On 11/24/11 07:24, Petr Sumbera wrote:
On 11/24/11 16:20, Shawn Walker wrote:
On 11/24/11 04:31, Petr Sumbera wrote:
Hi,
now since Solaris 11 is capable to run on 64bit HW only and Studio
compiler option '-m64' together with '-xO4' produce on Intel SSE2 code
we should allow also SSE2 code for 32bit binaries.
I did test with zlib compression and got about 18% time reduction for
compression. Which is I think very nice...
Not sure what we can do for sparc here. But I'm sure there can be done
something too.
Any comments?
I found the same thing during building python2.6/2.7 (well, better
performance, but not quite as much as you).
-xtarget=opteron -xarch=sse2 -xcache=generic
The above options gave the best performance on my Ultra 27, but should
still be compatible even with the oldest 64-bit x86 processors we
support (such as early steppings of AMD opteron).
I would note though that most of the performance boost you're seeing
probably came from using -m64 and not -xarch=sse2.
No, those were results from 32bit ZLIB with added -xarch=sse2.
Ah, then my advise still applies :-)
-Shawn
_______________________________________________
userland-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/userland-discuss
