sandmann at daimi.au.dk
Tue Aug 7 09:59:06 PDT 2007
Michel Dänzer <michel at tungstengraphics.com> writes:
> > And why is it not used pixman_blt_mmx instead?
> I was wondering the same thing, Søren?
I don't think there is any deep reason. When I did it originally, I
was generally profiler-driven, and this particular path probably just
never showed up.
In principle there is no reason memcpy() has to be slower than
pixman_blt_mmx(), but of course in practice it may well be. It does
look like at least one of the MEMCPY_WRAPPED() calls in fbBlt() could
be done with pixman_blt_mmx().
More information about the xorg