On Fri, Oct 22, 1999 at 06:52:10PM +0200, Goswin Brederlow wrote:
> arto.astala@nokia.com writes:
> > machines. I think this includes Pentium optimizations,
> Pentium optimisation will only benefit original Pentium-mmx cpus and
> slowdown all other cpus.
Not MMX - anything that uses MMX won't run on a non-MMX chip without
also providing a non-MMX version of the code, and in any case none of
the current free compilers make a good attempt at MMX optimization
(although there is a new MMX patch for GCC which may work better, but
it's not in any released versions).
The main argument against building optimized versions for newer
processors is that most of the time it's not actually going to achieve
anything noticable on the faster machine. If it's going to make a
visible difference for most applications it will be on the slower
machine.
> > instructions that just give a different result. Intel
> > has tried hard to avoid these, however, so it may be that
> > there are no such things.
> Ouch, I hope so. Everybody would shoot Intel if they have such and if
> for that reason windows crashes.
Well, with some code (particularly FP code) the exact set of operations
performed can make a noticable difference to the results. Robust code
working on good data shouldn't have much problem, but there can still be
lower order errors.
--
Mark Brown mailto:broonie@tardis.ed.ac.uk (Trying to avoid grumpiness)
http://www.tardis.ed.ac.uk/~broonie/
EUFS http://www.eusa.ed.ac.uk/societies/filmsoc/
Attachment:
pgpC3k3XJ1t2T.pgp
Description: PGP signature