Hey Laurent,it took me a couple of days to ask my friend if I could forward his code which he agreed upon yesterday so here is a tar.bz2 containing all the code. We ran the »test/minheap_performance_test.rb«. Note that it contains lines (35, 43, 51) which use the large number of method arguments. Obviously one would not have to do it like that but he wanted to play around a bit. So to run the tests you would have to comment them out. But before that you can run them on ruby 1.9.1 to see how they are doing ;)
As said before this test ran faster on ruby 1.9.1 from macports on my mac book pro 4,1
Kind regards, John
ruffman.tar.bz2
Description: BZip2 compressed data
On 05.09.2009, at 21:34, Laurent Sansonetti wrote:
Hi, On Sep 5, 2009, at 3:11 AM, John-Paul Bader wrote:Hey guys,a friend of mine implemented a heap in ruby for educational purposes. He did so with several implementations. He had a few benchmarks and I suggested to run them on macruby to see how well it would hold up.Unfortunately we didn't get really far because apparently there is a problem with globbed/splatted method arguments in macruby which raises an exception if more than 199 arguments are passed.Can be tested easily with: def foo(*bar); puts bar.length;end 1.upto(1000) {|x| foo(*([0]*x))}In vm.cpp:2977 #define MAX_DISPATCH_ARGS 200 is defined and a big fat todo is right below it addressing this issue. Now I know its not a super urgent problem because its not what you do in ruby every day. Still its obviously something that has to change.In line 3003 there is this assert: assert(real_argc + count < MAX_DISPATCH_ARGS); so it must be important i guess.So i'm just wondering why it is limited to 200 in the first place and if there is something I can to even though I'm not really a capable (objective) c programmer ?This limitation is purely temporary and is because of a limitation in the dispatcher. This will be addressed for the release.After commenting out the code with the globbed arguments the benchmarks on the remaining tests ran a bit slower than on ruby 1.9.1.Can you share these benchmarks? I would be very interested to see the results and also if we can be better.Laurent _______________________________________________ MacRuby-devel mailing list MacRuby-devel@lists.macosforge.org http://lists.macosforge.org/mailman/listinfo.cgi/macruby-devel
_______________________________________________ MacRuby-devel mailing list MacRuby-devel@lists.macosforge.org http://lists.macosforge.org/mailman/listinfo.cgi/macruby-devel
