| Message ID | 1401295810-2081-1-git-send-email-jbarnes@virtuousgeek.org (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Wed, 2014-05-28 at 09:50 -0700, Jesse Barnes wrote: > This saves many ms per call on my BYT by eliminating Punit communication > from the hw readout paths. > > Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org> > --- > drivers/gpu/drm/i915/intel_pm.c | 8 +++----- > 1 file changed, 3 insertions(+), 5 deletions(-) > > diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c > index 452518f..09a3677 100644 > --- a/drivers/gpu/drm/i915/intel_pm.c > +++ b/drivers/gpu/drm/i915/intel_pm.c > @@ -5572,13 +5572,11 @@ bool intel_display_power_enabled(struct drm_i915_private *dev_priv, > > mutex_lock(&power_domains->lock); > for_each_power_well_rev(i, power_well, BIT(domain), power_domains) { > - if (power_well->always_on) > + if (power_well->always_on || power_well->count) > continue; > > - if (!power_well->ops->is_enabled(dev_priv, power_well)) { > - is_enabled = false; > - break; > - } > + is_enabled = false; > + break; > } > mutex_unlock(&power_domains->lock); This was meant to return the HW state vs. the state based on the refcount. It would work in the above way now, because we enable all power wells for driver init time, but if remove that in the future the two states may not match. Perhaps we could maintain a cached version of the HW state in the power_well struct that we set in the sync_hw handler and update whenever we turn on/off the wells. --Imre
On Wed, 28 May 2014 21:09:27 +0300 Imre Deak <imre.deak@intel.com> wrote: > On Wed, 2014-05-28 at 09:50 -0700, Jesse Barnes wrote: > > This saves many ms per call on my BYT by eliminating Punit communication > > from the hw readout paths. > > > > Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org> > > --- > > drivers/gpu/drm/i915/intel_pm.c | 8 +++----- > > 1 file changed, 3 insertions(+), 5 deletions(-) > > > > diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c > > index 452518f..09a3677 100644 > > --- a/drivers/gpu/drm/i915/intel_pm.c > > +++ b/drivers/gpu/drm/i915/intel_pm.c > > @@ -5572,13 +5572,11 @@ bool intel_display_power_enabled(struct drm_i915_private *dev_priv, > > > > mutex_lock(&power_domains->lock); > > for_each_power_well_rev(i, power_well, BIT(domain), power_domains) { > > - if (power_well->always_on) > > + if (power_well->always_on || power_well->count) > > continue; > > > > - if (!power_well->ops->is_enabled(dev_priv, power_well)) { > > - is_enabled = false; > > - break; > > - } > > + is_enabled = false; > > + break; > > } > > mutex_unlock(&power_domains->lock); > > This was meant to return the HW state vs. the state based on the > refcount. It would work in the above way now, because we enable all > power wells for driver init time, but if remove that in the future the > two states may not match. > > Perhaps we could maintain a cached version of the HW state in the > power_well struct that we set in the sync_hw handler and update whenever > we turn on/off the wells. Yeah that works too, and I almost didn't send this out because I knew the hw state is something we want to read out sometimes. But on BYT it turns out to be pretty expensive to do the simple "get power well" calls at the top of the display functions, so it would be nice to find a way to avoid it.
diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 452518f..09a3677 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -5572,13 +5572,11 @@ bool intel_display_power_enabled(struct drm_i915_private *dev_priv, mutex_lock(&power_domains->lock); for_each_power_well_rev(i, power_well, BIT(domain), power_domains) { - if (power_well->always_on) + if (power_well->always_on || power_well->count) continue; - if (!power_well->ops->is_enabled(dev_priv, power_well)) { - is_enabled = false; - break; - } + is_enabled = false; + break; } mutex_unlock(&power_domains->lock);
This saves many ms per call on my BYT by eliminating Punit communication from the hw readout paths. Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org> --- drivers/gpu/drm/i915/intel_pm.c | 8 +++----- 1 file changed, 3 insertions(+), 5 deletions(-)