Messages in this thread |  | | | Date | Tue, 25 Mar 2008 08:49:13 -0700 | | From | "Ray Lee" <> | | Subject | Re: Serious performance regression in Wine applications and Linux 2.6.24.* |
| |
On Tue, Mar 25, 2008 at 5:12 AM, Kai <epimetreus@fastmail.fm> wrote: > > On Sat, 22 Mar 2008 21:47:52 -0700, "Ray Lee" <ray-lk@madrabbit.org> > said: > > > > On Sat, Mar 22, 2008 at 7:49 PM, Kai <epimetreus@fastmail.fm> wrote: > > > Performance degrades by about 60% when I run Wine under any of the > > > 2.6.24.* kernels. Attached are the output of lspci -vv and the two > > > config files of each kernel. > > > > > > I upgraded to 2.6.24, back to 2.6.23.(not sure), then tried 2.6.24.1 and > > > 2.6.24.3 and the issue is present in both; it's severe enough that Wine > > > apps are virtually unusable for me with this version of the kernel; I'm > > > having to use 2.6.23 until this somehow is resolved. > > > > > > I'd like some help figuring out why this performance regression exists, > > > and what can be done to mitigate it. > > > > As wine has a 'wineserver' running in a separate process, it may be > > related to scheduler changes. > > > > Regardless, if you have the time, please retest using he latest git > > head (or nightly snapshot), and see if the performance regression is > > still there, and report back. (There have been a lot of changes > > between 2.6.24 and current git head that impact the scheduler.) > > > > Please ensure that the fair group scheduler is disabled in your tests > > (just as you have in your 2.6.24 config you attached). > > As mentioned in another response, it was happening as recently as > 2.6.25-rc6-git7; I'm currently performing a git bisect between 2.6.23 > and 2.6.24, unless someone has a better idea; it seems my best option, > as I'm not really very experienced with kernel hacking or debugging. >
Andi's idea of looking for excessive context switches is good -- I didn't see a response to that one. Other than that, if you're only noticing the issue in 3d games, then it could be several things (not just the scheduler). Even just a few bisects (or testings of nightly snapshots) would help narrow it down.
Ray
|  |