Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: A Few Comments

Author: Matt Taylor

Date: 13:06:10 01/20/03

Go up one level in this thread


On January 20, 2003 at 15:55:22, Gerd Isenberg wrote:

>On January 20, 2003 at 15:10:53, Matt Taylor wrote:
>
><snip>
>>I've never found a way around this limitation. My best answer is to say that the
>>loads/stores can be optimized out by a post-optimizer.
>>
>>I have tools which would enable me to easily write such a program, but the tools
>>are only half-working. I need the ability to reverse engineer an executable file
>>-- right now I miss some sections of code completely, and finding them does not
>>look like an easy task. Once that is done, it would be trivial to inline
>>functions and do some simple load/store optimization.
>>
>>-Matt
>
>Thanks Matt,
>that's a real pity with inlined asm - some more bytes (8 if already in register)
>for each parameter. But it seems the temporary load/stores on the stack are not
>so expensive, at least not so expensive that even small inlined asm-routines pay
>off.
>
>Cheers,
>Gerd

Yeah, some 4-6 clocks penalty which can be hidden...it pays off since the
call/ret pair are more expensive. It is annoying and unnecessary, though. I can
appreciate gcc's inline assembly because they avoid this problem.

It wouldn't be a stellar improvement, but elimination of that extra memory
access would be excellent.

-Matt



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.