This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] New Optimization: Partitioning hot & cold basic blocks
- From: Zdenek Dvorak <rakdver at atrey dot karlin dot mff dot cuni dot cz>
- To: ctice at apple dot com
- Cc: gcc-patches at gcc dot gnu dot org, pcarlini at unitus dot it
- Date: Fri, 10 Oct 2003 01:47:41 +0200
- Subject: Re: [PATCH] New Optimization: Partitioning hot & cold basic blocks
Hello,
> > On Wednesday, October 8, 2003, at 12:28 PM, Paolo Carlini wrote:
> > ... I'm too naive in believing that people would be also
> > interested in seeing some numbers, both about the runtime
> > improvements and about the compile time cost???
>
> In my unofficial SPEC runs on my machine, I get an improvement in the SPEC ratio
> of anywhere from 0.2% to 1.7% for gzip, mcf, parser, and twolf. vpr got worse by 1.2%.
> The compile time differences were negligible.
could you please provide more detailes? I would be especially
interested in following:
1) What architecture are these timings from?
2) What are the differences between results using profile feedback and
static estimates?
3) Change in sizes of binaries?
4) If possible, some oprofile data, especially impact on instruction
cache and branch prediction.
It might be also interesting to perform measurements on several
architectures.
Zdenek