Parrot Benchmarking

Lucian Branescu lucian.branescu at
Thu Nov 25 10:13:42 UTC 2010

        On Tuesday, 23 November 2010 at 14:43, Jon Gentle wrote:
            There was some discussions last week about the speed of parrot and being able to benchmark it.  As it turns out, I had time for a quick and dirty project, so I built .  I whipped up a quick sqlite database and catalyst app, and thanks to dukeleto, I'm using Tool::Bench to actually generate the benchmarks.  I've posted the code onto github at and I went ahead and attempted to populate it with benchmarks from each release tag since 2.0.0.
Obviously, it's still pretty young and it has a few flaws.  First of all, there is no way to submit benchmarks to it, and in fact, at this point, it's ran manually.  Second, the data points need some context information so it's obvious exactly what they represent. 
Perhaps most importantly, it only has 1 real benchmark.  The rakudo start time benchmark is broken since I don't have a good way to checkout a previous rakudo that matches the parrot checkout for a release.  It's not a problem when I do daily benchmarking, and I could do it manually for the previous releases, but I'd rather not.
If anyone has any questions or suggestions, I'm all ears.  Even better, pull requests are welcomed.Perhaps it would be helpful to reuse existing code. The PyPy developers have, which is an instance of
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <>

More information about the parrot-dev mailing list