On Wed, Feb 08, 2012 at 10:00:54AM +0000, Chris Wilson wrote: > On Wed, 25 Jan 2012 14:04:00 +0100, Daniel Vetter <daniel.vetter at ffwll.ch> wrote: > > We still have reports of missed irqs even on Sandybridge with the > > HWSTAM workaround in place. Testing by the bug reporter gets rid of > > them with the forcewake voodoo and no HWSTAM writes. > > > > Because I've slightly botched the rebasing I've left out the ACTHD > > readback which is also required to get IVB working. Seems to still > > work on the tester's machine, so I think we should go with the more > > minmal approach on SNB. Especially since I've only found weak evidence > > for holding forcewake while waiting for an interrupt to arrive, but > > none for the ACTHD readback. > > > > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=45181 > > Tested-by: Nicolas Kalkhof nkalkhof()at()web.de > > Signed-Off-by: Daniel Vetter <daniel.vetter at ffwll.ch> > > Whilst this patch doesn't directly enable the ACTHD w/a, I have observed > that with a very light load (drawing a single large transformed texture) > reading back the ACTHD (along with the GT forcewake dance) becomes the > predominant consumer of CPU time for the system. (The rate-limiting step > is still the GPU, it just irked me to see the kernel consume more CPU time > than X.) > > This workaround appears more successful than the last, and doesn't > appear to break anything else, so > > Acked-by: Chris Wilson <chris at chris-wilson.co.uk> I've picked up patches 1,2 & 4 of this series, thanks for taking a look at them. I'll postpone 3 until we do the seqno/request refactoring for real, atm there's too much stuff in this area outstanding and we need to judge this one in the context of the real thing. -Daniel -- Daniel Vetter Mail: daniel at ffwll.ch Mobile: +41 (0)79 365 57 48
