Scalabench

Scalabench is a utility for running microbenchmarks in Scala. It is here to replace the functions you write over and over again in your REPL, not to replace your profiling tool.

Usage

Import

The framework is available through a package object:

import com.rossabaker.scalabench._

time

Executes a block. Returns a case class with its result and the number of nanoseconds it took to run.

scala> val xs = scala.util.Random.shuffle(1 to 10000 toList)
xs: List[Int] = List(7600, 5918, 1862, 1266, 331, 4342, 1798,

scala> time { xs.sorted }                                   
res0: TimedResult[List[Int]] = TimedResult(List(1, 2, ..., 10000),92164651) 

benchmark

Takes a number of untimed warmup trials, a number of timed trials, and a block. Returns a list of the nanoseconds elapsed to execute the block per timed trial.

scala> benchmark(100,1000) { xs.sorted }  
res1: List[Long] = List(2254514, 2321843, 2386586, 2286572, 2251094, ...)

report

Takes the output of benchmark, and prints some statistics on the results:

scala> report(res1)
Mean:          0.005289402 s
Min:           0.002092340 s
25th %ile:     0.002224026 s
Median:        0.002305149 s
75th %ile:     0.002403889 s
Max:           1.167523561 s

TODO

• Standard deviation
• Outlier detection
• Parallel execution of trials