Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: An Opteron note

Author: Gerd Isenberg

Date: 03:34:39 11/27/03

Go up one level in this thread


On November 27, 2003 at 06:05:04, Daniel Clausen wrote:

>On November 27, 2003 at 05:54:15, Gerd Isenberg wrote:
>
>>On November 27, 2003 at 05:33:17, Gian-Carlo Pascutto wrote:
>>
>>>On November 27, 2003 at 05:29:43, Gerd Isenberg wrote:
>>>
>>>>Be aware of the "signed extension" penalty if using signed 32-bit variables as
>>>>array indicies, "long" is still 32-bit with msc but 64-bit in gcc:
>>>
>>>Ugh, that sucks! Of course all indexes are int here...
>>
>>Yes, the point is to avoid signed 32-bit int as index, due to the additional
>>sign extension instruction. A signed 64-bit register variable as index register
>>don't has this problem, but of course that should not imply to use explicitely
>>64-bit long (gcc) or long long (gcc) or __int64 (msc) as index variables, unless
>>your arrays are that huge ;-)
>>
>>Gerd
>>
>>>
>>>--
>>>GCP
>
>So using unsigned 32-bit would be a good idea, right?
>
>Sargon

Yes.

I don't have such a box so far to try it by myself and to look at the generated
code. But if i understand the amd-manual right that's the point.

Gerd



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.