On Sun, Apr 13, 2008 at 07:21:06PM +0100, Nuno 'smash' Carvalho wrote: > On Sun, Apr 13, 2008 at 7:06 PM, Patrick R. Michaud <[EMAIL PROTECTED]> wrote: > > > > On Sun, Apr 13, 2008 at 06:57:26PM +0100, Nuno 'smash' Carvalho wrote: > > > Greetings all, > > > > > > We did another Parrot benchmarking, this time using a common > > > programming technique: recursion. We created a function to calculate > > > the number of nodes in a full binary tree given the tree's height. I > > > guess this time the results where not so satisfactory, for Parrot. You > > > can see the details here: http://use.perl.org/~smash/journal/36140 > > > > I left a comment on the use.perl post as well, but is there an easy way > > we could look at the sources being used for the benchmarks -- i.e., > > if we wanted to duplicate the results on our systems? > > Sure. You can download the source code used for every implementation here: > > http://nrc.homelinux.org/parrot/bintree/
Thanks, this is very helpful, as is the overall benchmarking effort you're putting together. The take-away result I get from the bintree benchmark, both from the graphic and from tests on my system, is that for this benchmark Parrot is about twice as slow as Perl 5. Personally, since this test was intended to benchmark recursion I suspect that the difference is in Parrot's subroutine call/return, but we probably need to do some additional profiling to verify that this is the case. Pm
