I don't know if it will help, but I used another flag. I have not my computer at hand but I think it was in the lines of -XX agressiveoptimizations.
Moreover, the performance goes from 200 000 to 800 000 depending on the size of the chunks. I tried to launch for 1 or 10 millions iterations to find the fastest on my computer. That is quite strange the fact that the size xhange so much the result. Really a fun exercize in writing benchmarking clojure, though. Best, Nicolas. On Jul 5, 2009 6:53 PM, "igorrumiha" <igorrum...@gmail.com> wrote: On Jul 5, 3:05 pm, Nicolas Oury <nicolas.o...@gmail.com> wrote: > Actually, that would be so good, ... OK, I did my tests, I found and fixed two small bugs and as far as I can tell the VM is now working correctly. I couldn't, however, reproduce the timings Nicolas had, on my machine (a 2 Ghz MacBook with java 1.6 build 1.6.0_13-b03-211) I got timings between 570 and 637 seconds for 100 000 000 executions. So, this translates, on average, to 163585 executions/second. That translates to 45.9 clock cycles per guest VM instruction. I used these tuning options for the Java VM: -XX:InitialCodeCacheSize=768m -XX:ReservedCodeCacheSize=1024m -Xms768M -Xmx768M -server I consider this an excellent result, about 2.5 times faster than my simple Java implementation of the VM. I can't explain the 5.2x difference between my and Nicolas' timings, though :) What I can say is that my quest for Orbital Virtual Machine speed is satisfied. With a simulation time limit of 3 million seconds (set by the contest rules) this VM now reaches that limit in ~19 seconds of wallclock time. That's even good enough for interactive exploration of problem solutions (I think it's very likely that it will usually take more than 19 seconds to write a new version of a solution and then run it in the REPL :)) I think it's safe to say that once again it's proved that Clojure easily matches the Java level of performance. What I've learned from this is: 1. I need to read the documentation more carefully :), 2. I need to learn how to write macros, 3. I need to get myself acquainted with the underlying JVM platform, 4. When in trouble, the clojure group is an excellent place to get help. 5. Write lots and lots of Clojure code, that is the only way to get enough experience for the next ICFP contest :) Thanks a bunch everyone, especially Nicolas for being so persistent :) p.s. I'll update my git repo in a day or two, I need to clean up the code a bit and port the game logic code to use the new VM. Igor. --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subs... --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "Clojure" group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en -~----------~----~----~----~------~----~------~--~---
