Date: Tue, 2 Dec 1997 10:30:14 +0800 (GMT)
From: Orlando Andico <orly AT dilnet DOT upd DOT edu DOT ph>
To: Paul Derbyshire <ao950 AT FreeNet DOT Carleton DOT CA>
cc: djgpp AT delorie DOT com
Subject: Re: 32bit memcpy function? _NEW_ Tried FPU memcpy (problem with CWSDPMI)
In-Reply-To: <65r887$fr9@freenet-news.carleton.ca>
Message-ID: <Pine.SGI.3.96.971202102850.3658A-100000@gibson.eee.upd.edu.ph>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Precedence: bulk

On 30 Nov 1997, Paul Derbyshire wrote:

> 
> Michal Mertl (xmerm05 AT manes DOT vse DOT cz) writes:
> > On Sun, 23 Nov 1997, Speed wrote:
> > 
> >> So it is equivalent to a 'rep movsl'? (assuming, of course, the source,
> >> dest, and count are already set)
> >> 
> > Yes, it is. Why you don't just look at the output. There are at least 3 good
> > debuggers for DJGPP.
> > 
> > Other thing is that I tried to write memcpy using 64bit FPU registers as
> > someone here suggested. It's about _20% faster_!!
> > 
> > __here is the source code of memcpyfpu.c __
> 
> NICE... anyone want to consider replacing the default memcpy with this
> feller? (Of course, to avoid it running 50000% SLOWER on FPU-less machines
> you might want to make the regular one be used instead if the FPU isn't
> found. :-))

NOT NICE. Only works on Pentium/PPro/Pentium MMX/Pentium-II. Does NOT work
on Cyrix 6x86, AMD K5 and K6, even if these machines do have an FPU. You
would have to implement CPUID to detect the CPU before using the FPU-based
memcpy() in order to be safe, and that's not exactly trivial.



-------------------------------------------------------------------
Orlando Alcantara Andico              Email: orly AT dilnet DOT upd DOT edu DOT ph
ICBM:  14 30 00 N  120 59 00 E               POTS:  (+632) 932-2385
"If I shed a tear I won't cage it, I won't fear love, and if I feel
 a rage I won't deny it, I won't fear love."      - Sarah Mclachlan