Re: [cc65] Optimization for size

From: Groepaz <groepaz1gmx.net>
Date: 2006-04-05 13:16:37
On Wednesday 05 April 2006 13:10, Ullrich von Bassewitz wrote:
> On Wed, Apr 05, 2006 at 09:46:25AM +0200, Groepaz wrote:
> > i'm just wondering because i did a simelar thing before (using
> > handcrafted rules though) and the result wasnt quite THAT good.
>
> There are two differences: The rule set is automatically generated,
> therefore it is really large (about 13000 rules/patterns). And, it is
> generated from one set of sources, so the rule set will be optimal for this
> set of sources. Both differences mean that the rule set by itself is not
> really generic. But the general approach is. You can always run a program
> over your .s files that detects common patterns, generate a set of
> replacement rules, and then replace common code sequences by a set of
> subroutines.
>
> However, the approach is beyond what would normally be done by a compiler,
> because to generate an optimal set of patterns, you will have to look at
> all assembly files for a complete project, not just at one module (as the
> compiler does).

yes ofcourse :)

as for the compiler, what about adding a little bit of support for undocs to 
the codegenerator (or rather, the internal optimizer) ? might also safe some 
memory (and gain speed) in some cases, and it can be done very generic :)

-- 

http://www.hitmen-console.org
http://www.gc-linux.org/docs/yagcd.html
http://www.pokefinder.org
http://ftp.pokefinder.org

Wenn A für Erfolg steht, gilt die Formel A = X + Y + Z. X ist Arbeit, Y ist 
Muse, und Z heißt Mundhalten. 
<Albert Einstein>

----------------------------------------------------------------------
To unsubscribe from the list send mail to majordomo@musoftware.de with
the string "unsubscribe cc65" in the body(!) of the mail.
Received on Wed Apr 5 13:16:44 2006

This archive was generated by hypermail 2.1.8 : 2006-04-05 13:16:47 CEST