[racket-dev] Wow; racket master at least 2x faster than 5.3.1 on my rb tree benchmark?

From: Matthew Flatt (mflatt at cs.utah.edu)
Date: Sun Nov 18 19:36:59 EST 2012

Yes, a program that is all structure creation, access, and mutation
should run around twice as fast compared to v5.3.1.

At Sun, 18 Nov 2012 17:07:12 -0600, Robby Findler wrote:
> I don't know if this is the reason, but I do know that Matthew made
> the jit able to see thru some structure operations. Perhaps that
> enables some other optimizations now that weren't in 5.3.1.
> 
> Robby
> 
> On Sun, Nov 18, 2012 at 4:50 PM, Danny Yoo <dyoo at hashcollision.org> wrote:
> > I'm doing some micro-optimizations on my rb-tree implementation.  One thing
> > I'm testing is inserting the entire contents of /usr/share/dict/words.  It's
> > heavily dominated by structure-mutation code.
> >
> > Under 5.3.1, I see the following times:
> >
> >     Timing construction of /usr/share/dict/words:
> >     inserting 235886 words at the end...
> >     cpu time: 446 real time: 446 gc time: 0
> >     dropping all those words...
> >     cpu time: 355 real time: 374 gc time: 0
> >     inserting 235886 words at the front...
> >     cpu time: 437 real time: 436 gc time: 0
> >
> >
> > Out of curiosity, I wanted to see how fast this ran under Racket under git
> > (8d30f173).  Under that version, I'm seeing:
> >
> >     Timing construction of /usr/share/dict/words:
> >     inserting 235886 words at the end...
> >     cpu time: 195 real time: 195 gc time: 0
> >     dropping all those words...
> >     cpu time: 203 real time: 217 gc time: 0
> >     inserting 235886 words at the front...
> >     cpu time: 200 real time: 199 gc time: 0
> >
> >
> > So I don't know what exactly happened between then and now, but whatever it
> > is, keep doing it!  :)
> >
> > _________________________
> >   Racket Developers list:
> >   http://lists.racket-lang.org/dev
> >
> _________________________
>   Racket Developers list:
>   http://lists.racket-lang.org/dev

Posted on the dev mailing list.