Brian Paul wrote: > Brian Paul wrote: >> Ian Romanick wrote: >>> -----BEGIN PGP SIGNED MESSAGE----- >>> Hash: SHA1 >>> >>> Michael Clark wrote: >>> >>>> I did a git bisect and found the revision where I see the biggest >>>> drop in performance (roughly 30% drop with my app): >>>> >>>> $ git bisect bad >>>> Bisecting: 0 revisions left to test after this >>>> [20f3497e4b6756e330f7b3f54e8acaa1d6c92052] i965: re-org of some of >>>> the new constant buffer code >>>> >>>> I think there may be some other performance regressions in there as >>>> the commit before this is also not as fast as mesa_7_4_branch >>>> although the above commit has the most noticeable drop. >>>> >>>> I think gears is the same speed before and after the commit so I'm >>>> not yet sure what part if my app is exercising the code that has >>>> the performance drop. I'll try to work up a small test program if I >>>> get time. >>> >>> I've been experiencing this with some of our internal test cases. The >>> killer seems to be programs that do lots of state changes. glxgears >>> does almost no state changes, so it's not surprising that it doesn't >>> take a big hit. >> >> That's one of my commits. I'll take a look... > > OK, I've just pushed a commit which should restore performance back to > what it was.
OK great! I now don't see any appreciable performance difference between master and mesa_7_4_branch. Michael. ------------------------------------------------------------------------------ Register Now & Save for Velocity, the Web Performance & Operations Conference from O'Reilly Media. Velocity features a full day of expert-led, hands-on workshops and two days of sessions from industry leaders in dedicated Performance & Operations tracks. Use code vel09scf and Save an extra 15% before 5/3. http://p.sf.net/sfu/velocityconf _______________________________________________ Mesa3d-dev mailing list Mesa3d-dev@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/mesa3d-dev
