Xref: news2.mv.net comp.os.msdos.djgpp:4408
From: brennan AT mack DOT rt66 DOT com (Brennan "Mr. Wacko" Underwood)
Newsgroups: comp.os.msdos.djgpp
Subject: Re: Making the nearptr hacks use longs.
Date: 30 May 1996 13:36:56 -0600
Organization: None, eh?
Lines: 27
Message-ID: <4oktco$9u1@mack.rt66.com>
References: <4nqe0f$som AT geraldo DOT cc DOT utexas DOT edu> <4nqp27$8uu AT waldorf DOT csc DOT calpoly DOT edu>
NNTP-Posting-Host: mack.rt66.com
To: djgpp AT delorie DOT com
DJ-Gateway: from newsgroup comp.os.msdos.djgpp

In article <4nqp27$8uu AT waldorf DOT csc DOT calpoly DOT edu>,
Jan Louwerens <jlouwere AT galaxy DOT csc DOT calpoly DOT edu> wrote:
>of the loop, as well as the (4*i) and the add. Why not just make an inline
>assembler function to do it. That seems fastest to me. Something like:
>
>	mov	esi, (pointer to Virtual Screen)
>	mov	edi, (pointer ~csc/sample/ada/to VGA Memory)
>	mov	ecx, 16000	(320*200 / 4 (for ~csc/sample/ada/dwords))
>	rep
>	movsl
>
>(This code may not be exactly correct, but it gets the idea across)

You'll get pretty much exactly this if you do:

memcpy((pointer to Virtual Screen), (pointer to VGA memory), 64000);

with optimization turned on. Trust me.

As for speed, you can sometimes beat a rep;movsl on P5 and up, but only
on some memory subsystems (Overlapping pushes/pops, check comp.lang.asm.x86).
For general purpose code, this is the fastest you're likely to get.


--Brennan
-- 
brennan AT rt66 DOT com  |  "Barney vs. Mentos II: Unholy Alliance." Coming Soon.