help-smalltalk
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Help-smalltalk] Re: GNU ST Speed


From: Tomas Vanak
Subject: [Help-smalltalk] Re: GNU ST Speed
Date: Mon, 13 Jan 2003 12:58:50 +0100

Hi Paolo,
I wanted to try the same code in both smalltalks, that is why I used the
given code.
>From the output of ./configure I can see I'm not using the GMP library:

checking how to link with GMP... not found

even when I'm using ./configure --with-gmp=/usr/lib
I have a libgmp.so.3.1.1 library there. Do I need another version? Am I
putting the -with-gmp parameter wrong?

Thanks
Best regards
Tomas


----- Original Message -----
From: "Paolo Bonzini" <address@hidden>
To: "Tomas Vanak" <address@hidden>
Cc: <address@hidden>
Sent: Tuesday, January 07, 2003 10:34 AM
Subject: Re: GNU ST Speed


> > I tried the version 2.0h and I saw a lot of improvements!
> > Amazing work!
>
> Thanks!
>
> > I was trying to compare the speed of St/X and GNU St.
> > I tried just a simple example of a factorial:
> > Integer>>factorial
> >  |p i|
> >  p := 2.
> >  i := 3.
> >  [i <= self] whileTrue: [
> >   p := p * i.
> >   i := i + 1.
> >  ].
> >  ^p
>
> Why don't you use the bundled factorial method?  It is optimized to the
> extreme... also check that you are using GMP.  ST/X's LargeIntegers are
> coded in C, without GMP GNU Smalltalk's are coded in Smalltalk.
>
> > Then I did-it: 1000 timesRepeat: [1000 factorial], on GNU St it took
many
> > minutes to complete and it was scavenging all the time.
>
> A scavenging takes a few milliseconds.  Even using -g might improve the
> performance by avoiding printing the messages.  It might also be
worthwhile
> to increase the size of new-space (see methods in ObjectMemory) -- in
> general, not only for this example -- for example having 1 MB of new-space
> and 200 Kb for each survivor space (total 1,5 MB; currently I have 300 Kb
> and 40 Kb).  More info on the new GC/memory manager is found in the info
> documentation
>
> Paolo
>





reply via email to

[Prev in Thread] Current Thread [Next in Thread]