delorie.com/archives/browse.cgi   search  
Mail Archives: djgpp/1996/07/11/19:45:15

Date: Fri, 12 Jul 1996 07:43:46 +0800 (GMT)
From: Orlando Andico <orly AT gibson DOT eee DOT upd DOT edu DOT ph>
To: "Eric J. Korpela" <korpela AT islay DOT ssl DOT berkeley DOT edu>
cc: djgpp AT delorie DOT com
Subject: Re: What is the quickiest way to clear memory?
In-Reply-To: <4s3ki2$btf@agate.berkeley.edu>
Message-ID: <Pine.SGI.3.93.960712074101.1646D-100000@gibson.eee.upd.edu.ph>
MIME-Version: 1.0

On 11 Jul 1996, Eric J. Korpela wrote:

> Unless you're not using a 386.  On pentium the fastest method seems
> to be an unrolled loop of fistq. (64 bits at a time)  On 486 machines
> that are clock doubled or tripled, an unrolled loop of movl instructions
> seems to be fastest.  rep stosd is only fastest on 386 and non-multiplied
> 486 machines.

Hmm.. well I don't program in assembly at all, I'm both lazy and
incompetent  :)

My question is, if anyone's been using gcc-2.7.2p-pl9 (the Pentium
optimized GCC) does it do this kind of optimization? I haven't had the
time to rebuild (DJGPP) GCC with the Pentium patches, but I *have* built
the compiler for Linux and I get a 15% speed increase for the dhrystone2
register benchmark.. if that's significant..

Cheers,

Orlando Andico                           http://gibson.eee.upd.edu.ph/~orly/
orly AT gibson DOT eee DOT upd DOT edu DOT ph                "Who knows what's going to happen,
IRC Lab/EE Dept/UP Diliman     lottery or car crash, or you'll join a cult."


- Raw text -


  webmaster     delorie software   privacy  
  Copyright © 2019   by DJ Delorie     Updated Jul 2019