Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Speeds intel c++ versus gcc 3.1

Author: Andrew Williams

Date: 10:17:50 05/18/02

Go up one level in this thread


On May 18, 2002 at 08:21:29, Vincent Diepeveen wrote:

>Hello,
>
>Zeke Smigel did for me a test and it was rather surprising.
>
>At his K7 1.4Ghz MP:
>
>  intel c++ 6.0 (reinforced with profile branch information):
>
>  64919.8 nodes a second
>
>
>  gcc 3.1 (not reinforced with profile branch information as
>           i didn't get this to work yet):
>
>  68135.2 nodes a second
>
>
>At intel P3-800Mhz under linux gcc 3.1 was only 5.1% slower than
>the intel c++ 6.0 version. This is very good result for gcc 3.1
>(consider it is without profiling branch optimization yet)
>
>So in short GCC 3.1 kicks butt!!

My preliminary tests results are quite impressive. It's now easy to use
execution profiling to improve results. Without execution profiling, I get a
speedup of between 5 and 6 percent. With execution profiling I'm getting over
10% speedup. I'm going to more tests, but so far I'm very happy with this.

Andrew

PS Thanks a lot to Vincent Diepeveen and Steffen Jakob for help with getting me
to this point.



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.