2009/2/3 Pekka Pietikainen pp@ee.oulu.fi:
On Tue, Feb 03, 2009 at 12:11:59PM -0500, Warren Togami wrote:
Jakub recommended doing comparison tests of -m32 -march=i586 -mtune=generic vs. -m32 -march=i686 -mtune=generic with Spec2k and Spec2006 benchmark tools. Anyone interested in trying this?
In related news, cebbert wants to do the following to the F-11 kernel:
- Eliminate the current i686 kernel.
- i686 hardware would get i686 PAE by default.
- i586 kernel becomes i686, except without cmov. This is primarily so
people don't complain when they realize they have the "i586" kernel.
Just a note, I believe that the ia32 support should "concentrate" on netbooks (So Intel Atom), those are the only new things on the market that can't run x86_64, and the userbase of those will just grow and every percent of extra performance will make many people happy.
A benchmark run on one of those might be very useful to make the final decision, not that the more server-oriented tests are very useful for the typical netbook userbase.
And then there's the "should work even with this" decision, and i686 without cmov is a reasonable minimum there. Could be i686+cmov or +sse2, but those will make a lot of people unhappy.
Does the right wording for this wouldn't be to move i586 to secondary architecture and welcome i686 with sse(2?) as the new default x86_32 and primary arch ?
Having a project to build fews packages targeted as i586 and eventually using dietlibc for dhcp/dns server usage (other?) would be more valuable than having the fast majority of packages compiled with old options?
Nicolas (kwizart)