From: Paul Shirley <Paul AT foobar DOT co DOT uk DOT chocolat>
Newsgroups: comp.os.msdos.djgpp
Subject: Re: Intel MMX with DJGPP?
Date: Sun, 22 Dec 1996 21:47:40 +0000
Organization: wot? me?
Lines: 18
Distribution: world
Message-ID: <H7IkgFA8xavyEwLw@chocolat.foobar.co.uk>
References: <32BA24E1 DOT 72E1 AT stud DOT warande DOT ruu DOT nl>
 <199612202105 DOT NAA11697 AT netcom5 DOT netcom DOT com>
Reply-To: Paul Shirley <junk AT defeating DOT email DOT address>
NNTP-Posting-Host: chocolat.foobar.co.uk
Mime-Version: 1.0
To: djgpp AT delorie DOT com
DJ-Gateway: from newsgroup comp.os.msdos.djgpp

In article <199612202105 DOT NAA11697 AT netcom5 DOT netcom DOT com>, Marc Singer
<elf AT netcom DOT com> writes
>Though I have not seen it, I hope that *someone* has pointed out that
>Intel is suggesting that software developers avoid using the MMX
>instructions.  In the present Intel implementation, Windows (and any
>other operating system not explicitly coded to handle the MMX mode)
>cannot prevent the MMX/floating point corruption.

Where did they say that? I've seen it claimed that MMX uses the FPU
partly because the code *already* in use in Windows will automagically
save the FPU/MMX state correctly. (ie an fpusave op will work whichever
mode the FPU is in). (Of course there are some bugs in Windoze that
corrupt the fpu anyway...)

MMX would have been a lot more useful if it was independent of the fpu.

---
Paul Shirley: shuffle chocolat before foobar for my real email address