kindaro / seq-vs-pseq

Profiling a contrived seq / pseq application.

seq versus pseq

Profiling a contrived seq / pseq application.

In this repository:

• Three *.hs files. It's a contrived sum program with a custom fold.

  • One is without strictness: SumNaive.
  • The other is with two-side-strict seq: SumSeq.
  • The third one is with left-side-strict pseq from parallel: SumPseq.

• A shell script that builds these programs and collects their running parameters: running time and RTS output, as well as some profiling data.

  The script will build the programs with all the combinations of -prof -O -O2, so that the effect of optimizations and the influence of profiling were exposed.

• Some stack.yaml configurations for various GHC versions.

  The suggestion here is to create a copy of the repository for each stack.*.yaml and separately explore the execution, so that differences between the way different GHC versions compile, optimize and profile were exposed.

  TODO: Why not run everything from one place, simply giving stack different yaml files via --stack-yaml?

TODO: Add Criterion to the mix for a more accurate time report.