Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Hammer info. And som SMP musings.

Author: Eugene Nalimov

Date: 15:28:51 03/27/02

Go up one level in this thread


Some of the Spec2k int benchmarks are very sensitive to the memory latency. Take
a look at the submitted Spec2k results for Itanium. There are results for HP-UX
(where pointers are 32-bit) and Windows (where pointers are 64-bit). Look at the
MCF. MCF is *very* sensitive for memory latency. For 32-bit pointers it's much
faster than for 64-bit ones; of course different compilers were used, but
believe me, main reason is much better cache utilization when pointers are
smaller.

And now look at the MCF result for POWER4. It's unproportinally high. I believe
that's because entire dataset was in the L3 cache.

Memory latency for the Spec2k FP is much less important than for the Spec2k int
-- and that's true for almost any FP application.

Eugene

On March 27, 2002 at 14:21:04, Tom Kerrigan wrote:

>On March 26, 2002 at 23:45:17, Eugene Nalimov wrote:
>
>>I agree that POWER4 is good processor, I just cannot believe it's so much better
>>than others.
>
>It's not THAT much better than the others. It's slower than a 21264C @ 1.3GHz,
>assuming linear speedup. I find that very believable. I don't think the cache
>really matters. Maybe on SPECfp.
>
>-Tom



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.