Re: gcc vs. Sun Workshop/Forte / was: Re: Can someone do us a favour? Need core trace back.

From: Roland Mainz (roland.mainz_at_nrubsig.org)
Date: 05/26/04 

Date: Wed, 26 May 2004 03:38:48 +0200

Peter Smithson wrote:
> > Well... did you you ever tought about working about the issue with the
> > "trial" version of Sun Workshop/Forte 8 ? Install it and then you have
> > around 30 days to get rid of the crasher... :)
>
> Our sys admin had just finished doing that when I realised gdb would
> work! And it's 60 days by the way so quite usefull. (Luckily I made it
> up to him today by finding out why our AIX compiler was failing).
>
> Since I posted, I've had a look around the net and saw those Mozilla
> results but I think they were quite old so I didn't know if they applied
> to the more recent versions of gcc.

As I said... I made my benchmarks long ago (and I never tried to figure
out the "why ?" - that's why I said that I never did in-depth research
on the issue) ... but the tests on my old SPARCstation-5/70MHz shows
that the gcc builds of Mozilla are still horrible slow when compared the
same source compiled with Sun Workshop. The Workshop builds of Mozilla
re useable on such a machine assuming you have enougth memory... but the
gcc builds are too slow for any useable browsing on such an old machine.

> I also found this article -
>
> http://www.osnews.com/story.php?news_id=6136
>
> Not much use for comparing to Forte but interesting that you can get
> double the speed by generating for more recent processors. Perhaps
> Forte has that too?

Yeah, but the gain isn't that big. I make the release builds in a way
that they still work on a SPARCstation-5. You can squish a lot more out
of the binaries with compiler option tweaking... but I avoid that for
the official builds stored at http://www.mozilla.org/releases/
The Mozilla.org default for Sun Workshop is "-xO4" but I reduce that to
"-xO2" for stability reasons.

> Hmm - just looking around again and found the authors site -
> http://vegan.net/tony/ which in turn takes us to another article he did
> comparing the compilers -
>
> http://www.osnews.com/story.php?news_id=5830
>
> Of interest is the last test which is on mySQL. One of our products is
> an SQL server so these tests are more relavant to us.

Mhhhh, the article doesn't say which optimizer options were used... I am
wondering what stuff like "-xipo" would squish out of MySQL... :) 

----
Bye,
Roland
-- 
  __ .  . __
 (o.\ \/ /.o) roland.mainz@nrubsig.org
  \__\/\/__/  MPEG specialist, C&&JAVA&&Sun&&Unix programmer
  /O /==\ O\  TEL +49 641 7950090
 (;O/ \/ \O;)

Relevant Pages

  • Re: Java versus Lisp. Comparison needed.
    ... It is like comparing some reasonable ... It is not wise for a Lisper to criticize performance of gcc. ... compared to gcc as GCL is compared to another not as good C compiler. ... There are also Clisp, ECL and ABCL ...
    (comp.lang.lisp)
  • Re: gcc and Static Variables
    ... > I am currently working on changing the compiler from Sun Workshop ... > static variables defined and declared in the file that contains ... how gcc identifies the language, ...
    (comp.lang.c)
  • Automake, sun workshop, and dependency
    ... a build system working with both GCC and SUN workshop. ... checking for gawk... ... checking for C compiler default output file name... ...
    (comp.unix.programmer)
  • Re: gcc vs. Sun Workshop/Forte / was: Re: Can someone do us a favour? Need core trace back.
    ... > same source compiled with Sun Workshop. ... gcc counterpart. ... If the app uses/links to GNU ... Sun ONE Studio 8 Compiler Collection ...
    (comp.unix.solaris)
  • Re: regex
    ... I tried it with the Sun Workshop 6.x CC and it worked fine for ... Just habit (gcc runs the C++ compiler any way). ...
    (comp.unix.programmer)
Quantcast