Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Crafty bench & caches experiment on a Powerbook

Author: Rick Bischoff

Date: 14:20:03 08/20/04

Go up one level in this thread


Hello,

How did you get it to compile cleanly with xlc?  My compiliation always seems to
bug out!  (I think I only tried it with -O5 though)

On August 20, 2004 at 17:04:49, Frank E. Oldham wrote:

>On my Powerbook I can selectively enable/disable caches.
>Here's the bench results for crafty19.15.
>
>1 GHz G4 (7455) with
>L1 cache = 32KB + 32Kb
>L2 cache (fast) = 256KB
>L3 cache (slow) = 1MB
>
>crafty19.15 compiled with IBM xlc compiler (fastest G4 32-bit binary)
>
>L2      L3      bench nps
>-------------------------
> +       +        472948
> +       -        449448
> -       +        374928
> -       -        190424
>
>From these numbers it looks like an increase from 256K to 512K of L2 cache
>would probably speed up crafty by 10% or so.
>
>Note -- I recall that in making the NUMA changes for his Opteron machine,
>Bob improved crafty's overall cache performance quite a bit.  I think this was
>during the progression from around 19.05 up to 19.10.  Bench for the Mac G4 & G5
>sped up 5-10% alone just from these "data layout" changes.
>
>Frank



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.