[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: SUMMARY: pentium optimized debian discussion



On 29-Jul-1998, Stephen Zander <gibreel@pobox.com> wrote:
> >>>>> "Joey" == Joey Hess <joey@kitenet.net> writes:
>     Joey> If we can clear up the naming problem and people agree with
>     Joey> the points in this summary about speed, then I think we can
>     Joey> set up the architecture on the ftp site now.
> 
> I've sat on one side through all of this but one thing occurs to me:
> 
> Waht, if any, difference do the MMX instructions make?  My IBM 760XL
> laptop (166MHz MMX) reports 330.96 bogo-mips & I've always presumed
> that MMX was responsible.  Does pegcs/egcs -mpentim support the MMX
> instructions yet?

MMX makes no difference for most users.

The MMX CPU was cunningly shipped with a larger L1 cache, so it was
always going to be faster than non-MMX smaller cache pentiums.

The MMX instructions are SIMD instructions (single instruction, multiple
data) -- they perform integer operations on several bytes at a time.
This is useful for some graphics manipulation code and a few other
applications.  You need to write special assembly code to do it, it's
pretty difficult for a compiler to take advantage of.

I have heard the AMD-K6-2 has "3d instructions" which as similar, but perform
multiple floating point instructions at once.  This is very useful for
some 3d applications (games).

And, what's more, bogo-mips are bogus, so don't rely on them as anything
more than a conversation piece.

Tyson.


--  
To UNSUBSCRIBE, email to debian-devel-request@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org


Reply to: