diff mbox

drm/i915: use hsw rps tuning values everywhere on gen6+

Message ID 1345020105-5933-1-git-send-email-daniel.vetter@ffwll.ch (mailing list archive)
State New, archived
Headers show

Commit Message

Daniel Vetter Aug. 15, 2012, 8:41 a.m. UTC 
James Bottomley reported [1] a massive power regression, due to the
enabling of semaphores by default in 3.5. A workaround for him is to
again disable semaphores. And indeed, his system has a very hard time
to entre rc6 with semaphores enabled.

Ben Widawsky run around with a kill-a-watt a lot and noticed:
- There are indeed a few rare systems that seem to have a hard time
  entering rc6 when desktop-idle.
- One machine, The Indestructible Toshiba regressed in this behaviour
  between 3.5 and 3.6 in a merge commit! So rc6 behaviour with the
  current setting seems to be highly timing dependent and not robust
  at all.
- The behaviour James reported wrt semaphores seems to be a freak
  timing thing that only happens on his specific machine, confirming
  that enabling semaphores shouldn't reduce rc6 residency.

Now furthermore the Google ChromeOS guys reported [2] a while ago that
at least on some machines a simply a blinking cursor can keep the gpu
turbo at the highest frequency. This is because the current rps limits
used on snb/ivb are highly asymmetric.

On the theory that gpu turbo and rc6 tuning values are related, we've
tried whether the much saner looking (since much less asymmetric) rps
tuning values used for hsw would also help entering rc6 more robustly.

And it seems to work.

Reference[1]: http://lists.freedesktop.org/archives/dri-devel/2012-July/025675.html
Reference[2]: http://lists.freedesktop.org/archives/intel-gfx/2012-July/018692.html
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=53393
Tested-by: Ben Widawsky <ben@bwidawsk.net>
Cc: stable@vger.kernel.org
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
---
 drivers/gpu/drm/i915/intel_pm.c |   15 ++++-----------
 1 file changed, 4 insertions(+), 11 deletions(-)

Comments

Paul Menzel Aug. 15, 2012, 2:05 p.m. UTC | #1 
Am Mittwoch, den 15.08.2012, 10:41 +0200 schrieb Daniel Vetter:
> James Bottomley reported [1] a massive power regression, due to the
> enabling of semaphores by default in 3.5. A workaround for him is to
> again disable semaphores. And indeed, his system has a very hard time
> to entre rc6 with semaphores enabled.

s,entre,enter,

> Ben Widawsky run around with a kill-a-watt a lot and noticed:
> - There are indeed a few rare systems that seem to have a hard time
>   entering rc6 when desktop-idle.
> - One machine, The Indestructible Toshiba regressed in this behaviour

Never heard that Toshiba name.

>   between 3.5 and 3.6 in a merge commit! So rc6 behaviour with the
>   current setting seems to be highly timing dependent and not robust
>   at all.
> - The behaviour James reported wrt semaphores seems to be a freak
>   timing thing that only happens on his specific machine, confirming
>   that enabling semaphores shouldn't reduce rc6 residency.
> 
> Now furthermore the Google ChromeOS guys reported [2] a while ago that
> at least on some machines a simply a blinking cursor can keep the gpu
> turbo at the highest frequency. This is because the current rps limits
> used on snb/ivb are highly asymmetric.
> 
> On the theory that gpu turbo and rc6 tuning values are related, we've
> tried whether the much saner looking (since much less asymmetric) rps
> tuning values used for hsw would also help entering rc6 more robustly.
> 
> And it seems to work.
> 
> Reference[1]: http://lists.freedesktop.org/archives/dri-devel/2012-July/025675.html
> Reference[2]: http://lists.freedesktop.org/archives/intel-gfx/2012-July/018692.html
> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=53393
> Tested-by: Ben Widawsky <ben@bwidawsk.net>

Did James already confirm, that this fixes his problem?

> Cc: stable@vger.kernel.org
> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
> ---
>  drivers/gpu/drm/i915/intel_pm.c |   15 ++++-----------
>  1 file changed, 4 insertions(+), 11 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c
> index d64dffb..cf10a1d 100644
> --- a/drivers/gpu/drm/i915/intel_pm.c
> +++ b/drivers/gpu/drm/i915/intel_pm.c
> @@ -2482,17 +2482,10 @@ static void gen6_enable_rps(struct drm_device *dev)
>  		   dev_priv->rps.max_delay << 24 |
>  		   dev_priv->rps.min_delay << 16);
>  
> -	if (IS_HASWELL(dev)) {
> -		I915_WRITE(GEN6_RP_UP_THRESHOLD, 59400);
> -		I915_WRITE(GEN6_RP_DOWN_THRESHOLD, 245000);
> -		I915_WRITE(GEN6_RP_UP_EI, 66000);
> -		I915_WRITE(GEN6_RP_DOWN_EI, 350000);
> -	} else {
> -		I915_WRITE(GEN6_RP_UP_THRESHOLD, 10000);
> -		I915_WRITE(GEN6_RP_DOWN_THRESHOLD, 1000000);
> -		I915_WRITE(GEN6_RP_UP_EI, 100000);
> -		I915_WRITE(GEN6_RP_DOWN_EI, 5000000);
> -	}
> +	I915_WRITE(GEN6_RP_UP_THRESHOLD, 59400);
> +	I915_WRITE(GEN6_RP_DOWN_THRESHOLD, 245000);
> +	I915_WRITE(GEN6_RP_UP_EI, 66000);
> +	I915_WRITE(GEN6_RP_DOWN_EI, 350000);
>  
>  	I915_WRITE(GEN6_RP_IDLE_HYSTERSIS, 10);
>  	I915_WRITE(GEN6_RP_CONTROL,

Acked-by: Paul Menzel <paulepanter@users.sourceforge.net>


Thanks,

Paul
Daniel Vetter Aug. 15, 2012, 10:38 p.m. UTC | #2 
On Wed, Aug 15, 2012 at 03:36:34PM +0100, James Bottomley wrote:
> On Wed, 2012-08-15 at 16:05 +0200, Paul Menzel wrote:
> > Am Mittwoch, den 15.08.2012, 10:41 +0200 schrieb Daniel Vetter:
> > >   between 3.5 and 3.6 in a merge commit! So rc6 behaviour with the
> > >   current setting seems to be highly timing dependent and not robust
> > >   at all.
> > > - The behaviour James reported wrt semaphores seems to be a freak
> > >   timing thing that only happens on his specific machine, confirming
> > >   that enabling semaphores shouldn't reduce rc6 residency.
> > > 
> > > Now furthermore the Google ChromeOS guys reported [2] a while ago that
> > > at least on some machines a simply a blinking cursor can keep the gpu
> > > turbo at the highest frequency. This is because the current rps limits
> > > used on snb/ivb are highly asymmetric.
> > > 
> > > On the theory that gpu turbo and rc6 tuning values are related, we've
> > > tried whether the much saner looking (since much less asymmetric) rps
> > > tuning values used for hsw would also help entering rc6 more robustly.
> > > 
> > > And it seems to work.
> > > 
> > > Reference[1]: http://lists.freedesktop.org/archives/dri-devel/2012-July/025675.html
> > > Reference[2]: http://lists.freedesktop.org/archives/intel-gfx/2012-July/018692.html
> > > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=53393
> > > Tested-by: Ben Widawsky <ben@bwidawsk.net>
> > 
> > Did James already confirm, that this fixes his problem?
> 
> Well, no ... I think no-one cc'd me on anything after the initial bug
> report, but the patch won't apply to 3.5, so cc stable isn't really
> going to work ... it will need backporting.
> 
> I can test either the backport or 3.6-rc1 with the patch if there's
> interest.

Sorry, the cc: you got lost, testing feedback highly welcome. The ChromeOS
guys just reported back that for them fully symmetric values actually lead
to the least power consumption for light gpu loads (which these are not
yet), so maybe we need to tune things some more even.

Thanks, Daniel
Daniel Vetter Aug. 20, 2012, 6:44 p.m. UTC | #3 
On Wed, Aug 15, 2012 at 10:41:45AM +0200, Daniel Vetter wrote:
> James Bottomley reported [1] a massive power regression, due to the
> enabling of semaphores by default in 3.5. A workaround for him is to
> again disable semaphores. And indeed, his system has a very hard time
> to entre rc6 with semaphores enabled.
> 
> Ben Widawsky run around with a kill-a-watt a lot and noticed:
> - There are indeed a few rare systems that seem to have a hard time
>   entering rc6 when desktop-idle.
> - One machine, The Indestructible Toshiba regressed in this behaviour
>   between 3.5 and 3.6 in a merge commit! So rc6 behaviour with the
>   current setting seems to be highly timing dependent and not robust
>   at all.
> - The behaviour James reported wrt semaphores seems to be a freak
>   timing thing that only happens on his specific machine, confirming
>   that enabling semaphores shouldn't reduce rc6 residency.
> 
> Now furthermore the Google ChromeOS guys reported [2] a while ago that
> at least on some machines a simply a blinking cursor can keep the gpu
> turbo at the highest frequency. This is because the current rps limits
> used on snb/ivb are highly asymmetric.
> 
> On the theory that gpu turbo and rc6 tuning values are related, we've
> tried whether the much saner looking (since much less asymmetric) rps
> tuning values used for hsw would also help entering rc6 more robustly.
> 
> And it seems to work.
> 
> Reference[1]: http://lists.freedesktop.org/archives/dri-devel/2012-July/025675.html
> Reference[2]: http://lists.freedesktop.org/archives/intel-gfx/2012-July/018692.html
> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=53393
> Tested-by: Ben Widawsky <ben@bwidawsk.net>
> Cc: stable@vger.kernel.org
> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

Ok, I've merged this one here, making the commit message a bit more
cautious. Unfortnately we don't seem to have the understanding (nor the
resources to gain it) of the hw to do better than copy&paste what we get
from the hw team :(
-Daniel
diff mbox

Patch

diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c
index d64dffb..cf10a1d 100644
--- a/drivers/gpu/drm/i915/intel_pm.c
+++ b/drivers/gpu/drm/i915/intel_pm.c
@@ -2482,17 +2482,10 @@  static void gen6_enable_rps(struct drm_device *dev)
 		   dev_priv->rps.max_delay << 24 |
 		   dev_priv->rps.min_delay << 16);
 
-	if (IS_HASWELL(dev)) {
-		I915_WRITE(GEN6_RP_UP_THRESHOLD, 59400);
-		I915_WRITE(GEN6_RP_DOWN_THRESHOLD, 245000);
-		I915_WRITE(GEN6_RP_UP_EI, 66000);
-		I915_WRITE(GEN6_RP_DOWN_EI, 350000);
-	} else {
-		I915_WRITE(GEN6_RP_UP_THRESHOLD, 10000);
-		I915_WRITE(GEN6_RP_DOWN_THRESHOLD, 1000000);
-		I915_WRITE(GEN6_RP_UP_EI, 100000);
-		I915_WRITE(GEN6_RP_DOWN_EI, 5000000);
-	}
+	I915_WRITE(GEN6_RP_UP_THRESHOLD, 59400);
+	I915_WRITE(GEN6_RP_DOWN_THRESHOLD, 245000);
+	I915_WRITE(GEN6_RP_UP_EI, 66000);
+	I915_WRITE(GEN6_RP_DOWN_EI, 350000);
 
 	I915_WRITE(GEN6_RP_IDLE_HYSTERSIS, 10);
 	I915_WRITE(GEN6_RP_CONTROL,