lkml.org 
[lkml]   [1997]   [Feb]   [8]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
SubjectRe: MMX for kernel
Date
In article <m0vsxdq-000QdzC@nevets.oau.org> ssd@nevets.oau.org (Steven S. Dick) writes:

> >Actually, what we could do is put in better logic to
> >determine whether or not the FPU is in use by a task (which is harder
> >than you think because crt0 does zero them out, so you have to filter
> >out that case)
>
> Would it be possible (or even useful) to change the crt0 code to
> zero the fpu area in the task struct rather than actually touch
> the fpu itself?

Tytso's wording wasn't exact. What crt0/crt1 does is to set the FPU
status word to the rounding mode etc. This has to be once at the
startup since there is no way to control the use of the FPU
afterwards.

If you *really* want to do weird things with the FPU and optimize for
non-FPU using programs (which are rather infrequent) then you should
think about a way to intercept the first status word setup
instruction. I don't know how easy and/or useful it is but I don't
see a way how on user-level you could implement lazy initialization.

--
-- Uli
---------------. drepper@cygnus.com ,-. Rubensstrasse 5
Ulrich Drepper \ ,-------------------' \ 76149 Karlsruhe/Germany
Cygnus Solutions `--' drepper@gnu.ai.mit.edu `------------------------
\
 
 \ /
  Last update: 2005-03-22 13:38    [from the cache]
©2003-2011 Jasper Spaans