[PATCH 4/7] drm/i915: Improved w/a for rps on Baytrail

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Rewrite commit 31685c258e0b0ad6aa486c5ec001382cf8a64212
Author: Deepak S <deepak.s@xxxxxxxxxxxxxxx>
Date:   Thu Jul 3 17:33:01 2014 -0400

    drm/i915/vlv: WA for Turbo and RC6 to work together.

Other than code clarity, the major improvement is to disable the extra
interrupts generated when idle.  However, the reclocking remains rather
slow under the new manual regime, in particular it fails to downclock as
quickly as desired.

Signed-off-by: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx>
Cc: Deepak S <deepak.s@xxxxxxxxxxxxxxx>
Cc: Ville Syrjälä <ville.syrjala@xxxxxxxxxxxxxxx>
Cc: Rodrigo Vivi <rodrigo.vivi@xxxxxxxxx>
Cc: Daniel Vetter <daniel.vetter@xxxxxxxx>
---
 drivers/gpu/drm/i915/i915_irq.c      | 166 ++++++++++++-----------------------
 drivers/gpu/drm/i915/i915_reg.h      |   4 +-
 drivers/gpu/drm/i915/intel_display.c |   2 +
 drivers/gpu/drm/i915/intel_drv.h     |   2 +
 drivers/gpu/drm/i915/intel_pm.c      |  13 +++
 5 files changed, 73 insertions(+), 114 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c
index 8e19d031c05d..2db5dbb87ced 100644
--- a/drivers/gpu/drm/i915/i915_irq.c
+++ b/drivers/gpu/drm/i915/i915_irq.c
@@ -1282,129 +1282,72 @@ static void notify_ring(struct drm_device *dev,
 	i915_queue_hangcheck(dev);
 }
 
-static u32 vlv_c0_residency(struct drm_i915_private *dev_priv,
-			    struct intel_rps_ei *rps_ei)
+static void vlv_c0_read(struct drm_i915_private *dev_priv,
+			struct intel_rps_ei *ei)
 {
-	u32 cz_ts, cz_freq_khz;
-	u32 render_count, media_count;
-	u32 elapsed_render, elapsed_media, elapsed_time;
-	u32 residency = 0;
-
-	cz_ts = vlv_punit_read(dev_priv, PUNIT_REG_CZ_TIMESTAMP);
-	cz_freq_khz = DIV_ROUND_CLOSEST(dev_priv->mem_freq * 1000, 4);
-
-	render_count = I915_READ(VLV_RENDER_C0_COUNT_REG);
-	media_count = I915_READ(VLV_MEDIA_C0_COUNT_REG);
-
-	if (rps_ei->cz_clock == 0) {
-		rps_ei->cz_clock = cz_ts;
-		rps_ei->render_c0 = render_count;
-		rps_ei->media_c0 = media_count;
-
-		return dev_priv->rps.cur_freq;
-	}
-
-	elapsed_time = cz_ts - rps_ei->cz_clock;
-	rps_ei->cz_clock = cz_ts;
-
-	elapsed_render = render_count - rps_ei->render_c0;
-	rps_ei->render_c0 = render_count;
-
-	elapsed_media = media_count - rps_ei->media_c0;
-	rps_ei->media_c0 = media_count;
-
-	/* Convert all the counters into common unit of milli sec */
-	elapsed_time /= VLV_CZ_CLOCK_TO_MILLI_SEC;
-	elapsed_render /=  cz_freq_khz;
-	elapsed_media /= cz_freq_khz;
-
-	/*
-	 * Calculate overall C0 residency percentage
-	 * only if elapsed time is non zero
-	 */
-	if (elapsed_time) {
-		residency =
-			((max(elapsed_render, elapsed_media) * 100)
-				/ elapsed_time);
-	}
-
-	return residency;
+	ei->cz_clock = vlv_punit_read(dev_priv, PUNIT_REG_CZ_TIMESTAMP);
+	ei->render_c0 = I915_READ(VLV_RENDER_C0_COUNT);
+	ei->media_c0 = I915_READ(VLV_MEDIA_C0_COUNT);
 }
 
-/**
- * vlv_calc_delay_from_C0_counters - Increase/Decrease freq based on GPU
- * busy-ness calculated from C0 counters of render & media power wells
- * @dev_priv: DRM device private
- *
- */
-static u32 vlv_calc_delay_from_C0_counters(struct drm_i915_private *dev_priv)
+static bool vlv_c0_above(struct drm_i915_private *dev_priv,
+			 const struct intel_rps_ei *old,
+			 const struct intel_rps_ei *now,
+			 int threshold)
 {
-	u32 residency_C0_up = 0, residency_C0_down = 0;
-	u8 new_delay, adj;
+	u64 time = now->cz_clock - old->cz_clock;
+	u64 c0 = max(now->render_c0 - old->render_c0,
+		     now->media_c0 - old->media_c0);
 
-	dev_priv->rps.ei_interrupt_count++;
+	c0 *= 100 * VLV_CZ_CLOCK_TO_MILLI_SEC * 4 / 1000;
+	time *= threshold * dev_priv->mem_freq;
+	return c0 >= time;
+}
 
-	WARN_ON(!mutex_is_locked(&dev_priv->rps.hw_lock));
+void gen6_rps_reset_ei(struct drm_i915_private *dev_priv)
+{
+	vlv_c0_read(dev_priv, &dev_priv->rps.down_ei);
+	dev_priv->rps.up_ei = dev_priv->rps.down_ei;
+	dev_priv->rps.ei_interrupt_count = 0;
+}
 
+static u32 vlv_wa_c0_ei(struct drm_i915_private *dev_priv, u32 pm_iir)
+{
+	struct intel_rps_ei now;
+	u32 events = 0;
 
-	if (dev_priv->rps.up_ei.cz_clock == 0) {
-		vlv_c0_residency(dev_priv, &dev_priv->rps.up_ei);
-		vlv_c0_residency(dev_priv, &dev_priv->rps.down_ei);
-		return dev_priv->rps.cur_freq;
-	}
+	if ((pm_iir & GEN6_PM_RP_UP_EI_EXPIRED) == 0)
+		return 0;
 
+	vlv_c0_read(dev_priv, &now);
 
 	/*
 	 * To down throttle, C0 residency should be less than down threshold
 	 * for continous EI intervals. So calculate down EI counters
 	 * once in VLV_INT_COUNT_FOR_DOWN_EI
 	 */
-	if (dev_priv->rps.ei_interrupt_count == VLV_INT_COUNT_FOR_DOWN_EI) {
-
+	if (++dev_priv->rps.ei_interrupt_count >= VLV_INT_COUNT_FOR_DOWN_EI) {
+		pm_iir |= GEN6_PM_RP_DOWN_EI_EXPIRED;
 		dev_priv->rps.ei_interrupt_count = 0;
-
-		residency_C0_down = vlv_c0_residency(dev_priv,
-						     &dev_priv->rps.down_ei);
-	} else {
-		residency_C0_up = vlv_c0_residency(dev_priv,
-						   &dev_priv->rps.up_ei);
 	}
 
-	new_delay = dev_priv->rps.cur_freq;
-
-	adj = dev_priv->rps.last_adj;
-	/* C0 residency is greater than UP threshold. Increase Frequency */
-	if (residency_C0_up >= VLV_RP_UP_EI_THRESHOLD) {
-		if (adj > 0)
-			adj *= 2;
-		else
-			adj = 1;
-
-		if (dev_priv->rps.cur_freq < dev_priv->rps.max_freq_softlimit)
-			new_delay = dev_priv->rps.cur_freq + adj;
-
-		/*
-		 * For better performance, jump directly
-		 * to RPe if we're below it.
-		 */
-		if (new_delay < dev_priv->rps.efficient_freq)
-			new_delay = dev_priv->rps.efficient_freq;
+	if (pm_iir & GEN6_PM_RP_DOWN_EI_EXPIRED) {
+		if (!vlv_c0_above(dev_priv,
+				  &dev_priv->rps.down_ei, &now,
+				  VLV_RP_DOWN_EI_THRESHOLD))
+			events |= GEN6_PM_RP_DOWN_THRESHOLD;
+		dev_priv->rps.down_ei = now;
+	}
 
-	} else if (!dev_priv->rps.ei_interrupt_count &&
-			(residency_C0_down < VLV_RP_DOWN_EI_THRESHOLD)) {
-		if (adj < 0)
-			adj *= 2;
-		else
-			adj = -1;
-		/*
-		 * This means, C0 residency is less than down threshold over
-		 * a period of VLV_INT_COUNT_FOR_DOWN_EI. So, reduce the freq
-		 */
-		if (dev_priv->rps.cur_freq > dev_priv->rps.min_freq_softlimit)
-			new_delay = dev_priv->rps.cur_freq + adj;
+	if (pm_iir & GEN6_PM_RP_UP_EI_EXPIRED) {
+		if (vlv_c0_above(dev_priv,
+				 &dev_priv->rps.up_ei, &now,
+				 VLV_RP_UP_EI_THRESHOLD))
+			events |= GEN6_PM_RP_UP_THRESHOLD;
+		dev_priv->rps.up_ei = now;
 	}
 
-	return new_delay;
+	return events;
 }
 
 static void gen6_pm_rps_work(struct work_struct *work)
@@ -1433,14 +1376,14 @@ static void gen6_pm_rps_work(struct work_struct *work)
 
 	mutex_lock(&dev_priv->rps.hw_lock);
 
+	pm_iir |= vlv_wa_c0_ei(dev_priv, pm_iir);
+
 	adj = dev_priv->rps.last_adj;
 	if (pm_iir & GEN6_PM_RP_UP_THRESHOLD) {
 		if (adj > 0)
 			adj *= 2;
-		else {
-			/* CHV needs even encode values */
-			adj = IS_CHERRYVIEW(dev_priv) ? 2 : 1;
-		}
+		else
+			adj = 1;
 		new_delay = dev_priv->rps.cur_freq + adj;
 
 		/*
@@ -1455,15 +1398,11 @@ static void gen6_pm_rps_work(struct work_struct *work)
 		else
 			new_delay = dev_priv->rps.min_freq_softlimit;
 		adj = 0;
-	} else if (pm_iir & GEN6_PM_RP_UP_EI_EXPIRED) {
-		new_delay = vlv_calc_delay_from_C0_counters(dev_priv);
 	} else if (pm_iir & GEN6_PM_RP_DOWN_THRESHOLD) {
 		if (adj < 0)
 			adj *= 2;
-		else {
-			/* CHV needs even encode values */
-			adj = IS_CHERRYVIEW(dev_priv) ? -2 : -1;
-		}
+		else
+			adj = -1;
 		new_delay = dev_priv->rps.cur_freq + adj;
 	} else { /* unknown event */
 		new_delay = dev_priv->rps.cur_freq;
@@ -1475,6 +1414,9 @@ static void gen6_pm_rps_work(struct work_struct *work)
 	new_delay = clamp_t(int, new_delay,
 			    dev_priv->rps.min_freq_softlimit,
 			    dev_priv->rps.max_freq_softlimit);
+	/* CHV needs even encode values */
+	if (IS_CHERRYVIEW(dev_priv))
+		new_delay = new_delay & ~1;
 
 	dev_priv->rps.last_adj = new_delay - dev_priv->rps.cur_freq;
 
diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
index 59eaebce2b02..66de6e3e99bf 100644
--- a/drivers/gpu/drm/i915/i915_reg.h
+++ b/drivers/gpu/drm/i915/i915_reg.h
@@ -5568,8 +5568,8 @@ enum punit_power_well {
 
 #define GEN6_GT_GFX_RC6p			0x13810C
 #define GEN6_GT_GFX_RC6pp			0x138110
-#define VLV_RENDER_C0_COUNT_REG		0x138118
-#define VLV_MEDIA_C0_COUNT_REG			0x13811C
+#define VLV_RENDER_C0_COUNT			0x138118
+#define VLV_MEDIA_C0_COUNT			0x13811C
 
 #define GEN6_PCODE_MAILBOX			0x138124
 #define   GEN6_PCODE_READY			(1<<31)
diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
index a17c2e80f9f2..960a3be39ad7 100644
--- a/drivers/gpu/drm/i915/intel_display.c
+++ b/drivers/gpu/drm/i915/intel_display.c
@@ -8850,6 +8850,8 @@ void intel_mark_busy(struct drm_device *dev)
 
 	intel_runtime_pm_get(dev_priv);
 	i915_update_gfx_val(dev_priv);
+	if (INTEL_INFO(dev)->gen >= 6)
+		gen6_rps_busy(dev_priv);
 	dev_priv->mm.busy = true;
 }
 
diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h
index 59748497b7ef..857b18df974e 100644
--- a/drivers/gpu/drm/i915/intel_drv.h
+++ b/drivers/gpu/drm/i915/intel_drv.h
@@ -1005,6 +1005,8 @@ void intel_suspend_gt_powersave(struct drm_device *dev);
 void intel_reset_gt_powersave(struct drm_device *dev);
 void ironlake_teardown_rc6(struct drm_device *dev);
 void gen6_update_ring_freq(struct drm_device *dev);
+void gen6_rps_busy(struct drm_i915_private *dev_priv);
+void gen6_rps_reset_ei(struct drm_i915_private *dev_priv);
 void gen6_rps_idle(struct drm_i915_private *dev_priv);
 void gen6_rps_boost(struct drm_i915_private *dev_priv);
 void intel_queue_rps_boost(struct drm_device *dev);
diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c
index 1302e1bc9136..eb6c54a5b112 100644
--- a/drivers/gpu/drm/i915/intel_pm.c
+++ b/drivers/gpu/drm/i915/intel_pm.c
@@ -3281,6 +3281,18 @@ static void vlv_set_rps_idle(struct drm_i915_private *dev_priv)
 		   gen6_rps_pm_mask(dev_priv, dev_priv->rps.cur_freq));
 }
 
+void gen6_rps_busy(struct drm_i915_private *dev_priv)
+{
+	mutex_lock(&dev_priv->rps.hw_lock);
+	if (dev_priv->rps.enabled) {
+		if (dev_priv->pm_rps_events & (GEN6_PM_RP_DOWN_EI_EXPIRED | GEN6_PM_RP_UP_EI_EXPIRED))
+			gen6_rps_reset_ei(dev_priv);
+		I915_WRITE(GEN6_PMINTRMSK,
+			   gen6_rps_pm_mask(dev_priv, dev_priv->rps.cur_freq));
+	}
+	mutex_unlock(&dev_priv->rps.hw_lock);
+}
+
 void gen6_rps_idle(struct drm_i915_private *dev_priv)
 {
 	struct drm_device *dev = dev_priv->dev;
@@ -3292,6 +3304,7 @@ void gen6_rps_idle(struct drm_i915_private *dev_priv)
 		else
 			gen6_set_rps(dev_priv->dev, dev_priv->rps.min_freq_softlimit);
 		dev_priv->rps.last_adj = 0;
+		I915_WRITE(GEN6_PMINTRMSK, 0xffffffff);
 	}
 	mutex_unlock(&dev_priv->rps.hw_lock);
 }
-- 
2.0.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
http://lists.freedesktop.org/mailman/listinfo/intel-gfx





[Index of Archives]     [Linux USB Devel]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]
  Powered by Linux