lkml.org 
[lkml]   [1997]   [Feb]   [6]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
SubjectRe: MMX for kernel
Date
> 
> On Wed, 5 Feb 1997, Marc Lehmann wrote:
>
> There are a few things that MMX does well. Most obviously, it does
> parallel adds, shifts, and multiplies. But I believe that since it's using
> the FPU data paths, it can also do 64-bit wide memory cycles (ala
> pentium-memcpy patch). Looking at the spec, it should be possible to read
> eight bytes, unpack them into eight shorts, and accumulate them in four
> accumulators quite quickly, possibly faster than the standard memcpy. I
> don't have the hardware to try it with yet though.

Hmm. I've been thinking about this, and since MMX instructions use the
FPU data paths, I'm not sure it's such a good idea to use MMX instructions.
I read somewhere that some special instruction is used to kind of
'switch' between MMX and FPU mode, and that this takes some 50 cycles.
Now, wouldn't that mean that we'd waste a lot of time here,
for instance, if some programs make heavy use of FPU and at the same time
makes a lot of system calls involving MMX instructions?

/stefan


\
 
 \ /
  Last update: 2005-03-22 13:38    [W:0.264 / U:0.340 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site