Massive power regression going 3.4->3.5

[Ben Widawsky]

On 2012-08-02 00:20, James Bottomley wrote:
> On Wed, 2012-08-01 at 22:08 -0700, bwidawsk wrote:
>> On 2012-08-01 03:06, Chris Wilson wrote:
>> > On Wed, 01 Aug 2012 10:38:36 +0100, James Bottomley
>> > <James.Bottomley at HansenPartnership.com> wrote:
>> >> On Wed, 2012-08-01 at 09:58 +0100, Chris Wilson wrote:
>> >> > On Wed, 01 Aug 2012 09:45:04 +0100, James Bottomley
>> >> <James.Bottomley at HansenPartnership.com> wrote:
>> >> > > On Wed, 2012-08-01 at 09:16 +0100, Chris Wilson wrote:
>> >> > > > On Wed, 01 Aug 2012 09:06:12 +0100, James Bottomley
>> >> <James.Bottomley at HansenPartnership.com> wrote:
>> >> > > > > I got the attached to apply and it doesn't really improve
>> >> the idle power
>> >> > > > > much (12.5W).
>> >> > > >
>> >> > > > That's good to know. Next step is to try overriding
>> >> i915.semaphores.
>> >> > > > Can you please test with i915.semaphores=0 and
>> >> i915.semaphores=1?
>> >> > >
>> >> > > There's not much point doing i915_semaphores=1 since that's 
>> the
>> >> default
>> >> > > on gen 6 hardware, but i915_semaphores=0 recovers and idle 
>> power
>> >> of
>> >> > > ~6.5W
>> >> >
>> >> > It is only the default if iommu is off, and changing the 
>> default
>> >> > was one of the side-effects of the patch you bisected.
>> >> >
>> >> > Can you please login to the desktop, let it idle, record
>> >> > /sys/kernel/debug/dri/0/i915_cur_delayinfo and 
>> .../i915_drpc_info.
>> >> > Then trace-cmd record -e i915 sleep 10s,
>> >>
>> >> OK, what is trace-cmd?  It looks similar to perf tools ... is 
>> that
>> >> it?
>> >
>> > Yes, it is roughly equivalent and you should be able to achieve 
>> the
>> > same
>> > with perf trace - except I haven't done it before so I don't have
>> > quick
>> > advice on how to drive it. :)
>> > -Chris
>>
>> Should be something like:
>> perf record -f -g -e i915:* -a
>
> I already sent the output of trace-cmd ... is that enough?
>
> James

Yes, should do. Have we already eliminated the obvious? GPU semaphores 
will
give time back to the GPU clients normally waiting on such things, X, 
XFCE,
whatever else. It'd probably be handy to begin investigating what those 
guys
are doing with their new extra time. Chris, this is what I was getting 
at on
IRC the other day. What do you think?