Skip to contentSkip to conversation

drm/i915/skl: Retrieve the Rpe value from Pcode

PATCH In review [email protected] 4 participants 11 messages May 5, 2015 – Jun 4, 2015

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/2 drm/i915/skl: Retrieve the Rpe value from Pcode

From: Akash Goel <[email protected]>

Read the efficient frequency (aka RPe) value through the the mailbox command (0x1A) from the pcode, as done on Haswell and Broadwell. The turbo minimum frequency softlimit is not revised as per the efficient frequency value.

Issue: VIZ-5143

---
 drivers/gpu/drm/i915/intel_pm.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)
drivers/gpu/drm/i915/intel_pm.c +6 -2
index a7516ed..421b78d 100644
@@ -4274,16 +4274,20 @@ static void gen6_init_rps_frequencies(struct drm_device *dev)
 	dev_priv->rps.max_freq		= dev_priv->rps.rp0_freq;
 
 	dev_priv->rps.efficient_freq = dev_priv->rps.rp1_freq;
-	if (IS_HASWELL(dev) || IS_BROADWELL(dev)) {
+	if (IS_HASWELL(dev) || IS_BROADWELL(dev) || IS_SKYLAKE(dev)) {
 		ret = sandybridge_pcode_read(dev_priv,
 					HSW_PCODE_DYNAMIC_DUTY_CYCLE_CONTROL,
 					&ddcc_status);
-		if (0 == ret)
+		if (0 == ret) {
 			dev_priv->rps.efficient_freq =
 				clamp_t(u8,
 					((ddcc_status >> 8) & 0xff),
 					dev_priv->rps.min_freq,
 					dev_priv->rps.max_freq);
+
+			dev_priv->rps.efficient_freq *=
+				(IS_SKYLAKE(dev) ? GEN9_FREQ_SCALER : 1);
+		}
 	}
 
 	dev_priv->rps.idle_freq = dev_priv->rps.min_freq;
Signature · 6 lines
1.9.2

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#1 lore raw

PATCH 2/2 drm/i915/skl: Ring frequency table programming changes

From: Akash Goel <[email protected]>

Ring frequency table programming changes for SKL. No need for a floor on ring frequency, as the issue of performance impact with ring running below DDR frequency, is believed to be fixed on SKL

Issue: VIZ-5144

---
 drivers/gpu/drm/i915/intel_pm.c | 26 +++++++++++++++++++++-----
 1 file changed, 21 insertions(+), 5 deletions(-)
drivers/gpu/drm/i915/intel_pm.c +21 -5
index 421b78d..d1bdea7 100644
@@ -4582,6 +4582,7 @@ static void __gen6_update_ring_freq(struct drm_device *dev)
 	int min_freq = 15;
 	unsigned int gpu_freq;
 	unsigned int max_ia_freq, min_ring_freq;
+	unsigned int max_gpu_freq, min_gpu_freq;
 	int scaling_factor = 180;
 	struct cpufreq_policy *policy;
 
@@ -4606,17 +4607,31 @@ static void __gen6_update_ring_freq(struct drm_device *dev)
 	/* convert DDR frequency from units of 266.6MHz to bandwidth */
 	min_ring_freq = mult_frac(min_ring_freq, 8, 3);
 
+	if (IS_SKYLAKE(dev)) {
+		/* Convert GT frequency to 50 HZ units */
+		min_gpu_freq = dev_priv->rps.min_freq / GEN9_FREQ_SCALER;
+		max_gpu_freq = dev_priv->rps.max_freq / GEN9_FREQ_SCALER;
+	} else {
+		min_gpu_freq = dev_priv->rps.min_freq;
+		max_gpu_freq = dev_priv->rps.max_freq;
+	}
+
 	/*
 	 * For each potential GPU frequency, load a ring frequency we'd like
 	 * to use for memory access.  We do this by specifying the IA frequency
 	 * the PCU should use as a reference to determine the ring frequency.
 	 */
-	for (gpu_freq = dev_priv->rps.max_freq; gpu_freq >= dev_priv->rps.min_freq;
-	     gpu_freq--) {
-		int diff = dev_priv->rps.max_freq - gpu_freq;
+	for (gpu_freq = max_gpu_freq; gpu_freq >= min_gpu_freq; gpu_freq--) {
+		int diff = max_gpu_freq - gpu_freq;
 		unsigned int ia_freq = 0, ring_freq = 0;
 
-		if (INTEL_INFO(dev)->gen >= 8) {
+		if (IS_SKYLAKE(dev)) {
+			/*
+			 * ring_freq = 2 * GT. ring_freq is in 100MHz units
+			 * No floor required for ring frequency on SKL.
+			 */
+			ring_freq = gpu_freq;
+		} else if (INTEL_INFO(dev)->gen >= 8) {
 			/* max(2 * GT, DDR). NB: GT is 50MHz units */
 			ring_freq = max(min_ring_freq, gpu_freq);
 		} else if (IS_HASWELL(dev)) {
@@ -5770,7 +5785,8 @@ static void intel_gen6_powersave_work(struct work_struct *work)
 	} else if (INTEL_INFO(dev)->gen >= 9) {
 		gen9_enable_rc6(dev);
 		gen9_enable_rps(dev);
-		__gen6_update_ring_freq(dev);
+		if (IS_SKYLAKE(dev))
+			__gen6_update_ring_freq(dev);
 	} else if (IS_BROADWELL(dev)) {
 		gen8_enable_rps(dev);
 		__gen6_update_ring_freq(dev);
Signature · 6 lines
1.9.2

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Tested-by on v1 2/2
#2 lore raw

Re: [PATCH 2/2] drm/i915/skl: Ring frequency table programming changes

Task id: 6320 -------------------------------------Summary-------------------------------------
Platform Delta drm-intel-nightly Series Applied
PNV 276/276 276/276
ILK -1 302/302 301/302

SNB                                  316/316              316/316
IVB                                  342/342              342/342
BYT                                  286/286              286/286
BDW                                  318/318              318/318

-------------------------------------Detailed-------------------------------------
Platform Test drm-intel-nightly Series Applied
*ILK igt@kms_flip@flip-vs-dpms-interruptible PASS(2) DMESG_WARN(1)PASS(1)
(dmesg patch applied)drm:intel_pch_fifo_underrun_irq_handler[i915]]*ERROR*PCH_transcoder_A_FIFO_underrun@PCH transcoder A FIFO underrun
Note: You need to pay more attention to line start with '*'
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

#3 lore raw

PATCH 3/3 drm/i915/skl: Updated the i915_ring_freq_table debugfs function

In reply to [email protected]

From: Akash Goel <[email protected]>

Updated the i915_ring_freq_table debugfs function to allow read of ring frequency table through Punit interface, for SKL also.

---
 drivers/gpu/drm/i915/i915_debugfs.c | 21 ++++++++++++++++-----
 1 file changed, 16 insertions(+), 5 deletions(-)
drivers/gpu/drm/i915/i915_debugfs.c +16 -5
index 9e5a56c..32527de 100644
@@ -1769,8 +1769,9 @@ static int i915_ring_freq_table(struct seq_file *m, void *unused)
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	int ret = 0;
 	int gpu_freq, ia_freq;
+	unsigned int max_gpu_freq, min_gpu_freq;
 
-	if (!(IS_GEN6(dev) || IS_GEN7(dev))) {
+	if (!(IS_GEN6(dev) || IS_GEN7(dev) || IS_SKYLAKE(dev))) {
 		seq_puts(m, "unsupported on this chipset\n");
 		return 0;
 	}
@@ -1783,17 +1784,27 @@ static int i915_ring_freq_table(struct seq_file *m, void *unused)
 	if (ret)
 		goto out;
 
+	if (IS_SKYLAKE(dev)) {
+		/* Convert GT frequency to 50 HZ units */
+		min_gpu_freq =
+			dev_priv->rps.min_freq_softlimit / GEN9_FREQ_SCALER;
+		max_gpu_freq =
+			dev_priv->rps.max_freq_softlimit / GEN9_FREQ_SCALER;
+	} else {
+		min_gpu_freq = dev_priv->rps.min_freq_softlimit;
+		max_gpu_freq = dev_priv->rps.max_freq_softlimit;
+	}
+
 	seq_puts(m, "GPU freq (MHz)\tEffective CPU freq (MHz)\tEffective Ring freq (MHz)\n");
 
-	for (gpu_freq = dev_priv->rps.min_freq_softlimit;
-	     gpu_freq <= dev_priv->rps.max_freq_softlimit;
-	     gpu_freq++) {
+	for (gpu_freq = min_gpu_freq; gpu_freq <= max_gpu_freq; gpu_freq++) {
 		ia_freq = gpu_freq;
 		sandybridge_pcode_read(dev_priv,
 				       GEN6_PCODE_READ_MIN_FREQ_TABLE,
 				       &ia_freq);
 		seq_printf(m, "%d\t\t%d\t\t\t\t%d\n",
-			   intel_gpu_freq(dev_priv, gpu_freq),
+			   intel_gpu_freq(dev_priv, (gpu_freq *
+				(IS_SKYLAKE(dev) ? GEN9_FREQ_SCALER : 1))),
 			   ((ia_freq >> 0) & 0xff) * 100,
 			   ((ia_freq >> 8) & 0xff) * 100);
 	}
Signature · 6 lines
1.9.2

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#4 lore raw

Re: [PATCH 3/3] drm/i915/skl: Updated the i915_ring_freq_table debugfs function

Quoted 15 lines · On Tue, May 12, 2015 at 12:49 AM, <[email protected]> wrote:

From: Akash Goel <[email protected]>

Updated the i915_ring_freq_table debugfs function to allow read of ring frequency table through Punit interface, for SKL also.

---
 drivers/gpu/drm/i915/i915_debugfs.c | 21 ++++++++++++++++-----
 1 file changed, 16 insertions(+), 5 deletions(-)
drivers/gpu/drm/i915/i915_debugfs.c +0 -0
index 9e5a56c..32527de 100644

drivers/gpu/drm/i915/i915_debugfs.c static int i915_ring_freq_table(struct seq_file *m, void *unused) · line 1774

Quoted 7 lines
drivers/gpu/drm/i915/i915_debugfs.c +2 -1
@@ -1769,8 +1769,9 @@ static int i915_ring_freq_table(struct seq_file *m, void *unused)
        struct drm_i915_private *dev_priv = dev->dev_private;
        int ret = 0;
        int gpu_freq, ia_freq;
+       unsigned int max_gpu_freq, min_gpu_freq;
 
-       if (!(IS_GEN6(dev) || IS_GEN7(dev))) {
+       if (!(IS_GEN6(dev) || IS_GEN7(dev) || IS_SKYLAKE(dev))) {

Why HSW and BDW aren't here and SKL is back?
Again, I'm without doc for this, sorry...

Quoted 22 lines
drivers/gpu/drm/i915/i915_debugfs.c +11 -2
                seq_puts(m, "unsupported on this chipset\n");
                return 0;
        }
@@ -1783,17 +1784,27 @@ static int i915_ring_freq_table(struct seq_file *m, void *unused)
        if (ret)
                goto out;
 
+       if (IS_SKYLAKE(dev)) {
+               /* Convert GT frequency to 50 HZ units */
+               min_gpu_freq =
+                       dev_priv->rps.min_freq_softlimit / GEN9_FREQ_SCALER;
+               max_gpu_freq =
+                       dev_priv->rps.max_freq_softlimit / GEN9_FREQ_SCALER;
+       } else {
+               min_gpu_freq = dev_priv->rps.min_freq_softlimit;
+               max_gpu_freq = dev_priv->rps.max_freq_softlimit;
+       }
+
        seq_puts(m, "GPU freq (MHz)\tEffective CPU freq (MHz)\tEffective Ring freq (MHz)\n");
 
-       for (gpu_freq = dev_priv->rps.min_freq_softlimit;
-            gpu_freq <= dev_priv->rps.max_freq_softlimit;

drivers/gpu/drm/i915/i915_debugfs.c static int i915_ring_freq_table(struct seq_file *m, void *unused) · line 1807

Quoted 10 lines
drivers/gpu/drm/i915/i915_debugfs.c +3 -2
-            gpu_freq++) {
+       for (gpu_freq = min_gpu_freq; gpu_freq <= max_gpu_freq; gpu_freq++) {
                ia_freq = gpu_freq;
                sandybridge_pcode_read(dev_priv,
                                       GEN6_PCODE_READ_MIN_FREQ_TABLE,
                                       &ia_freq);
                seq_printf(m, "%d\t\t%d\t\t\t\t%d\n",
-                          intel_gpu_freq(dev_priv, gpu_freq),
+                          intel_gpu_freq(dev_priv, (gpu_freq *
+                               (IS_SKYLAKE(dev) ? GEN9_FREQ_SCALER : 1))),

here the scaler conversion made sense to me...

Quoted 10 lines
drivers/gpu/drm/i915/i915_debugfs.c +0 -1
                           ((ia_freq >> 0) & 0xff) * 100,
                           ((ia_freq >> 8) & 0xff) * 100);
        }
--

1.9.2

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Signature · 6 lines
Rodrigo Vivi
Blog: http://blog.vivi.eng.br
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
author
#5 lore raw

Re: [PATCH 3/3] drm/i915/skl: Updated the i915_ring_freq_table debugfs function

Quoted 21 lines · Rodrigo Vivi
Quoted 20 lines · On Tue, May 12, 2015 at 12:49 AM, <[email protected]> wrote:

From: Akash Goel <[email protected]>

Updated the i915_ring_freq_table debugfs function to allow read of ring frequency table through Punit interface, for SKL also.

---
 drivers/gpu/drm/i915/i915_debugfs.c | 21 ++++++++++++++++-----
 1 file changed, 16 insertions(+), 5 deletions(-)
drivers/gpu/drm/i915/i915_debugfs.c +1 -0
index 9e5a56c..32527de 100644
@@ -1769,8 +1769,9 @@ static int i915_ring_freq_table(struct seq_file *m, void *unused)
        struct drm_i915_private *dev_priv = dev->dev_private;
        int ret = 0;
        int gpu_freq, ia_freq;
+       unsigned int max_gpu_freq, min_gpu_freq;
 
Quoted 5 lines

drivers/gpu/drm/i915/i915_debugfs.c static int i915_ring_freq_table(struct seq_file *m, void *unused) · line 1774

Quoted 2 lines
drivers/gpu/drm/i915/i915_debugfs.c +1 -1
-       if (!(IS_GEN6(dev) || IS_GEN7(dev))) {
+       if (!(IS_GEN6(dev) || IS_GEN7(dev) || IS_SKYLAKE(dev))) {

Why HSW and BDW aren't here and SKL is back?
Again, I'm without doc for this, sorry...

Sorry not sure that why HSW/BDW aren't used here. For them also this interface should work.

Quoted 21 lines
Quoted 21 lines
drivers/gpu/drm/i915/i915_debugfs.c +11 -0
                seq_puts(m, "unsupported on this chipset\n");
                return 0;
        }
@@ -1783,17 +1784,27 @@ static int i915_ring_freq_table(struct seq_file *m, void *unused)
        if (ret)
                goto out;
 
+       if (IS_SKYLAKE(dev)) {
+               /* Convert GT frequency to 50 HZ units */
+               min_gpu_freq =
+                       dev_priv->rps.min_freq_softlimit / GEN9_FREQ_SCALER;
+               max_gpu_freq =
+                       dev_priv->rps.max_freq_softlimit / GEN9_FREQ_SCALER;
+       } else {
+               min_gpu_freq = dev_priv->rps.min_freq_softlimit;
+               max_gpu_freq = dev_priv->rps.max_freq_softlimit;
+       }
+
        seq_puts(m, "GPU freq (MHz)\tEffective CPU freq (MHz)\tEffective Ring freq (MHz)\n");
 
Quoted 28 lines

drivers/gpu/drm/i915/i915_debugfs.c static int i915_ring_freq_table(struct seq_file *m, void *unused) · line 1807

Quoted 12 lines
drivers/gpu/drm/i915/i915_debugfs.c +3 -4
-       for (gpu_freq = dev_priv->rps.min_freq_softlimit;
-            gpu_freq <= dev_priv->rps.max_freq_softlimit;
-            gpu_freq++) {
+       for (gpu_freq = min_gpu_freq; gpu_freq <= max_gpu_freq; gpu_freq++) {
                ia_freq = gpu_freq;
                sandybridge_pcode_read(dev_priv,
                                       GEN6_PCODE_READ_MIN_FREQ_TABLE,
                                       &ia_freq);
                seq_printf(m, "%d\t\t%d\t\t\t\t%d\n",
-                          intel_gpu_freq(dev_priv, gpu_freq),
+                          intel_gpu_freq(dev_priv, (gpu_freq *
+                               (IS_SKYLAKE(dev) ? GEN9_FREQ_SCALER : 1))),

here the scaler conversion made sense to me...

Quoted 10 lines
drivers/gpu/drm/i915/i915_debugfs.c +0 -1
                           ((ia_freq >> 0) & 0xff) * 100,
                           ((ia_freq >> 8) & 0xff) * 100);
        }
--

1.9.2

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

#6 lore raw

Re: [PATCH 3/3] drm/i915/skl: Updated the i915_ring_freq_table debugfs function

Quoted 22 lines · Akash Goel
Quoted 21 lines · Rodrigo Vivi
Quoted 20 lines · On Tue, May 12, 2015 at 12:49 AM, <[email protected]> wrote:

From: Akash Goel <[email protected]>

Updated the i915_ring_freq_table debugfs function to allow read of ring frequency table through Punit interface, for SKL also.

---
 drivers/gpu/drm/i915/i915_debugfs.c | 21 ++++++++++++++++-----
 1 file changed, 16 insertions(+), 5 deletions(-)
drivers/gpu/drm/i915/i915_debugfs.c +1 -0
index 9e5a56c..32527de 100644
@@ -1769,8 +1769,9 @@ static int i915_ring_freq_table(struct seq_file *m, void *unused)
        struct drm_i915_private *dev_priv = dev->dev_private;
        int ret = 0;
        int gpu_freq, ia_freq;
+       unsigned int max_gpu_freq, min_gpu_freq;
 
Quoted 8 lines
Quoted 5 lines

drivers/gpu/drm/i915/i915_debugfs.c static int i915_ring_freq_table(struct seq_file *m, void *unused) · line 1774

Quoted 2 lines
drivers/gpu/drm/i915/i915_debugfs.c +1 -1
-       if (!(IS_GEN6(dev) || IS_GEN7(dev))) {
+       if (!(IS_GEN6(dev) || IS_GEN7(dev) || IS_SKYLAKE(dev))) {

Why HSW and BDW aren't here and SKL is back?
Again, I'm without doc for this, sorry...

Sorry not sure that why HSW/BDW aren't used here. For them also this interface should work.

HSW is included in the GEN7 check. But so is VLV and that one should be rejected. BDW is genuinely misising. Can you fix it all up while you're poking at it?

Quoted 56 lines
Quoted 22 lines
Quoted 22 lines
drivers/gpu/drm/i915/i915_debugfs.c +11 -0
                seq_puts(m, "unsupported on this chipset\n");
                return 0;
        }
@@ -1783,17 +1784,27 @@ static int i915_ring_freq_table(struct seq_file *m, void *unused)
        if (ret)
                goto out;
 
+       if (IS_SKYLAKE(dev)) {
+               /* Convert GT frequency to 50 HZ units */
+               min_gpu_freq =
+                       dev_priv->rps.min_freq_softlimit / GEN9_FREQ_SCALER;
+               max_gpu_freq =
+                       dev_priv->rps.max_freq_softlimit / GEN9_FREQ_SCALER;
+       } else {
+               min_gpu_freq = dev_priv->rps.min_freq_softlimit;
+               max_gpu_freq = dev_priv->rps.max_freq_softlimit;
+       }
+
        seq_puts(m, "GPU freq (MHz)\tEffective CPU freq (MHz)\tEffective Ring freq (MHz)\n");
 
Quoted 28 lines

drivers/gpu/drm/i915/i915_debugfs.c static int i915_ring_freq_table(struct seq_file *m, void *unused) · line 1807

Quoted 12 lines
drivers/gpu/drm/i915/i915_debugfs.c +3 -4
-       for (gpu_freq = dev_priv->rps.min_freq_softlimit;
-            gpu_freq <= dev_priv->rps.max_freq_softlimit;
-            gpu_freq++) {
+       for (gpu_freq = min_gpu_freq; gpu_freq <= max_gpu_freq; gpu_freq++) {
                ia_freq = gpu_freq;
                sandybridge_pcode_read(dev_priv,
                                       GEN6_PCODE_READ_MIN_FREQ_TABLE,
                                       &ia_freq);
                seq_printf(m, "%d\t\t%d\t\t\t\t%d\n",
-                          intel_gpu_freq(dev_priv, gpu_freq),
+                          intel_gpu_freq(dev_priv, (gpu_freq *
+                               (IS_SKYLAKE(dev) ? GEN9_FREQ_SCALER : 1))),

here the scaler conversion made sense to me...

Quoted 10 lines
drivers/gpu/drm/i915/i915_debugfs.c +0 -1
                           ((ia_freq >> 0) & 0xff) * 100,
                           ((ia_freq >> 8) & 0xff) * 100);
        }
--

1.9.2

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Signature · 6 lines
Ville Syrjälä
Intel OTC
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#7 lore raw

Re: [PATCH 2/2] drm/i915/skl: Ring frequency table programming changes

In reply to [email protected]
Quoted 25 lines · On Tue, May 5, 2015 at 4:30 AM, <[email protected]> wrote:

From: Akash Goel <[email protected]>

Ring frequency table programming changes for SKL. No need for a floor on ring frequency, as the issue of performance impact with ring running below DDR frequency, is believed to be fixed on SKL

Issue: VIZ-5144

---
 drivers/gpu/drm/i915/intel_pm.c | 26 +++++++++++++++++++++-----
 1 file changed, 21 insertions(+), 5 deletions(-)
drivers/gpu/drm/i915/intel_pm.c +1 -0
index 421b78d..d1bdea7 100644
@@ -4582,6 +4582,7 @@ static void __gen6_update_ring_freq(struct drm_device *dev)
        int min_freq = 15;
        unsigned int gpu_freq;
        unsigned int max_ia_freq, min_ring_freq;
+       unsigned int max_gpu_freq, min_gpu_freq;
        int scaling_factor = 180;
        struct cpufreq_policy *policy;
 

drivers/gpu/drm/i915/intel_pm.c static void __gen6_update_ring_freq(struct drm_device *dev) · line 4613

Quoted 7 lines
drivers/gpu/drm/i915/intel_pm.c +4 -0
@@ -4606,17 +4607,31 @@ static void __gen6_update_ring_freq(struct drm_device *dev)
        /* convert DDR frequency from units of 266.6MHz to bandwidth */
        min_ring_freq = mult_frac(min_ring_freq, 8, 3);
 
+       if (IS_SKYLAKE(dev)) {
+               /* Convert GT frequency to 50 HZ units */
+               min_gpu_freq = dev_priv->rps.min_freq / GEN9_FREQ_SCALER;
+               max_gpu_freq = dev_priv->rps.max_freq / GEN9_FREQ_SCALER;

I got even more confused here with the scale differences... Doesn't look so consistent..

Quoted 14 lines
drivers/gpu/drm/i915/intel_pm.c +6 -3
+       } else {
+               min_gpu_freq = dev_priv->rps.min_freq;
+               max_gpu_freq = dev_priv->rps.max_freq;
+       }
+
        /*
         * For each potential GPU frequency, load a ring frequency we'd like
         * to use for memory access.  We do this by specifying the IA frequency
         * the PCU should use as a reference to determine the ring frequency.
         */
-       for (gpu_freq = dev_priv->rps.max_freq; gpu_freq >= dev_priv->rps.min_freq;
-            gpu_freq--) {
-               int diff = dev_priv->rps.max_freq - gpu_freq;
+       for (gpu_freq = max_gpu_freq; gpu_freq >= min_gpu_freq; gpu_freq--) {

drivers/gpu/drm/i915/intel_pm.c static void __gen6_update_ring_freq(struct drm_device *dev) · line 4633

Quoted 10 lines
drivers/gpu/drm/i915/intel_pm.c +7 -1
+               int diff = max_gpu_freq - gpu_freq;
                unsigned int ia_freq = 0, ring_freq = 0;
 
-               if (INTEL_INFO(dev)->gen >= 8) {
+               if (IS_SKYLAKE(dev)) {
+                       /*
+                        * ring_freq = 2 * GT. ring_freq is in 100MHz units
+                        * No floor required for ring frequency on SKL.
+                        */
+                       ring_freq = gpu_freq;

Aren't we using the 16.6 now? Shouldn't it be converted anyway?

drivers/gpu/drm/i915/intel_pm.c static void intel_gen6_powersave_work(struct work_struct *work) · line 5788

Quoted 10 lines
drivers/gpu/drm/i915/intel_pm.c +2 -1
+               } else if (INTEL_INFO(dev)->gen >= 8) {
                        /* max(2 * GT, DDR). NB: GT is 50MHz units */
                        ring_freq = max(min_ring_freq, gpu_freq);
                } else if (IS_HASWELL(dev)) {
@@ -5770,7 +5785,8 @@ static void intel_gen6_powersave_work(struct work_struct *work)
        } else if (INTEL_INFO(dev)->gen >= 9) {
                gen9_enable_rc6(dev);
                gen9_enable_rps(dev);
-               __gen6_update_ring_freq(dev);
+               if (IS_SKYLAKE(dev))

why this if here?

If it is a bxt restriction shouldn't it be in a separated patch?

Quoted 11 lines
drivers/gpu/drm/i915/intel_pm.c +1 -1
+                       __gen6_update_ring_freq(dev);
        } else if (IS_BROADWELL(dev)) {
                gen8_enable_rps(dev);
                __gen6_update_ring_freq(dev);
--

1.9.2

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Signature · 6 lines
Rodrigo Vivi
Blog: http://blog.vivi.eng.br
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
author
#8 lore raw

Re: [PATCH 2/2] drm/i915/skl: Ring frequency table programming changes

Quoted 29 lines · Rodrigo Vivi
Quoted 28 lines · On Tue, May 5, 2015 at 4:30 AM, <[email protected]> wrote:

From: Akash Goel <[email protected]>

Ring frequency table programming changes for SKL. No need for a floor on ring frequency, as the issue of performance impact with ring running below DDR frequency, is believed to be fixed on SKL

Issue: VIZ-5144

---
 drivers/gpu/drm/i915/intel_pm.c | 26 +++++++++++++++++++++-----
 1 file changed, 21 insertions(+), 5 deletions(-)
drivers/gpu/drm/i915/intel_pm.c +1 -0
index 421b78d..d1bdea7 100644
@@ -4582,6 +4582,7 @@ static void __gen6_update_ring_freq(struct drm_device *dev)
        int min_freq = 15;
        unsigned int gpu_freq;
        unsigned int max_ia_freq, min_ring_freq;
+       unsigned int max_gpu_freq, min_gpu_freq;
        int scaling_factor = 180;
        struct cpufreq_policy *policy;
 
@@ -4606,17 +4607,31 @@ static void __gen6_update_ring_freq(struct drm_device *dev)
        /* convert DDR frequency from units of 266.6MHz to bandwidth */
        min_ring_freq = mult_frac(min_ring_freq, 8, 3);
 
Quoted 7 lines

drivers/gpu/drm/i915/intel_pm.c static void __gen6_update_ring_freq(struct drm_device *dev) · line 4613

Quoted 4 lines
drivers/gpu/drm/i915/intel_pm.c +4 -0
+       if (IS_SKYLAKE(dev)) {
+               /* Convert GT frequency to 50 HZ units */
+               min_gpu_freq = dev_priv->rps.min_freq / GEN9_FREQ_SCALER;
+               max_gpu_freq = dev_priv->rps.max_freq / GEN9_FREQ_SCALER;

I got even more confused here with the scale differences... Doesn't look so consistent..

For Ring frequency table programming, as an input GT frequency has to be specified in units of 50 MHZ only. For SKL natural hardware unit is 16.667 MHZ and Driver is internally storing in the same unit only, hence an extra conversion required here.

Quoted 40 lines
Quoted 15 lines
drivers/gpu/drm/i915/intel_pm.c +6 -3
+       } else {
+               min_gpu_freq = dev_priv->rps.min_freq;
+               max_gpu_freq = dev_priv->rps.max_freq;
+       }
+
        /*
         * For each potential GPU frequency, load a ring frequency we'd like
         * to use for memory access.  We do this by specifying the IA frequency
         * the PCU should use as a reference to determine the ring frequency.
         */
-       for (gpu_freq = dev_priv->rps.max_freq; gpu_freq >= dev_priv->rps.min_freq;
-            gpu_freq--) {
-               int diff = dev_priv->rps.max_freq - gpu_freq;
+       for (gpu_freq = max_gpu_freq; gpu_freq >= min_gpu_freq; gpu_freq--) {

drivers/gpu/drm/i915/intel_pm.c static void __gen6_update_ring_freq(struct drm_device *dev) · line 4633

Quoted 10 lines
drivers/gpu/drm/i915/intel_pm.c +7 -1
+               int diff = max_gpu_freq - gpu_freq;
                unsigned int ia_freq = 0, ring_freq = 0;
 
-               if (INTEL_INFO(dev)->gen >= 8) {
+               if (IS_SKYLAKE(dev)) {
+                       /*
+                        * ring_freq = 2 * GT. ring_freq is in 100MHz units
+                        * No floor required for ring frequency on SKL.
+                        */
+                       ring_freq = gpu_freq;

Aren't we using the 16.6 now? Shouldn't it be converted anyway?

drivers/gpu/drm/i915/intel_pm.c static void intel_gen6_powersave_work(struct work_struct *work) · line 5788

Quoted 10 lines
drivers/gpu/drm/i915/intel_pm.c +2 -1
+               } else if (INTEL_INFO(dev)->gen >= 8) {
                        /* max(2 * GT, DDR). NB: GT is 50MHz units */
                        ring_freq = max(min_ring_freq, gpu_freq);
                } else if (IS_HASWELL(dev)) {
@@ -5770,7 +5785,8 @@ static void intel_gen6_powersave_work(struct work_struct *work)
        } else if (INTEL_INFO(dev)->gen >= 9) {
                gen9_enable_rc6(dev);
                gen9_enable_rps(dev);
-               __gen6_update_ring_freq(dev);
+               if (IS_SKYLAKE(dev))

why this if here?

Quoted 1 line

If it is a bxt restriction shouldn't it be in a separated patch?

Yes for BXT, its not supported. Would move it to separate patch.

Quoted 8 lines
Quoted 8 lines
drivers/gpu/drm/i915/intel_pm.c +1 -1
+                       __gen6_update_ring_freq(dev);
        } else if (IS_BROADWELL(dev)) {
                gen8_enable_rps(dev);
                __gen6_update_ring_freq(dev);
--

1.9.2

Quoted 7 lines
Quoted 4 lines

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

#9 lore raw
In reply to [email protected]
Quoted 18 lines · On Tue, May 5, 2015 at 4:30 AM, <[email protected]> wrote:

From: Akash Goel <[email protected]>

Read the efficient frequency (aka RPe) value through the the mailbox command (0x1A) from the pcode, as done on Haswell and Broadwell. The turbo minimum frequency softlimit is not revised as per the efficient frequency value.

Issue: VIZ-5143

---
 drivers/gpu/drm/i915/intel_pm.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)
drivers/gpu/drm/i915/intel_pm.c +0 -0
index a7516ed..421b78d 100644

drivers/gpu/drm/i915/intel_pm.c static void gen6_init_rps_frequencies(struct drm_device *dev) · line 4277

Quoted 5 lines
drivers/gpu/drm/i915/intel_pm.c +1 -1
@@ -4274,16 +4274,20 @@ static void gen6_init_rps_frequencies(struct drm_device *dev)
        dev_priv->rps.max_freq          = dev_priv->rps.rp0_freq;
 
        dev_priv->rps.efficient_freq = dev_priv->rps.rp1_freq;
-       if (IS_HASWELL(dev) || IS_BROADWELL(dev)) {
+       if (IS_HASWELL(dev) || IS_BROADWELL(dev) || IS_SKYLAKE(dev)) {

Could you please give me a pointer in the spec to this? Couldn't find on new one.

drivers/gpu/drm/i915/intel_pm.c static void gen6_init_rps_frequencies(struct drm_device *dev) · line 4289

Quoted 13 lines
drivers/gpu/drm/i915/intel_pm.c +4 -1
                ret = sandybridge_pcode_read(dev_priv,
                                        HSW_PCODE_DYNAMIC_DUTY_CYCLE_CONTROL,
                                        &ddcc_status);
-               if (0 == ret)
+               if (0 == ret) {
                        dev_priv->rps.efficient_freq =
                                clamp_t(u8,
                                        ((ddcc_status >> 8) & 0xff),
                                        dev_priv->rps.min_freq,
                                        dev_priv->rps.max_freq);
+
+                       dev_priv->rps.efficient_freq *=
+                               (IS_SKYLAKE(dev) ? GEN9_FREQ_SCALER : 1);

I thought you were managing everything on 50MHz and converting to 16.6 only when writing. So this here seems to me that will compare to values in different bases..

Quoted 11 lines
drivers/gpu/drm/i915/intel_pm.c +1 -1
+               }
        }
 
        dev_priv->rps.idle_freq = dev_priv->rps.min_freq;
--

1.9.2

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Signature · 6 lines
Rodrigo Vivi
Blog: http://blog.vivi.eng.br
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
author
#10 lore raw
Quoted 21 lines · Rodrigo Vivi
Quoted 20 lines · On Tue, May 5, 2015 at 4:30 AM, <[email protected]> wrote:

From: Akash Goel <[email protected]>

Read the efficient frequency (aka RPe) value through the the mailbox command (0x1A) from the pcode, as done on Haswell and Broadwell. The turbo minimum frequency softlimit is not revised as per the efficient frequency value.

Issue: VIZ-5143

---
 drivers/gpu/drm/i915/intel_pm.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)
drivers/gpu/drm/i915/intel_pm.c +0 -0
index a7516ed..421b78d 100644
@@ -4274,16 +4274,20 @@ static void gen6_init_rps_frequencies(struct drm_device *dev)
        dev_priv->rps.max_freq          = dev_priv->rps.rp0_freq;
 
Quoted 6 lines

drivers/gpu/drm/i915/intel_pm.c static void gen6_init_rps_frequencies(struct drm_device *dev) · line 4277

Quoted 3 lines
drivers/gpu/drm/i915/intel_pm.c +1 -1
        dev_priv->rps.efficient_freq = dev_priv->rps.rp1_freq;
-       if (IS_HASWELL(dev) || IS_BROADWELL(dev)) {
+       if (IS_HASWELL(dev) || IS_BROADWELL(dev) || IS_SKYLAKE(dev)) {

Could you please give me a pointer in the spec to this? Couldn't find on new one.

Actually got this piece of info from fellow VPG engineers. Sorry I myself don't have the access to the Punit spec.

Quoted 14 lines

drivers/gpu/drm/i915/intel_pm.c static void gen6_init_rps_frequencies(struct drm_device *dev) · line 4289

Quoted 13 lines
drivers/gpu/drm/i915/intel_pm.c +4 -1
                ret = sandybridge_pcode_read(dev_priv,
                                        HSW_PCODE_DYNAMIC_DUTY_CYCLE_CONTROL,
                                        &ddcc_status);
-               if (0 == ret)
+               if (0 == ret) {
                        dev_priv->rps.efficient_freq =
                                clamp_t(u8,
                                        ((ddcc_status >> 8) & 0xff),
                                        dev_priv->rps.min_freq,
                                        dev_priv->rps.max_freq);
+
+                       dev_priv->rps.efficient_freq *=
+                               (IS_SKYLAKE(dev) ? GEN9_FREQ_SCALER : 1);
Quoted 3 lines

I thought you were managing everything on 50MHz and converting to 16.6 only when writing. So this here seems to me that will compare to values in different bases..

The frequency values are being internally maintained in their natural hardware units only (and functions intel_gpu_freq & intel_freq_opcode help in the conversion).
For SKL, although unit is 16.667 MHz, but there are few exceptions like RP_STATE register, which still reports frequency in 50 MHZ units, so an extra conversion is required there to internally store the values in 16.667 MHZ. The Ring frequency table programming is another instance where update has to be done as per 50 MHZ unit.

Best regards
Akash

Quoted 4 lines
Quoted 4 lines
drivers/gpu/drm/i915/intel_pm.c +1 -0
+               }
        }
 
Quoted 11 lines
Quoted 8 lines
drivers/gpu/drm/i915/intel_pm.c +0 -1
        dev_priv->rps.idle_freq = dev_priv->rps.min_freq;
--

1.9.2

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx