linux, and macOS for the amd64 processor architecture. If a different operating system or architecture is required, you have to compile pa from source. For this, clone pa and run the following command:

GOOS=<os> GOARCH=<arch> go build

where <os> and <arch> are supported by the Go tool chain.

Next you need to copy the generated pa binary to jmh-core/src/main/resources/ and adapt the method executableName in org.openjdk.jmh.reconfigure.statistics.ci.CIHelper to return your pa binary.

[1] Christoph Laaber, Stefan Würsten, Harald C. Gall, Philipp Leitner (2020). Dynamically Reconfiguring Software Microbenchmarks: Reducing Execution Time Without Sacrificing Result Quality. In Proceedings of the 2020 ACM Joint European Software Engineering Conference and Symposium on the Foundations of Software Engineering (ESEC/FSE).