Merge pull request #859 from bwignall/bmk_compiling

Compile benchmark directory

benchmark/src/main/scala/breeze/benchmark/BreezeBenchmark.scala

@@ -18,12 +18,15 @@
 
 package breeze.benchmark
 
-import com.google.caliper.{Runner, SimpleBenchmark}
+import breeze.stats.distributions.RandBasis
+import com.google.caliper.Benchmark
+import com.google.caliper.runner.CaliperMain
 
 /**
  * Extend this to create an actual benchmarking class.
  */
-trait BreezeBenchmark extends SimpleBenchmark {
+trait BreezeBenchmark {
+  implicit val randBasis: RandBasis = RandBasis.mt0
 
   /**
    * Sugar to run 'f' for 'reps' number of times.
@@ -68,6 +71,6 @@ trait BreezeBenchmark extends SimpleBenchmark {
 /**
  * Extend this to create a main object which will run 'cls' (a benchmark).
  */
-abstract class MyRunner(val cls: java.lang.Class[_ <: com.google.caliper.Benchmark]) {
-  def main(args: Array[String]): Unit = Runner.main(cls, args: _*)
+abstract class MyRunner(val cls: java.lang.Class[_]) {
+  def main(args: Array[String]): Unit = CaliperMain.main(cls, args)
 }

benchmark/src/main/scala/breeze/linalg/DenseMatrixBenchmark.scala

@@ -8,7 +8,7 @@ import breeze.macros._
 object DenseMatrixBenchmark extends MyRunner(classOf[DenseMatrixBenchmark])
 
 trait BuildsRandomMatrices {
-  private val uniform = Uniform(0, 1)
+  private val uniform = Uniform(0, 1)(RandBasis.mt0)
   def randomMatrix(m: Int, n: Int, transpose: Boolean = false): DenseMatrix[Double] = {
     if (!transpose) {
       DenseMatrix.rand[Double](m, n)
@@ -52,7 +52,7 @@ class DenseMatrixBenchmark extends BreezeBenchmark with BuildsRandomMatrices {
 //    sin(dm)
 //  }
 
-  def timeIntMatrixMultiply(reps: Int) = runWith(reps, randomIntMatrix(2500, 2500)): Unit = { dm =>
+  def timeIntMatrixMultiply(reps: Int) = runWith(reps, randomIntMatrix(2500, 2500)) { dm =>
     dm * dm
   }
 }

benchmark/src/main/scala/breeze/linalg/DenseVectorBenchmark.scala

@@ -7,14 +7,14 @@ import breeze.macros._
 object DenseVectorBenchmark extends MyRunner(classOf[DenseVectorBenchmark])
 
 trait BuildsRandomVectors {
-  private val uniform = Uniform(0, 1)
+  private val uniform = Uniform(0, 1)(RandBasis.mt0)
   def randomArray(size: Int, offset: Int = 0, stride: Int = 1): DenseVector[Double] = {
     require(offset >= 0)
     require(stride >= 1)
     val result = new DenseVector(new Array[Double](offset + stride * size), offset, stride, size)
     var i = 0
     while (i < size) {
-      result.unsafeUpdate(i, uniform.draw())
+      result.update(i, uniform.draw())
       i += 1
     }
     result
@@ -84,21 +84,6 @@ class DenseVectorBenchmark extends BreezeBenchmark with BuildsRandomVectors {
   def timeValueAt(reps: Int) = valueAtBench(reps, 1024 * 8, 1)
   def timeValueAtStride4(reps: Int) = valueAtBench(reps, 1024 * 8, 4)
 
-  def unsafeValueAtBench(reps: Int, size: Int, stride: Int) =
-    runWith(reps, { randomArray(size, stride = stride) })(arr => {
-      var i = 0
-      var t: Double = 0
-      while (i < arr.size) {
-        t += arr
-          .unsafeValueAt(i) //This is not strictly part of the benchmark, but done so that the JIT doensn't eliminate everything
-        i += 1
-      }
-      t
-    })
-
-  def timeUnsafeValueAt(reps: Int) = unsafeValueAtBench(reps, 1024 * 8, 1)
-  def timeUnsafeValueAtStride4(reps: Int) = unsafeValueAtBench(reps, 1024 * 8, 4)
-
   def updateBench(reps: Int, size: Int, stride: Int) =
     runWith(reps, { randomArray(size, stride = stride) })(arr => {
       var i = 0
@@ -116,7 +101,7 @@ class DenseVectorBenchmark extends BreezeBenchmark with BuildsRandomVectors {
     runWith(reps, { randomArray(size, stride = stride) })(arr => {
       var i = 0
       while (i < arr.size) {
-        arr.unsafeUpdate(i, i.toDouble)
+        arr.update(i, i.toDouble)
         i += 1
       }
       arr

benchmark/src/main/scala/breeze/linalg/SparseVectorBenchmark.scala

@@ -23,7 +23,7 @@ import breeze.benchmark._
 object SparseVectorBenchmark extends MyRunner(classOf[SparseVectorBenchmark])
 
 class SparseVectorBenchmark extends BreezeBenchmark with BuildsRandomVectors {
-  def timeAllocate(reps: Int) = run(reps): Unit = {
+  def timeAllocate(reps: Int) = run(reps) {
     SparseVector.zeros[Double](1024)
   }
 

benchmark/src/main/scala/breeze/linalg/support/CanTraverseValues.scala

@@ -23,7 +23,7 @@ class CanTraverseValuesBenchmark extends BreezeBenchmark with BuildsRandomVector
     sum(arr)
   })
 
-  def timePrimitiveSum(reps: Int) = runWith(reps, {randomArray(1024 * 8)}): Unit = { arr =>
+  def timePrimitiveSum(reps: Int) = runWith(reps, {randomArray(1024 * 8)}){ arr =>
     val d = arr.data
     var sum = 0.0
     import breeze.macros._
@@ -37,7 +37,7 @@ class CanTraverseValuesBenchmark extends BreezeBenchmark with BuildsRandomVector
     sum(arr(0 to -1 by 5))
   })
 
-  def timePrimitiveSumStrided(reps: Int) = runWith(reps, {randomArray(1024 * 8 * 5)}): Unit = { arr =>
+  def timePrimitiveSumStrided(reps: Int) = runWith(reps, {randomArray(1024 * 8 * 5)}){ arr =>
     val d = arr.data
     var sum = 0.0
     import breeze.macros._
@@ -49,27 +49,27 @@ class CanTraverseValuesBenchmark extends BreezeBenchmark with BuildsRandomVector
    */
 
   /*
-  def timeSumMatrix(reps: Int) = runWith(reps, {randomMatrix(1024, 40)}): Unit = { arr =>
+  def timeSumMatrix(reps: Int) = runWith(reps, {randomMatrix(1024, 40)}){ arr =>
     sum(arr)
   }
 
-  def timeSumMatrixRows(reps: Int) = runWith(reps, {randomMatrix(1024, 40)}): Unit = { arr =>
+  def timeSumMatrixRows(reps: Int) = runWith(reps, {randomMatrix(1024, 40)}){ arr =>
     sum(arr(*, ::))
   }
 
-  def timeSumMatrixRowsLoop(reps: Int) = runWith(reps, {randomMatrix(1024, 40)}): Unit = { arr =>
+  def timeSumMatrixRowsLoop(reps: Int) = runWith(reps, {randomMatrix(1024, 40)}){ arr =>
     val result = DenseVector.zeros[Double](1024)
     for (i <- 0 until arr.cols) {
       result += arr(::, i)
     }
     result
   }
 
-  def timeSumMatrixCols(reps: Int) = runWith(reps, {randomMatrix(40, 1024)}): Unit = { arr =>
+  def timeSumMatrixCols(reps: Int) = runWith(reps, {randomMatrix(40, 1024)}){ arr =>
     sum(arr(::, *))
   }
 
-  def timeSumMatrixColsLoop(reps: Int) = runWith(reps, {randomMatrix(40, 1024)}): Unit = { arr =>
+  def timeSumMatrixColsLoop(reps: Int) = runWith(reps, {randomMatrix(40, 1024)}){ arr =>
     val result = DenseVector.zeros[Double](1024)
     for (i <- 0 until arr.rows) {
       result += arr(i, ::).t
@@ -78,19 +78,19 @@ class CanTraverseValuesBenchmark extends BreezeBenchmark with BuildsRandomVector
   }
    */
 
-  def timeMaxMatrixCols(reps: Int) = runWith(reps, { randomMatrix(40, 1024) }): Unit = { arr =>
+  def timeMaxMatrixCols(reps: Int) = runWith(reps, { randomMatrix(40, 1024) }) { arr =>
     max(arr(::, *))
   }
 
-  def timeMaxMatrixRows(reps: Int) = runWith(reps, { randomMatrix(40, 1024) }): Unit = { arr =>
+  def timeMaxMatrixRows(reps: Int) = runWith(reps, { randomMatrix(40, 1024) }) { arr =>
     max(arr(*, ::))
   }
 
-  def timeMinMatrixCols(reps: Int) = runWith(reps, { randomMatrix(40, 1024) }): Unit = { arr =>
+  def timeMinMatrixCols(reps: Int) = runWith(reps, { randomMatrix(40, 1024) }) { arr =>
     min(arr(::, *))
   }
 
-  def timeMinMatrixRows(reps: Int) = runWith(reps, { randomMatrix(40, 1024) }): Unit = { arr =>
+  def timeMinMatrixRows(reps: Int) = runWith(reps, { randomMatrix(40, 1024) }) { arr =>
     max(arr(*, ::))
   }
 

benchmark/src/main/scala/breeze/polynomial/DensePolynomialBenchmark.scala

@@ -4,6 +4,7 @@ import breeze.benchmark._
 import breeze.linalg.BuildsRandomVectors
 import breeze.stats.distributions._
 
+import algebra.instances.all.doubleAlgebra
 import spire.math._
 import spire.math.poly._
 import breeze.macros._

benchmark/src/main/scala/breeze/stats/ProbMonad.scala

@@ -18,7 +18,7 @@ class ProbMonadBenchmark extends BreezeBenchmark {
 
   val size = 1024 * 1024
 
-  def timeMonad(reps: Int) = run(reps): Unit = {
+  def timeMonad(reps: Int) = run(reps) {
     /* The purpose of this benchmark is to compare monadic usage of rand to non-monadic usage (see timeRaw).
      */
     val monadic = for {
@@ -28,7 +28,7 @@ class ProbMonadBenchmark extends BreezeBenchmark {
     monadic.samplesVector(size)
   }
 
-  def timeRaw(reps: Int) = run(reps): Unit = {
+  def timeRaw(reps: Int) = run(reps) {
     /* The purpose of this benchmark is to compare monadic usage of rand to non-monadic usage (see timeMonad).
      */
     val nonmonadic = new Rand[Double] {
@@ -37,45 +37,45 @@ class ProbMonadBenchmark extends BreezeBenchmark {
     nonmonadic.samplesVector(size)
   }
 
-  def timeMap(reps: Int) = run(reps): Unit = {
+  def timeMap(reps: Int) = run(reps) {
     val mg = gaussian.map(f)
     mg.samplesVector(size)
   }
 
-  def timeMapRepeated(reps: Int) = run(reps): Unit = {
+  def timeMapRepeated(reps: Int) = run(reps) {
     val mg = gaussian.map(f).map(f2).map(f3)
     mg.samplesVector(size)
   }
 
-  def timeFlatMap(reps: Int) = run(reps): Unit = {
+  def timeFlatMap(reps: Int) = run(reps) {
     val mg = gaussian.flatMap(fm)
     mg.samplesVector(size)
   }
 
-  def timeFlatMapRepeated(reps: Int) = run(reps): Unit = {
+  def timeFlatMapRepeated(reps: Int) = run(reps) {
     val mg = gaussian.flatMap(fm).flatMap(fm).flatMap(fm)
     mg.samplesVector(size)
   }
 
-  def timeCondition(reps: Int) = run(reps): Unit = {
+  def timeCondition(reps: Int) = run(reps) {
     val mg = gaussian.condition(x => x > 0)
     mg.samplesVector(size)
   }
 
-  def timeRepeatCondition(reps: Int) = run(reps): Unit = {
+  def timeRepeatCondition(reps: Int) = run(reps) {
     val mg = gaussian.condition(x => x > 0).condition(x => x < 1).condition(x => x > -1)
     mg.samplesVector(size)
   }
 
-  def timeDrawOpt(reps: Int) = run(reps): Unit = {
+  def timeDrawOpt(reps: Int) = run(reps) {
     val mg = gaussian.condition(x => x > 0)
     val result = new Array[Option[Double]](size)
     cforRange(0 until size)(i => {
       result(i) = mg.drawOpt()
     })
     result
   }
-  def timeDrawOptMultipleCondition(reps: Int) = run(reps): Unit = {
+  def timeDrawOptMultipleCondition(reps: Int) = run(reps) {
     val mg = gaussian.condition(x => x > 0).condition(x => x < 1).condition(x => x > -1)
     val result = new Array[Option[Double]](size)
     cforRange(0 until size)(i => {

benchmark/src/main/scala/breeze/stats/mcmc/MetropolisHastings.scala

@@ -42,13 +42,13 @@ class MetropolisHastingsBenchmark extends BreezeBenchmark {
     result
   }
 
-  def timeMarkovChainEquiv(reps: Int) = run(reps): Unit = {
+  def timeMarkovChainEquiv(reps: Int) = run(reps) {
     val m =
       ArbitraryMetropolisHastings(likelihood _, gaussianJump _, gaussianJumpLogProb _, 0.5, burnIn = 0, dropCount = 0)
     pullAllSamples(m)
   }
 
-  def timeMetropolisHastings(reps: Int) = run(reps): Unit = {
+  def timeMetropolisHastings(reps: Int) = run(reps) {
     val m = ArbitraryMetropolisHastings(
       likelihood _,
       (_: Double) => Uniform(0, 1),
@@ -59,7 +59,7 @@ class MetropolisHastingsBenchmark extends BreezeBenchmark {
     pullAllSamples(m)
   }
 
-  def timeMetropolisHastingsWithWork(reps: Int) = run(reps): Unit = {
+  def timeMetropolisHastingsWithWork(reps: Int) = run(reps) {
     val m = ArbitraryMetropolisHastings(
       likelihood _,
       (_: Double) => Uniform(0, 1),
@@ -70,7 +70,7 @@ class MetropolisHastingsBenchmark extends BreezeBenchmark {
     pullAllSamplesWithWork(m)
   }
 
-  def timeThreadedBufferedWithWork(reps: Int) = run(reps): Unit = {
+  def timeThreadedBufferedWithWork(reps: Int) = run(reps) {
     val wrapped = ArbitraryMetropolisHastings(
       likelihood _,
       (_: Double) => Uniform(0, 1),