I noticed this today when I decided to try to update my patch
for SMP lazy FPU restore (but not lazy save if ncpus>1), after having
finally written a simple benchmark. This patch reduces the differences
between uniprocessor and SMP, although not as much as removal of lazy
floating point saves and restores.
When I originally wrote that patch, I was motivated by some
remarks claiming that the performance difference could be as high as
30% (perhaps for older CPU's?) for some benchmarks, but I had not gotten
around to writing a benchmark to test this patch until now. So, I am
rather curious to know if lazy FPU save and restore is worthwhile or not,
or if the truth is really not known.
I would be happy to update my SMP lazy FP patch and submit it
for integration, but if it is known that this is not worth doing, I'd
like to know.
Adam J. Richter __ ______________ 4880 Stevens Creek Blvd, Suite 205
adam@yggdrasil.com \ / San Jose, California 95129-1034
+1 408 261-6630 | g g d r a s i l United States of America
fax +1 408 261-6631 "Free Software For The Rest Of Us."
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.altern.org/andrebalsa/doc/lkml-faq.html