On Wed, Aug 24, 2005 at 03:28:12AM +1000, matthew green wrote:
> 
> so the question in my mind now is - do we handle FPU owner when
> someone executes VIS instructions properly?

I believe so. Since libc's memcpy() uses VIS we had this torture tested
(I'm not sure about the current default with gcc -O2, we might use the gcc
generated version mostly).

Anyway, what are you after? The register corruption we talked about earlier
in this thread happens to all locals and out registers - how would that
corelate to lazy fpu saving?

Martin