Actually, I think I have the major performance problem figured out.... String concatenation. The result << line is the major time sucker in this. On my laptop when I change RubyString to use a StringBuffer instead of String a 10,000 line file takes 0.6 seconds to run while it takes 242 seconds as a String. The full gemspec runs in about 6 seconds versus 0.21 in ruby so we are still a ways off, but it seems closer to general interpreter performance (e.g. eval'ing nodes). Regexp creation probably is the only other measurable difference.
I am looking into doing this conversion now (I did a fairly naive modification that sucks for all the non-stringbuffer-friendly functions). -Tom On Sat, 08 Apr 2006, Ola Bini defenestrated me: > Ok, next point of interest, regexp substitions with String.sub > With my former optimizations for Regexps I got this: > D:\Project\jruby>ruby testRegSubPerformance.rb > Time1 0.581 > Time2 0.58 > > D:\Project\jruby>bin\jruby testRegSubPerformance.rb > Time1 8.623 > Time2 5.428 > > where the test is more or less similar to other one. > With the attached patch applied I get this: > D:\Project\jruby>bin\jruby testRegSubPerformance.rb > Time1 6.48 > Time2 5.1770000000000005 > > which is quite good. And the only difference was to replace a RubyString > with a StringBuffer in two lines which used append. I guess these small > kinds of optimizations could get us some good performance at last. > > Regards > Ola Bini > > > ----- Original Message ----- > From: Ola Bini <[EMAIL PROTECTED]> > Date: Saturday, April 8, 2006 11:27 am > Subject: Re: [Jruby-devel] Regexp performance. > To: jruby-devel@lists.sourceforge.net > > > Ok, the second approach gave a little bit more: > > > > D:\Project\jruby>bin\jruby testRegexpPerformance.rb > > Time1 5.168 for sum 150000 > > Time2 3.9450000000000003 for sum 150000 > > > > But it's still not great. Anyway, I've attached the patch if this is > > something we should continue work on. > > > > Regards > > Ola Bini > > > > ----- Original Message ----- > > From: Ola Bini <[EMAIL PROTECTED]> > > Date: Saturday, April 8, 2006 11:23 am > > Subject: Re: [Jruby-devel] Regexp performance. > > To: jruby-devel@lists.sourceforge.net > > > > > Once again, a very small test where I replaced the > > Pattern.compile in > > > RegexpTranslator into a getPattern which saves all earlier > > patterns > > > in a > > > hashmap. This actually worsens the performance slightly, so that's > > > probably not the right way to go. I will try to apply this > > approach to > > > the whole RegexpTranslator.translate method instead and see what > > > happens. > > > Regards > > > Ola Bini > > > > > > ----- Original Message ----- > > > From: Ola Bini <[EMAIL PROTECTED]> > > > Date: Saturday, April 8, 2006 10:53 am > > > Subject: [Jruby-devel] Regexp performance. > > > To: jruby-devel@lists.sourceforge.net > > > > > > > Hi there, > > > > > > > > I made a highly unscientific test of performance of regexps in > > Ruby> > compared to JRuby. > > > > The attached script is what I used to test this. The output is as > > > > follows for 100 000 repetitions of the same regexp applying: > > > > D:\Project\jruby>ruby testRegexpPerformance.rb > > > > Time1 0.201 for sum 150000 > > > > Time2 0.22 for sum 150000 > > > > > > > > D:\Project\jruby>bin\jruby testRegexpPerformance.rb > > > > Time1 5.698 for sum 150000 > > > > Time2 3.986 for sum 150000 > > > > > > > > This seems fairly good, actually, we're only an order of magnitude > > > > slower, which is not so bad if you think about all the stuff > > that's> > going on. The difference between Time1 and Time2 is what > > happens > > > if I > > > > save the regexp in a variable before the loop, or not. As you > > can > > > see,> we could save pretty much performance here. > > > > > > > > Regards > > > > Ola Bini > > > > > > > > > > > > > ------------------------------------------------------- > > > This SF.Net email is sponsored by xPML, a groundbreaking > > scripting > > > languagethat extends applications into web and mobile media. > > Attend > > > the live webcast > > > and join the prime developer group breaking into this new coding > > > territory!http://sel.as- > > > > > > us.falkag.net/sel?cmd=lnk&kid=110944&bid=241720&dat=121642_______________________________________________ > > > Jruby-devel mailing list > > > Jruby-devel@lists.sourceforge.net > > > https://lists.sourceforge.net/lists/listinfo/jruby-devel > > > > > -- + http://www.tc.umn.edu/~enebo +---- mailto:[EMAIL PROTECTED] ----+ | Thomas E Enebo, Protagonist | "Luck favors the prepared | | | mind." -Louis Pasteur | ------------------------------------------------------- This SF.Net email is sponsored by xPML, a groundbreaking scripting language that extends applications into web and mobile media. Attend the live webcast and join the prime developer group breaking into this new coding territory! http://sel.as-us.falkag.net/sel?cmd=lnk&kid=110944&bid=241720&dat=121642 _______________________________________________ Jruby-devel mailing list Jruby-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/jruby-devel
