🚀 go-pugleaf

In article <31ab1f2f.4414157@news-win.rinet.ru>, minister@rinet.ru says...
>P5-200) and don't  bother about MMX AT ALL. Because the software
>support for MMX is MUCH more important than hardware implementation.
>Just like P5's and P6's new operands - they probably could speed up
>some applications, but nobody use them just not to do their code not
>capable of running on 486, P5 clones or something else. So even when
>MMX processors will be available, you won't be able to use their
>features - software companies shouldn't make software that wont work
>on 99% processors. It's a waste of money for them. And this situation
>is going to continue for a couple of years. IMHO, of course.
>Sorry for poor English.

I have to disagee here.  I expect the MMX to catch on and quickly lead to a 
situation analogous to floating point coprocessors.  An emulator library is 
written to support "old" hardware without MMX so software auotmatically detects 
MMX and uses it, otherwise it calls the library code to accomplish the same 
result (although slower).  

Anyone who runs real calculations (fft's, matrix algebra, etc) doesn't consider 
a machine without FP hardware (this has been true since the XT), so people 
serious about multimedia won't consider machine without MMX asuming the 
hardware actually delivers the improvements promised.  I expect it will.

Don't forget, AMD has licensed this from Intel already, Cyrix will have to or 
else modify its existing extensions to be compatable.

--wally.