Related articles |
---|
Pentium Compiler? colem861@cs.uidaho.edu (1994-07-28) |
Re: Pentium Compiler? johnm@cory.EECS.Berkeley.EDU (1994-07-30) |
Re: Pentium Compiler? tgm@netcom.com (1994-08-01) |
Re: Pentium Compiler? hays@ichips.intel.com (1994-08-03) |
Newsgroups: | comp.compilers |
From: | hays@ichips.intel.com (Kirk Hays) |
Keywords: | 586, optimize, GCC |
Organization: | Intel Corp., Hillsboro, Oregon |
References: | 94-07-095 94-08-012 |
Date: | Wed, 3 Aug 1994 16:15:47 GMT |
Status: | RO |
James E Coleman (colem861@cs.uidaho.edu) wrote:
>: [I want a compiler that will optimize for the Pentium Processor.
Thomas G. McWilliams <tgm@netcom.com> wrote:
>Try ftp.intel.com:/pub/x86/gcc/compiler/gcc-i2.4.0p.tar.Z
>
>This is Intel's own version of gcc for the P5. I don't think any of the
>changes have been merged into the FSF/cygnus tree. This might be your only
>choice with respect to gcc. It is reported to generate code about 20%
>faster on P5 processors than standard gcc. According to the docs, this is
>Intel's only experiment with gcc for the P5 and they have ceased further
>work on this compiler.
I've used this compiler, and in my experience, you get about 20% on
integer code, and up to 10x on FP code. YMMV. I've found no penalty on
the 486 when using pentium optimized code, although the compiler notes do
warn of slow-downs.
It isn't the most stable compiler, and attempting to bring the diffs
forward into 2.5.8 proved to be futile - the resulting compiler couldn't
compile itself.
I find it's use worthwhile on FP kernels, and use 2.5.8 or 2.6.0 on the
non-FP portions of my applications.
I'd really like to see the modifications rolled into the current source
base for GCC, if at all possible. I'd do it myself, but I'm no expert on
GCC internals.
--
Kirk Hays
--
Return to the
comp.compilers page.
Search the
comp.compilers archives again.