| Message ID | 1489592584-10422-6-git-send-email-mika.kuoppala@intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Wed, Mar 15, 2017 at 05:43:04PM +0200, Mika Kuoppala wrote: > Set byt rc residency counters high level as chv does by > default. We lose some accuracy on byt but we can do the calculation > without extra hw read on both platforms, as now they behave > identically in this respect. > > v2: use ktime > > Cc: Chris Wilson <chris@chris-wilson.co.uk> > Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> > Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com> > --- > drivers/gpu/drm/i915/i915_irq.c | 9 ++------- > drivers/gpu/drm/i915/intel_pm.c | 9 +++------ > 2 files changed, 5 insertions(+), 13 deletions(-) > > diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c > index f73d8db..7fb35a5 100644 > --- a/drivers/gpu/drm/i915/i915_irq.c > +++ b/drivers/gpu/drm/i915/i915_irq.c > @@ -1100,11 +1100,6 @@ static u32 vlv_wa_c0_ei(struct drm_i915_private *dev_priv, u32 pm_iir) > if (prev->ktime) { > u64 time, c0; > u32 render, media; > - unsigned int mul; > - > - mul = 1000 * 100; /* scale to threshold% */ > - if (I915_READ(VLV_COUNTER_CONTROL) & VLV_COUNT_RANGE_HIGH) > - mul <<= 8; > > time = ktime_us_delta(now.ktime, prev->ktime); > time *= dev_priv->czclk_freq; > @@ -1116,8 +1111,8 @@ static u32 vlv_wa_c0_ei(struct drm_i915_private *dev_priv, u32 pm_iir) > */ > render = now.render_c0 - prev->render_c0; > media = now.media_c0 - prev->media_c0; > - c0 = max(render, media); > - c0 *= mul; > + c0 = max_t(u64, render, media) << 8; /* upper part of 40 bit */ Keep the comparison at u32, do the promotion afterwards, be kind to 32bit machines, i.e. c0 *= 1000 * 100 << 8; as before With that Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> -Chris
diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index f73d8db..7fb35a5 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -1100,11 +1100,6 @@ static u32 vlv_wa_c0_ei(struct drm_i915_private *dev_priv, u32 pm_iir) if (prev->ktime) { u64 time, c0; u32 render, media; - unsigned int mul; - - mul = 1000 * 100; /* scale to threshold% */ - if (I915_READ(VLV_COUNTER_CONTROL) & VLV_COUNT_RANGE_HIGH) - mul <<= 8; time = ktime_us_delta(now.ktime, prev->ktime); time *= dev_priv->czclk_freq; @@ -1116,8 +1111,8 @@ static u32 vlv_wa_c0_ei(struct drm_i915_private *dev_priv, u32 pm_iir) */ render = now.render_c0 - prev->render_c0; media = now.media_c0 - prev->media_c0; - c0 = max(render, media); - c0 *= mul; + c0 = max_t(u64, render, media) << 8; /* upper part of 40 bit */ + c0 *= 1000 * 100; /* to usecs and scale to threshold% */ if (c0 > time * dev_priv->rps.up_threshold) events = GEN6_PM_RP_UP_THRESHOLD; diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 19fd11b..41d85ce 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -6392,7 +6392,8 @@ static void valleyview_enable_rps(struct drm_i915_private *dev_priv) /* allows RC6 residency counter to work */ I915_WRITE(VLV_COUNTER_CONTROL, - _MASKED_BIT_ENABLE(VLV_MEDIA_RC0_COUNT_EN | + _MASKED_BIT_ENABLE(VLV_COUNT_RANGE_HIGH | + VLV_MEDIA_RC0_COUNT_EN | VLV_RENDER_RC0_COUNT_EN | VLV_MEDIA_RC6_COUNT_EN | VLV_RENDER_RC6_COUNT_EN)); @@ -8361,7 +8362,7 @@ static u64 vlv_residency_raw(struct drm_i915_private *dev_priv, spin_lock_irq(&dev_priv->uncore.lock); saved_ctl = I915_READ_FW(VLV_COUNTER_CONTROL); - if (!(saved_ctl & VLV_COUNT_RANGE_HIGH)) + if (WARN_ON(!(saved_ctl & VLV_COUNT_RANGE_HIGH))) I915_WRITE_FW(VLV_COUNTER_CONTROL, _MASKED_BIT_ENABLE(VLV_COUNT_RANGE_HIGH)); @@ -8382,10 +8383,6 @@ static u64 vlv_residency_raw(struct drm_i915_private *dev_priv, upper = I915_READ_FW(reg); } while (upper != tmp); - if (!(saved_ctl & VLV_COUNT_RANGE_HIGH)) - I915_WRITE_FW(VLV_COUNTER_CONTROL, - _MASKED_BIT_DISABLE(VLV_COUNT_RANGE_HIGH)); - spin_unlock_irq(&dev_priv->uncore.lock); return lower | (u64)upper << 8;
Set byt rc residency counters high level as chv does by default. We lose some accuracy on byt but we can do the calculation without extra hw read on both platforms, as now they behave identically in this respect. v2: use ktime Cc: Chris Wilson <chris@chris-wilson.co.uk> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com> --- drivers/gpu/drm/i915/i915_irq.c | 9 ++------- drivers/gpu/drm/i915/intel_pm.c | 9 +++------ 2 files changed, 5 insertions(+), 13 deletions(-)