tag:blogger.com,1999:blog-19626531.post8661498008290965627..comments2023-11-02T08:32:39.646+01:00Comments on Ola Bini: Programming Language Synchronicity: Mystery: An exposé on JRuby performanceOla Binihttp://www.blogger.com/profile/15793488672952593953noreply@blogger.comBlogger8125tag:blogger.com,1999:blog-19626531.post-9513866825234262972007-10-15T14:54:00.000+02:002007-10-15T14:54:00.000+02:00http://joyeur.com/2007/05/07/dtrace-for-ruby-is-av...http://joyeur.com/2007/05/07/dtrace-for-ruby-is-availableAnonymousnoreply@blogger.comtag:blogger.com,1999:blog-19626531.post-50288965412465713572007-10-13T21:59:00.000+02:002007-10-13T21:59:00.000+02:00Have you looked into using dtrace to profile what'...Have you looked into using dtrace to profile what's happening?joevandykhttps://www.blogger.com/profile/17835717350954324810noreply@blogger.comtag:blogger.com,1999:blog-19626531.post-25913574780071088982007-10-13T09:49:00.000+02:002007-10-13T09:49:00.000+02:00Route regognition does depend heavily on regular e...Route regognition does depend heavily on regular expressions, but as it only happens once a request it's unlikely to be a huge bottleneck.<BR/><BR/>You're running in production mode so the class-loading should be one-off.<BR/><BR/>Your benchmark rig seems quite a bit harder than the once I've been using(http://p.caboo.se/106832) not sure if it's worth trying that.<BR/><BR/>Grab me on IRC if you find anything interesting, always keen to apply benchmark driven patches :)Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-19626531.post-41268640365653317622007-10-12T23:20:00.000+02:002007-10-12T23:20:00.000+02:00It sounds a bit like "Look, all my unit tests succ...It sounds a bit like "Look, all my unit tests succeed so my application must work but what the ****, it crashes??" <BR/><BR/>Two possibilities imho:<BR/><BR/>1. Granularity of the benchmarks is not representative for a real life rails case.<BR/>2. Unforeseen behavior in rails that should have been benchmarked but did not made into the benchmarks<BR/><BR/>Its probably a combination. Rails simply stretches ruby to every edge it can find.<BR/><BR/>Finally, it is likely rails has been tested for performance on the MRI, not on JRuby.<BR/><BR/>Nevertheless, great figures. And given the JRuby progress that have been made lately, I bet it's nothing but a matter of time and you can show green benchmark results for Rails also.Okkehttps://www.blogger.com/profile/09887610259313782730noreply@blogger.comtag:blogger.com,1999:blog-19626531.post-74295440436167377432007-10-12T19:17:00.000+02:002007-10-12T19:17:00.000+02:00Any chance it's the rails class reloading? I woul...Any chance it's the rails class reloading? I would assume it is disabled, but I could see that having fairly different performance characteristics between MRI and JRuby.Adam Sandersonhttps://www.blogger.com/profile/10376742387880670801noreply@blogger.comtag:blogger.com,1999:blog-19626531.post-207763487293538212007-10-12T19:00:00.000+02:002007-10-12T19:00:00.000+02:00Hey, thanks for the benchmark numbers, wouldn't it...Hey, thanks for the benchmark numbers, wouldn't it be easier to just tag each line like so:<BR/><BR/>MRI: 0000.000 0000.0000 0000.0000<BR/>JRB: 1111.111 3333.3333 9999.9999Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-19626531.post-56357769757722069132007-10-12T15:45:00.000+02:002007-10-12T15:45:00.000+02:00Ugh... so much for us red/green color blind folks....Ugh... so much for us red/green color blind folks.<BR/><BR/>That thin colored text on the white background is hard to pick apart. :(Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-19626531.post-61621357734031246122007-10-12T13:14:00.000+02:002007-10-12T13:14:00.000+02:00If I remember correctly from last time I looked at...If I remember correctly from last time I looked at it, Rails's routing code leans very heavily on regular expressions for route recognition. If I had to put the blame somewhere for poor rails performance in JRuby, that would be the first place I'd look.Anonymousnoreply@blogger.com