diff options
author | Lukas Rytz <lukas.rytz@gmail.com> | 2016-06-01 10:20:59 +0200 |
---|---|---|
committer | Lukas Rytz <lukas.rytz@gmail.com> | 2016-06-01 10:20:59 +0200 |
commit | 20dd825ec6b5d4015ce36cf4373ba1c1d917da94 (patch) | |
tree | 8853ed0b590f37607fdedaed36a74ca6fada34c8 /test/benchmarks | |
parent | 697158414f563a9b43ada696a1bb949eab96208b (diff) | |
parent | cba585d41b9c2c47f256cbce45115bb205ae58a2 (diff) | |
download | scala-20dd825ec6b5d4015ce36cf4373ba1c1d917da94.tar.gz scala-20dd825ec6b5d4015ce36cf4373ba1c1d917da94.tar.bz2 scala-20dd825ec6b5d4015ce36cf4373ba1c1d917da94.zip |
Merge commit 'cba585d' into merge-2.11-to-2.12-june-1
Diffstat (limited to 'test/benchmarks')
-rw-r--r-- | test/benchmarks/.gitignore | 14 | ||||
-rw-r--r-- | test/benchmarks/README.md | 105 | ||||
-rw-r--r-- | test/benchmarks/build.sbt | 11 | ||||
-rw-r--r-- | test/benchmarks/project/plugins.sbt | 2 | ||||
-rw-r--r-- | test/benchmarks/src/main/scala/benchmark/JmhRunner.scala | 16 | ||||
-rw-r--r-- | test/benchmarks/src/main/scala/scala/collection/mutable/OpenHashMapBenchmark.scala | 204 | ||||
-rw-r--r-- | test/benchmarks/src/main/scala/scala/collection/mutable/OpenHashMapRunner.scala | 105 |
7 files changed, 457 insertions, 0 deletions
diff --git a/test/benchmarks/.gitignore b/test/benchmarks/.gitignore new file mode 100644 index 0000000000..ce4d893417 --- /dev/null +++ b/test/benchmarks/.gitignore @@ -0,0 +1,14 @@ +/project/project/ +/project/target/ +/target/ + +# what appears to be a Scala IDE-generated file +.cache-main + +# standard Eclipse output directory +/bin/ + +# sbteclipse-generated Eclipse files +/.classpath +/.project +/.settings/ diff --git a/test/benchmarks/README.md b/test/benchmarks/README.md new file mode 100644 index 0000000000..370d610bc4 --- /dev/null +++ b/test/benchmarks/README.md @@ -0,0 +1,105 @@ +# Scala library benchmarks + +This directory is a standalone SBT project, within the Scala project, +that makes use of the [SBT plugin](https://github.com/ktoso/sbt-jmh) for [JMH](http://openjdk.java.net/projects/code-tools/jmh/). + +## Running a benchmark + +The benchmarks require first building Scala into `../../build/pack` with `ant`. +If you want to build with `sbt dist/mkPack` instead, +you'll need to change `scalaHome` in this project. + +You'll then need to know the fully-qualified name of the benchmark runner class. +The benchmarking classes are organized under `src/main/scala`, +in the same package hierarchy as the classes that they test. +Assuming that we're benchmarking `scala.collection.mutable.OpenHashMap`, +the benchmark runner would likely be named `scala.collection.mutable.OpenHashMapRunner`. +Using this example, one would simply run + + jmh:runMain scala.collection.mutable.OpenHashMapRunner + +in SBT. +SBT should be run _from this directory_. + +The JMH results can be found under `target/jmh-results/`. +`target` gets deleted on an SBT `clean`, +so you should copy these files out of `target` if you wish to preserve them. + +## Creating a benchmark and runner + +The benchmarking classes use the same package hierarchy as the classes that they test +in order to make it easy to expose, in package scope, members of the class under test, +should that be necessary for benchmarking. + +There are two types of classes in the source directory: +those suffixed `Benchmark` and those suffixed `Runner`. +The former are benchmarks that can be run directly using `jmh:run`; +however, they are normally run from a corresponding class of the latter type, +which is run using `jmh:runMain` (as described above). +This …`Runner` class is useful for setting appropriate JMH command options, +and for processing the JMH results into files that can be read by other tools, such as Gnuplot. + +The `benchmark.JmhRunner` trait should be woven into any runner class, for the standard behavior that it provides. +This includes creating output files in a subdirectory of `target/jmh-results` +derived from the fully-qualified package name of the `Runner` class. + +## Some useful HotSpot options +Adding these to the `jmh:run` or `jmh:runMain` command line may help if you're using the HotSpot (Oracle, OpenJDK) compiler. +They require prefixing with `-jvmArgs`. +See [the Java documentation](http://docs.oracle.com/javase/8/docs/technotes/tools/unix/java.html) for more options. + +### Viewing JIT compilation events +Adding `-XX:+PrintCompilation` shows when Java methods are being compiled or deoptimized. +At the most basic level, +these messages will tell you whether the code that you're measuring is still being tuned, +so that you know whether you're running enough warm-up iterations. +See [Kris Mok's notes](https://gist.github.com/rednaxelafx/1165804#file-notes-md) to interpret the output in detail. + +### Consider GC events +If you're not explicitly performing `System.gc()` calls outside of your benchmarking code, +you should add the JVM option `-verbose:gc` to understand the effect that GCs may be having on your tests. + +### "Diagnostic" options +These require the `-XX:+UnlockDiagnosticVMOptions` JVM option. + +#### Viewing inlining events +Add `-XX:+PrintInlining`. + +#### Viewing the disassembled code +If you're running OpenJDK or Oracle JVM, +you may need to install the disassembler library (`hsdis-amd64.so` for the `amd64` architecture). +In Debian, this is available in +<a href="https://packages.debian.org/search?keywords=libhsdis0-fcml">the `libhsdis0-fcml` package</a>. +For an Oracle (or other compatible) JVM not set up by your distribution, +you may also need to copy or link the disassembler library +to the `jre/lib/`_`architecture`_ directory inside your JVM installation directory. + +To show the assembly code corresponding to the code generated by the JIT compiler for specific methods, +add `-XX:CompileCommand=print,scala.collection.mutable.OpenHashMap::*`, +for example, to show all of the methods in the `scala.collection.mutable.OpenHashMap` class. + +To show it for _all_ methods, add `-XX:+PrintAssembly`. +(This is usually excessive.) + +## Useful reading +* [OpenJDK advice on microbenchmarks](https://wiki.openjdk.java.net/display/HotSpot/MicroBenchmarks) +* Brian Goetz's "Java theory and practice" articles: + * "[Dynamic compilation and performance measurement](http://www.ibm.com/developerworks/java/library/j-jtp12214/)" + * "[Anatomy of a flawed benchmark](http://www.ibm.com/developerworks/java/library/j-jtp02225/)" +* [Doug Lea's JSR 166 benchmarks](http://gee.cs.oswego.edu/cgi-bin/viewcvs.cgi/jsr166/src/test/loops/) +* "[Measuring performance](http://docs.scala-lang.org/overviews/parallel-collections/performance.html)" of Scala parallel collections + +## Legacy frameworks + +An older version of the benchmarking framework is still present in this directory, in the following locations: + +<dl> +<dt><code>bench</code></dt> +<dd>A script to run the old benchmarks.</dd> +<dt><code>source.list</code></dt> +<dd>A temporary file used by <code>bench</code>.</dd> +<dt><code>src/scala/</code></dt> +<dd>The older benchmarks, including the previous framework.</dd> +</dl> + +Another, older set of benchmarks is present in `../benchmarking/`. diff --git a/test/benchmarks/build.sbt b/test/benchmarks/build.sbt new file mode 100644 index 0000000000..4806ecdde8 --- /dev/null +++ b/test/benchmarks/build.sbt @@ -0,0 +1,11 @@ +scalaHome := Some(file("../../build/pack")) +scalaVersion := "2.11.8" +scalacOptions ++= Seq("-feature", "-Yopt:l:classpath") + +lazy val root = (project in file(".")). + enablePlugins(JmhPlugin). + settings( + name := "test-benchmarks", + version := "0.0.1", + libraryDependencies += "org.openjdk.jol" % "jol-core" % "0.4" + ) diff --git a/test/benchmarks/project/plugins.sbt b/test/benchmarks/project/plugins.sbt new file mode 100644 index 0000000000..e11aa29f3b --- /dev/null +++ b/test/benchmarks/project/plugins.sbt @@ -0,0 +1,2 @@ +addSbtPlugin("com.typesafe.sbteclipse" % "sbteclipse-plugin" % "4.0.0") +addSbtPlugin("pl.project13.scala" % "sbt-jmh" % "0.2.6") diff --git a/test/benchmarks/src/main/scala/benchmark/JmhRunner.scala b/test/benchmarks/src/main/scala/benchmark/JmhRunner.scala new file mode 100644 index 0000000000..cc75be529d --- /dev/null +++ b/test/benchmarks/src/main/scala/benchmark/JmhRunner.scala @@ -0,0 +1,16 @@ +package benchmark + +import java.io.File + +/** Common code for JMH runner objects. */ +trait JmhRunner { + private[this] val parentDirectory = new File("target", "jmh-results") + + /** Return the output directory for this class, creating the directory if necessary. */ + protected def outputDirectory: File = { + val subdir = getClass.getPackage.getName.replace('.', File.separatorChar) + val dir = new File(parentDirectory, subdir) + if (!dir.isDirectory) dir.mkdirs() + dir + } +} diff --git a/test/benchmarks/src/main/scala/scala/collection/mutable/OpenHashMapBenchmark.scala b/test/benchmarks/src/main/scala/scala/collection/mutable/OpenHashMapBenchmark.scala new file mode 100644 index 0000000000..26e26b3065 --- /dev/null +++ b/test/benchmarks/src/main/scala/scala/collection/mutable/OpenHashMapBenchmark.scala @@ -0,0 +1,204 @@ +package scala.collection.mutable; + +import java.util.concurrent.TimeUnit +import org.openjdk.jmh.annotations._ +import org.openjdk.jmh.infra.Blackhole +import org.openjdk.jmh.infra.BenchmarkParams +import org.openjdk.jol.info.GraphLayout +import org.openjdk.jol.info.GraphWalker +import org.openjdk.jol.info.GraphVisitor +import org.openjdk.jmh.infra.IterationParams +import org.openjdk.jmh.runner.IterationType + +/** Utilities for the [[OpenHashMapBenchmark]]. + * + * The method calls are tested by looping to the size desired for the map; + * instead of using the JMH harness, which iterates for a fixed length of time. + */ +private object OpenHashMapBenchmark { + /** State container for the `put()` bulk calling tests. + * + * Provides an array of adequately-sized, empty maps to each invocation, + * so that hash table allocation won't be done during measurement. + * Provides enough maps to make each invocation long enough to avoid timing artifacts. + * Performs a GC after re-creating the empty maps before every invocation, + * so that only the GCs caused by the invocation contribute to the measurement. + * + * Records the memory used by all the maps in the last invocation of each iteration. + */ + @State(Scope.Thread) + @AuxCounters + class BulkPutState { + /** A lower-bound estimate of the number of nanoseconds per `put()` call */ + private[this] val nanosPerPut: Double = 5 + + /** Minimum number of nanoseconds per invocation, so as to avoid timing artifacts. */ + private[this] val minNanosPerInvocation = 1000000 // one millisecond + + /** Size of the maps created in this trial. */ + private[this] var size: Int = _ + + /** Total number of entries in all of the `maps` combined. */ + var mapEntries: Int = _ + + /** Number of operations performed in the current invocation. */ + var operations: Int = _ + + /** Bytes of memory used in the object graphs of all the maps. */ + var memory: Long = _ + + var maps: Array[OpenHashMap[Int,Int]] = null + + @Setup + def threadSetup(params: BenchmarkParams) { + size = params.getParam("size").toInt + val n = math.ceil(minNanosPerInvocation / (nanosPerPut * size)).toInt + mapEntries = size * n + maps = new Array(n) + } + + @Setup(Level.Iteration) + def iterationSetup { + operations = 0 + } + + @Setup(Level.Invocation) + def setup(params: IterationParams) { + for (i <- 0 until maps.length) maps(i) = new OpenHashMap[Int,Int](size) + + if (params.getType == IterationType.MEASUREMENT) { + operations += mapEntries + System.gc() // clean up after last invocation + } + } + + @TearDown(Level.Iteration) + def iterationTeardown(params: IterationParams) { + if (params.getType == IterationType.MEASUREMENT) { + // limit to smaller cases to avoid OOM + memory = if (mapEntries <= 1000000) GraphLayout.parseInstance(maps(0), maps.tail).totalSize else 0 + } + } + } + + /** State container for the `get()` bulk calling tests. + * + * Provides a thread-scoped map of the expected size. + * Performs a GC after loading the map. + */ + @State(Scope.Thread) + class BulkGetState { + val map = new OpenHashMap[Int,Int].empty + + /** Load the map with keys from `1` to `size`. */ + @Setup + def setup(params: BenchmarkParams) { + val size = params.getParam("size").toInt + put_Int(map, 1, size) + System.gc() + } + } + + /** State container for the `get()` bulk calling tests with deleted entries. + * + * Provides a thread-scoped map of the expected size, from which entries have been removed. + * Performs a GC after loading the map. + */ + @State(Scope.Thread) + class BulkRemovedGetState { + val map = new OpenHashMap[Int,Int].empty + + /** Load the map with keys from `1` to `size`, removing half of them. */ + @Setup + def setup(params: BenchmarkParams) { + val size = params.getParam("size").toInt + put_remove_Int(map, size) + System.gc() + } + } + + /** Put elements into the given map. */ + private def put_Int(map: OpenHashMap[Int,Int], from: Int, to: Int) { + var i = from + while (i <= to) { // using a `for` expression instead adds significant overhead + map.put(i, i) + i += 1 + } + } + + /** Put elements into the given map, removing half of them as they're added. + * + * @param size number of entries to leave in the map on return + */ + def put_remove_Int(map: OpenHashMap[Int,Int], size: Int) { + val blocks = 50 // should be a factor of `size` + val totalPuts = 2 * size // add twice as many, because we remove half of them + val blockSize: Int = totalPuts / blocks + var base = 0 + while (base < totalPuts) { + put_Int(map, base + 1, base + blockSize) + + // remove every other entry + var i = base + 1 + while (i <= base + blockSize) { + map.remove(i) + i += 2 + } + + base += blockSize + } + } + + /** Get elements from the given map. */ + def get_Int(map: OpenHashMap[Int,Int], size: Int, bh: Blackhole) { + var i = 1 + while (i <= size) { + bh.consume(map.get(i).getOrElse(0)) + i += 1 + } + } +} + +/** Benchmark for the library's [[OpenHashMap]]. */ +@BenchmarkMode(Array(Mode.AverageTime)) +@Fork(6) +@Threads(1) +@Warmup(iterations = 20) +@Measurement(iterations = 6) +@OutputTimeUnit(TimeUnit.NANOSECONDS) +@State(Scope.Benchmark) +class OpenHashMapBenchmark { + import OpenHashMapBenchmark._ + + @Param(Array("25", "50", "100", "250", "1000", "2500", "10000", "25000", "100000", "250000", "1000000", "2500000", + "5000000", "7500000", "10000000", "25000000")) + var size: Int = _ + + /** Test putting elements to a map of `Int` to `Int`. */ + @Benchmark + def put_Int(state: BulkPutState) { + var i = 0 + while (i < state.maps.length) { + OpenHashMapBenchmark.put_Int(state.maps(i), 1, size) + i += 1 + } + } + + /** Test putting and removing elements to a growing map of `Int` to `Int`. */ + @Benchmark + def put_remove_Int(state: BulkPutState) { + var i = 0 + while (i < state.maps.length) { + OpenHashMapBenchmark.put_remove_Int(state.maps(i), size) + i += 1 + } + } + + /** Test getting elements from a map of `Int` to `Int`. */ + @Benchmark + def put_get_Int(state: BulkGetState, bh: Blackhole) = OpenHashMapBenchmark.get_Int(state.map, size, bh) + + /** Test getting elements from a map of `Int` to `Int` from which elements have been removed. */ + @Benchmark + def put_remove_get_Int(state: BulkRemovedGetState, bh: Blackhole) = OpenHashMapBenchmark.get_Int(state.map, size, bh) +} diff --git a/test/benchmarks/src/main/scala/scala/collection/mutable/OpenHashMapRunner.scala b/test/benchmarks/src/main/scala/scala/collection/mutable/OpenHashMapRunner.scala new file mode 100644 index 0000000000..1a58b18ee9 --- /dev/null +++ b/test/benchmarks/src/main/scala/scala/collection/mutable/OpenHashMapRunner.scala @@ -0,0 +1,105 @@ +package scala.collection.mutable + +import java.io.BufferedWriter +import java.io.File +import java.io.FileOutputStream +import java.io.OutputStreamWriter +import java.io.PrintWriter +import scala.collection.JavaConversions +import scala.language.existentials +import org.openjdk.jmh.results.RunResult +import org.openjdk.jmh.runner.Runner +import org.openjdk.jmh.runner.options.CommandLineOptions +import org.openjdk.jmh.runner.options.Options +import benchmark.JmhRunner +import org.openjdk.jmh.runner.options.OptionsBuilder +import org.openjdk.jmh.runner.options.VerboseMode +import org.openjdk.jmh.results.Result + +/** Replacement JMH application that runs the [[OpenHashMap]] benchmark. + * + * Outputs the results in a form consumable by a Gnuplot script. + */ +object OpenHashMapRunner extends JmhRunner { + /** File that will be created for the output data set. */ + private[this] val outputFile = new File(outputDirectory, "OpenHashMap.dat") + + /** Qualifier to add to the name of a memory usage data set. */ + private[this] val memoryDatasetQualifier = "-memory" + + private[this] implicit class MyRunResult(r: RunResult) { + /** Return the dataset label. */ + def label = r.getPrimaryResult.getLabel + + /** Return the value of the JMH parameter for the number of map entries per invocation. */ + def size: String = r.getParams.getParam("size") + + /** Return the operation counts. */ + def operations = Option(r.getSecondaryResults.get("operations")) + + /** Return the number of map entries. */ + def entries = r.getSecondaryResults.get("mapEntries") + + /** Return the memory usage. */ + def memory = Option(r.getSecondaryResults.get("memory")) + } + + /** Return the statistics of the given result as a string. */ + private[this] def stats(r: Result[_]) = r.getScore + " " + r.getStatistics.getStandardDeviation + + + def main(args: Array[String]) { + import scala.collection.JavaConversions._ + import scala.language.existentials + + val opts = new CommandLineOptions(args: _*) + var builder = new OptionsBuilder().parent(opts).jvmArgsPrepend("-Xmx6000m") + if (!opts.verbosity.hasValue) builder = builder.verbosity(VerboseMode.SILENT) + + val results = new Runner(builder.build).run() + + // Sort the results + + /** Map from data set name to data set. */ + val datasetByName = Map.empty[String, Set[RunResult]] + + /** Ordering for the results within a data set. Orders by increasing number of map entries. */ + val ordering = Ordering.by[RunResult, Int](_.size.toInt) + + def addToDataset(key: String, result: RunResult): Unit = + datasetByName.getOrElseUpdate(key, SortedSet.empty(ordering)) += result + + results.foreach { result => + addToDataset(result.label, result) + + // Create another data set for trials that track memory usage + if (result.memory.isDefined) + addToDataset(result.label + memoryDatasetQualifier, result) + } + + //TODO Write out test parameters + // val jvm = params.getJvm + // val jvmArgs = params.getJvmArgs.mkString(" ") + + val f = new PrintWriter(outputFile, "UTF-8") + try { + datasetByName.foreach(_ match { case (label: String, dataset: Iterable[RunResult]) => { + f.println(s"# [$label]") + + val isMemoryUsageDataset = label.endsWith(memoryDatasetQualifier) + dataset.foreach { r => + f.println(r.size + " " + ( + if (isMemoryUsageDataset) + stats(r.entries) + " " + stats(r.memory.get) + else + stats(r.operations getOrElse r.getPrimaryResult) + )) + } + + f.println(); f.println() // data set separator + }}) + } finally { + f.close() + } + } +} |