[Intel-gfx] [PATCH 2/7] drm/i915: Fix computation of last_adjustment for RPS autotuning
Chris Wilson
chris at chris-wilson.co.uk
Wed Mar 18 02:20:54 PDT 2015
On Wed, Mar 18, 2015 at 12:26:49PM +0530, Deepak S wrote:
>
>
> On Friday 06 March 2015 08:36 PM, Chris Wilson wrote:
> >Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> >---
> > drivers/gpu/drm/i915/i915_irq.c | 27 ++++++++++++---------------
> > 1 file changed, 12 insertions(+), 15 deletions(-)
> >
> >diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c
> >index 9baecb79de8c..1296ce37e435 100644
> >--- a/drivers/gpu/drm/i915/i915_irq.c
> >+++ b/drivers/gpu/drm/i915/i915_irq.c
> >@@ -1150,21 +1150,20 @@ static void gen6_pm_rps_work(struct work_struct *work)
> > mutex_lock(&dev_priv->rps.hw_lock);
> > adj = dev_priv->rps.last_adj;
> >+ new_delay = dev_priv->rps.cur_freq;
> > if (pm_iir & GEN6_PM_RP_UP_THRESHOLD) {
> > if (adj > 0)
> > adj *= 2;
> >- else {
> >- /* CHV needs even encode values */
> >- adj = IS_CHERRYVIEW(dev_priv->dev) ? 2 : 1;
> >- }
> >- new_delay = dev_priv->rps.cur_freq + adj;
> >-
> >+ else /* CHV needs even encode values */
> >+ adj = IS_CHERRYVIEW(dev_priv) ? 2 : 1;
> > /*
> > * For better performance, jump directly
> > * to RPe if we're below it.
> > */
> >- if (new_delay < dev_priv->rps.efficient_freq)
> >+ if (new_delay < dev_priv->rps.efficient_freq - adj) {
> > new_delay = dev_priv->rps.efficient_freq;
> >+ adj = 0;
> >+ }
> > } else if (pm_iir & GEN6_PM_RP_DOWN_TIMEOUT) {
> > if (dev_priv->rps.cur_freq > dev_priv->rps.efficient_freq)
> > new_delay = dev_priv->rps.efficient_freq;
> >@@ -1176,24 +1175,22 @@ static void gen6_pm_rps_work(struct work_struct *work)
>
> I think we should modify adj in GEN6_PM_RP_UP_EI_EXPIRED?
> if not not we might request higher freq since we add adj to new_delay before request freq.
The best way to resolve the conflict appears to be just to reorder this
patch later after the removal of the vlv specific adj paths
-Chris
--
Chris Wilson, Intel Open Source Technology Centre
More information about the Intel-gfx
mailing list