Date:	Wed, 5 Mar 1997 08:51:06 +0800 (GMT)
From:	Orlando Andico <orly AT gibson DOT eee DOT upd DOT edu DOT ph>
To:	Mark T Logan <fwec AT juno DOT com>
cc:	ao950 AT FreeNet DOT Carleton DOT CA, djgpp AT delorie DOT com
Subject:	Re: Loop unrolling: Don't bother -off topic
In-Reply-To:	<19970303.162642.7191.1.fwec@juno.com>
Message-ID:	<Pine.SGI.3.93.970305084830.8941C-100000@gibson.eee.upd.edu.ph>
MIME-Version:	1.0

On Mon, 3 Mar 1997, Mark T Logan wrote:

> 
> MMX is simply a pentium with several new string instructions for improved
> multimedia capability.  It also runs 10-20% faster than a normal pentium.
> 
> -Fwec
> 

Not exactly. It runs 10-20% faster because the internal cache is 32K
instead of 16K (of course, if you have a PPro, the Pro's 256K internal
cache blows it away...)

Also, there are some new instructions which are allegedly faster for
things like blitting graphics onscreen and such. Allegedly. Because these
instructions use the FPU registers, so... you can't crunch heavy math at
the same time.

I personally think MMX is just Intel trying to perpetuate the "dumbing
down peripherals" thing -- like those Rockwell modems that implement MNP-5
compression using the main CPU to save on a lousy $20 DSP chip. And
they're not even original at it. I think the first CPU to have dedicated
DSP/graphics extensions was the Sun Ultra VIS.

.-----------------------------------------------------------------.
| Orlando Andico                email: orly AT gibson DOT eee DOT upd DOT edu DOT ph |
| IRC Lab/EE Dept/UP Diliman   http://gibson.eee.upd.edu.ph/~orly |
|  "through adventure we are not adventuresome" -- 10000 Maniacs  |
`-----------------------------------------------------------------'

- Raw text -