contrib/gdl-benchmark/README.markdown @ 16b422487296
Add profiling to the GDL benchmark
| author | Steve Losh <steve@stevelosh.com> |
|---|---|
| date | Wed, 24 Aug 2016 22:13:48 +0000 |
| parents | 1f0a36161f17 |
| children | cbfb813a3f82 |
Temperance GDL Reasoning Benchmark ================================== An thin wrapper to plug Temperance into [Stephan and Yngvi's GDL reasoner benchmarking library thing][paper]. [paper]: http://cgi.cse.unsw.edu.au/~mit/GGP/GIGA-13-Proceedings.pdf#page=55 Setup ----- Install [Roswell][]. Pick a Lisp implementation and install it with Roswell. Temperance is primarily focused on running fast with [SBCL][], but should run on at least CCL and ECL too. ros install sbcl Clone down the dependencies into Roswell's `local-projects` directory. You can clone them elsewhere and symlink them if you prefer: cd ~/.roswell/local-projects/ # Mercurial hg clone http://bitbucket.org/sjl/temperance hg clone http://bitbucket.org/sjl/cl-losh # Git git clone http://github.com/sjl/temperance git clone http://github.com/sjl/cl-losh Build the benchmarking binary with your chosen Lisp: cd ~/.roswell/local-projects/temperance/contrib/gdl-benchmark/ ros use sbcl ros build run-temperance.ros Run the binary just like you would any of the others in the suite: ~/.roswell/local-projects/temperance/contrib/gdl-benchmark/run-temperance 'dfs mc' 10 .../foo.gdl .../foo.trace [Roswell]: https://github.com/roswell/roswell [SBCL]: http://www.sbcl.org/ Performance ----------- The benchmark script ensures that Temperance will be compiled with sane optimization settings: `(debug 1) (safety 1) (speed 3)`. If you want to throw caution to the wind and see how fast it can get, you can set the `PLEASE_SEGFAULT` environment variable to `YES` **when building**: cd ~/.roswell/local-projects/temperance/contrib/gdl-benchmark/ ros use sbcl PLEASE_SEGFAULT=YES ros build run-temperance.ros This must be done when *building*. The variable has no effect when running the binary. In practice this results in a speed increase of around 20%.