park-bench: A quick-and-dirty, low-friction benchmark tool with immediate feedback

This is a package candidate release! Here you can preview how this package release will appear once published to the main package index (which can be accomplished via the 'maintain' link below). Please note that once a package has been published to the main package index it cannot be undone! Please consult the package uploading documentation for more information.

[maintain] [Publish]

A quick-and-dirty, low-friction benchmark tool with immediate feedback.


[Skip to Readme]

Properties

Versions 0.1.0, 0.1.0, 0.1.0.1
Change log CHANGELOG.md
Dependencies base (>=4.12 && <4.17), bytestring (>=0.10 && <0.12), text (>=1.1 && <1.3) [details]
License BSD-3-Clause
Copyright Copyright (C) 2020-2021 Mitchell Rosen, Travis Staton
Author Mitchell Rosen, Travis Staton
Maintainer Mitchell Rosen <mitchellwrosen@gmail.com>, Travis Staton <hello@travisstaton.com>
Category Concurrency
Home page https://github.com/awkward-squad/park-bench
Bug tracker https://github.com/awkward-squad/park-bench/issues
Source repo head: git clone https://github.com/awkward-squad/park-bench
Uploaded by mitchellwrosen at 2021-12-18T04:31:04Z

Modules

[Index] [Quick Jump]

Downloads

Maintainer's Corner

Package maintainers

For package maintainers and hackage trustees


Readme for park-bench-0.1.0

[back to package description]

Overview

park-bench is a quick-and-dirty benchmarking tool for comparing the performance of Haskell functions. Specifically, it is designed to optimize the workflow in which a programmer makes a small change to a function and wants to measure its performance impact with as little friction as possible.

Screenshot

Example usage

Say I am interested in improving the performance of fib, which is a function defined in module MyMathUtilities.Fib in a local package called my-math-utilities.

Step 1: Write the function you'd like to benchmark

First, I'm going to copy the implementation of fib to a new top-level definition called fastfib, tweak its implementation, and export both from module MyMathUtilities.Fib.

If fib was private before, that's ok. We only need to expose it for as long as we are interested in benchmarking.

module MyMathUtilities.Fib (fib, fastfib, ...) where

Step 2: Write a standalone bench/Main.hs module with a main function

Next, I'm going to write a standalone Main.hs in a subdirectory called bench, which will be compiled to an executable that runs my benchmark.

module Main where

-- The module in my local package that I want to benchmark
import MyMathUtilities.Fib

-- This library
import ParkBench

main :: IO ()
main =
  benchmark
    [ function "fib" fib 20
    , function "fastfib" fastfib 20
    ]

Step 3: Define an executable component

Next, I'm going to define an executable component for my benchmark in my my-math-utilities.cabal file.

executable bench
  build-depends:
    base,
    -- The local package that I want to benchmark
    my-math-utilities,
    -- This library
    park-bench
  default-language: Haskell2010
  ghc-options: -O -rtsopts -with-rtsopts=-T
  hs-source-dirs: bench
  main-is: Main.hs

I need to compile the benchmark with -rtsopts -with-rtsopts=-T, otherwise my benchmark will not be able to get RTS statistics from GHC at runtime.

Alternatively, I could compile the benchmark with only -rtsopts, but then I'll have to provide +RTS -T to the executable later.

Step 4: Run the benchmark

If all goes well, I'll have an executable component to run.

cabal run my-math-utilities:exe:bench
stack run my-math-utilities:exe:bench

Or, if I only compiled with -rtsopts, but not -with-rtsopts=-T,

cabal run my-math-utilities:exe:bench -- +RTS -T
stack run my-math-utilities:exe:bench -- +RTS -T

Step 5: Clean up

After benchmarking, I can choose to keep the benchmark (and associated executable component) around, but I'll probably delete them instead. I've learned something, collected some sweet screenshots for my PR, and I'm ready to move on.

Caveat emptor

The statistical analysis performed by park-bench is simplistic, written by a novice, and may have bugs. Results should not necessarily be trusted; please use (or at least compare to) a different tool.