Re: [PATCH 0/9] gpu hang and swizzle patches

[Chris Wilson <chris@chris-wilson.co.uk>]

On Thu, 10 Nov 2011 14:17:58 +0100, Daniel Vetter <daniel.vetter@ffwll.ch> wrote:
> Hi all,
> 
> This is a bit a mixed pile, but I've used all the earlier patches to test
> the gen6+ swizzling patch and I like to send out patch series somewhat
> resembling the setup I've tested them in.
> 
> Patches 1-2 refactor our debugfs code a bit.
> Patches 3-5 implement a debugfs interface to simulate a gpu hang Patch 6
> fixes the swizzle detection on i915G/i945G. This one is for stable and
> independent of the previous patches.
> Patches 7-8 add a debugfs file with information to debug swizzle issues
> Patch 9 implements swizzle support for snb/ivb when in dual-channel mode
> 
> Benchmarking on my snb is notoriously difficult, but swizzling seems to
> yield a few percent in some workloads, topping out at 5% for padman.
> 
> Review highly welcome.

I've had a play and could barely detect any improvement above the noise.
However, I've read through all the patches and passed on my comments and
I don't see anything fundamentally wrong with them (though I see enough
GPU hangs not to need to manually trigger one ;-).
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre