Skip to contentSkip to conversation

drm/i915: DP link training optimization

PATCH Superseded Mika Kahola 3 participants 10 messages Apr 28, 2015

Subsystems (2)

v3 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

PATCH v3 1/1 drm/i915: DP link training optimization

This is patch series optimizes DP link training by reusing the link parameter settings if DP link has bee previously trained. In case we are not able to train the link by reusing the known values, the link training parameters are set to zero and training is restarted.

The first patch is for eDP only where we reuse the previously trained link training values from cache i.e. voltage swing and pre-emphasis levels.

The second patch is a generalization to cover all cases.

Mika Kahola (1):
drm/i915: DP link training optimization

 drivers/gpu/drm/i915/intel_dp.c  | 29 ++++++++++++++++++++++++++---
 drivers/gpu/drm/i915/intel_drv.h |  1 +
 2 files changed, 27 insertions(+), 3 deletions(-)
Signature · 6 lines
1.9.1

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

PATCH v3 1/2 drm/i915: DP link training optimization

This is a first of series patches that optimize DP link training. The first patch is for eDP only where we reuse the previously trained link training values from cache i.e. voltage swing and pre-emphasis levels.

In case we are not able to train the link by reusing the known values, the link training parameters are set to zero and training is restarted.

V2:
- flag that indicates if DP link is trained and valid renamed from 'link_trained' to 'train_set_valid'
- removed routine 'intel_dp_reuse_link_train'

V3:
- rebased against the latest drm-intel-nightly

---
 drivers/gpu/drm/i915/intel_dp.c  | 29 ++++++++++++++++++++++++++---
 drivers/gpu/drm/i915/intel_drv.h |  1 +
 2 files changed, 27 insertions(+), 3 deletions(-)
drivers/gpu/drm/i915/intel_dp.c +26 -3
index 937ba31..15adafc 100644
@@ -3547,7 +3547,8 @@ static bool
 intel_dp_reset_link_train(struct intel_dp *intel_dp, uint32_t *DP,
 			uint8_t dp_train_pat)
 {
-	memset(intel_dp->train_set, 0, sizeof(intel_dp->train_set));
+	if (!intel_dp->train_set_valid)
+		memset(intel_dp->train_set, 0, sizeof(intel_dp->train_set));
 	intel_dp_set_signal_levels(intel_dp, DP);
 	return intel_dp_set_link_train(intel_dp, DP, dp_train_pat);
 }
@@ -3660,6 +3661,23 @@ intel_dp_start_link_train(struct intel_dp *intel_dp)
 			break;
 		}
 
+		/*
+		 * if we used previously trained voltage and pre-emphasis values
+		 * and we don't get clock recovery, reset link training values
+		 */
+		if (intel_dp->train_set_valid) {
+			DRM_DEBUG_KMS("clock recovery not ok, reset");
+			/* clear the flag as we are not reusing train set */
+			intel_dp->train_set_valid = false;
+			if (!intel_dp_reset_link_train(intel_dp, &DP,
+						       DP_TRAINING_PATTERN_1 |
+						       DP_LINK_SCRAMBLING_DISABLE)) {
+				DRM_ERROR("failed to enable link training\n");
+				return;
+			}
+			continue;
+		}
+
 		/* Check to see if we've tried the max voltage */
 		for (i = 0; i < intel_dp->lane_count; i++)
 			if ((intel_dp->train_set[i] & DP_TRAIN_MAX_SWING_REACHED) == 0)
@@ -3737,6 +3755,7 @@ intel_dp_complete_link_train(struct intel_dp *intel_dp)
 
 		/* Make sure clock is still ok */
 		if (!drm_dp_clock_recovery_ok(link_status, intel_dp->lane_count)) {
+			intel_dp->train_set_valid = false;
 			intel_dp_start_link_train(intel_dp);
 			intel_dp_set_link_train(intel_dp, &DP,
 						training_pattern |
@@ -3752,6 +3771,7 @@ intel_dp_complete_link_train(struct intel_dp *intel_dp)
 
 		/* Try 5 times, then try clock recovery if that fails */
 		if (tries > 5) {
+			intel_dp->train_set_valid = false;
 			intel_dp_start_link_train(intel_dp);
 			intel_dp_set_link_train(intel_dp, &DP,
 						training_pattern |
@@ -3773,9 +3793,10 @@ intel_dp_complete_link_train(struct intel_dp *intel_dp)
 
 	intel_dp->DP = DP;
 
-	if (channel_eq)
+	if (channel_eq) {
+		intel_dp->train_set_valid = is_edp(intel_dp);
 		DRM_DEBUG_KMS("Channel EQ done. DP Training successful\n");
-
+	}
 }
 
 void intel_dp_stop_link_train(struct intel_dp *intel_dp)
@@ -4822,6 +4843,8 @@ intel_dp_hpd_pulse(struct intel_digital_port *intel_dig_port, bool long_hpd)
 	intel_display_power_get(dev_priv, power_domain);
 
 	if (long_hpd) {
+		/* indicate that we need to restart link training */
+		intel_dp->train_set_valid = false;
 
 		if (HAS_PCH_SPLIT(dev)) {
 			if (!ibx_digital_port_connected(dev_priv, intel_dig_port))
drivers/gpu/drm/i915/intel_drv.h +1 -0
index 23a42a4..62bd08b 100644
@@ -739,6 +739,7 @@ struct intel_dp {
 				     bool has_aux_irq,
 				     int send_bytes,
 				     uint32_t aux_clock_divider);
+	bool train_set_valid;
 
 	/* Displayport compliance testing */
 	unsigned long compliance_test_type;
Signature · 6 lines
1.9.1

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

PATCH v3 2/2 drm/i915: DP link training optimization

In reply to Mika Kahola

This patch adds DP link training optimization by reusing the previously trained values.

v2:
- rebase

V3:
- rebase

---
 drivers/gpu/drm/i915/intel_dp.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/i915/intel_dp.c +1 -1
index 15adafc..bb1a8d0 100644
@@ -3794,7 +3794,7 @@ intel_dp_complete_link_train(struct intel_dp *intel_dp)
 	intel_dp->DP = DP;
 
 	if (channel_eq) {
-		intel_dp->train_set_valid = is_edp(intel_dp);
+		intel_dp->train_set_valid = true;
 		DRM_DEBUG_KMS("Channel EQ done. DP Training successful\n");
 	}
 }
Signature · 6 lines
1.9.1

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#3 lore raw
Quoted 19 lines · Mika Kahola

This patch adds DP link training optimization by reusing the previously trained values.

v2:
- rebase

V3:
- rebase

---
  drivers/gpu/drm/i915/intel_dp.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/i915/intel_dp.c +0 -0
index 15adafc..bb1a8d0 100644

drivers/gpu/drm/i915/intel_dp.c intel_dp_complete_link_train(struct intel_dp *intel_dp) · line 3800

Quoted 8 lines
drivers/gpu/drm/i915/intel_dp.c +1 -1
@@ -3794,7 +3794,7 @@ intel_dp_complete_link_train(struct intel_dp *intel_dp)
  	intel_dp->DP = DP;
 
  	if (channel_eq) {
-		intel_dp->train_set_valid = is_edp(intel_dp);
+		intel_dp->train_set_valid = true;
  		DRM_DEBUG_KMS("Channel EQ done. DP Training successful\n");
  	}
  }

any reason why this is split into a separate patch ?
_______________________________________________
Intel-gfx mailing list [email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

#4 lore raw
Quoted 23 lines · Sivakumar Thulasimani
Quoted 21 lines · Mika Kahola

This patch adds DP link training optimization by reusing the previously trained values.

v2:
- rebase

V3:
- rebase

---
  drivers/gpu/drm/i915/intel_dp.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/i915/intel_dp.c +0 -0
index 15adafc..bb1a8d0 100644
@@ -3794,7 +3794,7 @@ intel_dp_complete_link_train(struct intel_dp *intel_dp)
  	intel_dp->DP = DP;
 
Quoted 7 lines

drivers/gpu/drm/i915/intel_dp.c intel_dp_complete_link_train(struct intel_dp *intel_dp) · line 3800

Quoted 6 lines
drivers/gpu/drm/i915/intel_dp.c +1 -1
  	if (channel_eq) {
-		intel_dp->train_set_valid = is_edp(intel_dp);
+		intel_dp->train_set_valid = true;
  		DRM_DEBUG_KMS("Channel EQ done. DP Training successful\n");
  	}
  }

any reason why this is split into a separate patch ?

I discussed this with Jani and we agreed that one option would be to apply this optimization only for the eDP case. This second patch is for the possibility to cover the general DP.

-Mika-

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

#5 lore raw
Quoted 30 lines · Mika Kahola
Quoted 22 lines · Sivakumar Thulasimani
Quoted 21 lines · Mika Kahola

This patch adds DP link training optimization by reusing the previously trained values.

v2:
- rebase

V3:
- rebase

---
   drivers/gpu/drm/i915/intel_dp.c | 2 +-
   1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/i915/intel_dp.c +0 -0
index 15adafc..bb1a8d0 100644
@@ -3794,7 +3794,7 @@ intel_dp_complete_link_train(struct intel_dp *intel_dp)
   	intel_dp->DP = DP;
 
Quoted 7 lines

drivers/gpu/drm/i915/intel_dp.c intel_dp_complete_link_train(struct intel_dp *intel_dp) · line 3800

Quoted 6 lines
drivers/gpu/drm/i915/intel_dp.c +1 -1
   	if (channel_eq) {
-		intel_dp->train_set_valid = is_edp(intel_dp);
+		intel_dp->train_set_valid = true;
   		DRM_DEBUG_KMS("Channel EQ done. DP Training successful\n");
   	}
   }

any reason why this is split into a separate patch ?

Quoted 7 lines

I discussed this with Jani and we agreed that one option would be to apply this optimization only for the eDP case. This second patch is for the possibility to cover the general DP.

-Mika-

might be a nit pick so up to you to consider this or not :). since DP is part of the second patch, better to reset "train_set_valid" for long pulse HPD in the second patch as well. HPD is not enabled for eDP as of now.

Signature · 8 lines
regards,
Sivakumar


_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#6 lore raw
Quoted 37 lines · Sivakumar Thulasimani
Quoted 34 lines · Mika Kahola
Quoted 22 lines · Sivakumar Thulasimani
Quoted 21 lines · Mika Kahola

This patch adds DP link training optimization by reusing the previously trained values.

v2:
- rebase

V3:
- rebase

---
   drivers/gpu/drm/i915/intel_dp.c | 2 +-
   1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/i915/intel_dp.c +0 -0
index 15adafc..bb1a8d0 100644
@@ -3794,7 +3794,7 @@ intel_dp_complete_link_train(struct intel_dp *intel_dp)
   	intel_dp->DP = DP;
 
Quoted 7 lines

drivers/gpu/drm/i915/intel_dp.c intel_dp_complete_link_train(struct intel_dp *intel_dp) · line 3800

Quoted 6 lines
drivers/gpu/drm/i915/intel_dp.c +1 -1
   	if (channel_eq) {
-		intel_dp->train_set_valid = is_edp(intel_dp);
+		intel_dp->train_set_valid = true;
   		DRM_DEBUG_KMS("Channel EQ done. DP Training successful\n");
   	}
   }

any reason why this is split into a separate patch ?

I discussed this with Jani and we agreed that one option would be to apply this optimization only for the eDP case. This second patch is for the possibility to cover the general DP.

Quoted 7 lines
Quoted 3 lines

-Mika-

might be a nit pick so up to you to consider this or not :). since DP is part of the second patch, better to reset "train_set_valid" for long pulse HPD in the second patch as well. HPD is not enabled for eDP as of now.

So, you mean that the second patch could be written as independent of the first patch? In that way you could choose which patch you apply and you wouldn't have to apply the both patches to get the DP case covered.

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

#7 lore raw
Quoted 40 lines · Mika Kahola
Quoted 39 lines · Sivakumar Thulasimani
Quoted 37 lines · Mika Kahola
Quoted 22 lines · Sivakumar Thulasimani
Quoted 21 lines · Mika Kahola

This patch adds DP link training optimization by reusing the previously trained values.

v2:
- rebase

V3:
- rebase

---
    drivers/gpu/drm/i915/intel_dp.c | 2 +-
    1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/i915/intel_dp.c +0 -0
index 15adafc..bb1a8d0 100644
@@ -3794,7 +3794,7 @@ intel_dp_complete_link_train(struct intel_dp *intel_dp)
    	intel_dp->DP = DP;
 
Quoted 7 lines

drivers/gpu/drm/i915/intel_dp.c intel_dp_complete_link_train(struct intel_dp *intel_dp) · line 3800

Quoted 6 lines
drivers/gpu/drm/i915/intel_dp.c +1 -1
    	if (channel_eq) {
-		intel_dp->train_set_valid = is_edp(intel_dp);
+		intel_dp->train_set_valid = true;
    		DRM_DEBUG_KMS("Channel EQ done. DP Training successful\n");
    	}
    }

any reason why this is split into a separate patch ?

I discussed this with Jani and we agreed that one option would be to apply this optimization only for the eDP case. This second patch is for the possibility to cover the general DP.

-Mika-

Quoted 9 lines
Quoted 4 lines

might be a nit pick so up to you to consider this or not :). since DP is part of the second patch, better to reset "train_set_valid" for long pulse HPD in the second patch as well. HPD is not enabled for eDP as of now.

So, you mean that the second patch could be written as independent of the first patch? In that way you could choose which patch you apply and you wouldn't have to apply the both patches to get the DP case covered.

nope, i was referring to the following change alone.

@@ -4822,6 +4843,8 @@ intel_dp_hpd_pulse(struct intel_digital_port *intel_dig_port, bool long_hpd)
intel_display_power_get(dev_priv, power_domain);

if (long_hpd) {
+ /* indicate that we need to restart link training */
+ intel_dp->train_set_valid = false;

to be specific, if 2nd patch is to enable this optimization for DP, the above change will belong here as well. eDP never has HPD enabled so it will never be hit for eDP only scenario.

Signature · 7 lines
regards,
Sivakumar

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#8 lore raw
Quoted 60 lines · Sivakumar Thulasimani
Quoted 40 lines · Mika Kahola
Quoted 39 lines · Sivakumar Thulasimani
Quoted 37 lines · Mika Kahola
Quoted 22 lines · Sivakumar Thulasimani
Quoted 21 lines · Mika Kahola

This patch adds DP link training optimization by reusing the previously trained values.

v2:
- rebase

V3:
- rebase

---
    drivers/gpu/drm/i915/intel_dp.c | 2 +-
    1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/i915/intel_dp.c +0 -0
index 15adafc..bb1a8d0 100644
@@ -3794,7 +3794,7 @@ intel_dp_complete_link_train(struct intel_dp *intel_dp)
    	intel_dp->DP = DP;
 
Quoted 7 lines

drivers/gpu/drm/i915/intel_dp.c intel_dp_complete_link_train(struct intel_dp *intel_dp) · line 3800

Quoted 6 lines
drivers/gpu/drm/i915/intel_dp.c +1 -1
    	if (channel_eq) {
-		intel_dp->train_set_valid = is_edp(intel_dp);
+		intel_dp->train_set_valid = true;
    		DRM_DEBUG_KMS("Channel EQ done. DP Training successful\n");
    	}
    }

any reason why this is split into a separate patch ?

I discussed this with Jani and we agreed that one option would be to apply this optimization only for the eDP case. This second patch is for the possibility to cover the general DP.

-Mika-

Quoted 9 lines
Quoted 4 lines

might be a nit pick so up to you to consider this or not :). since DP is part of the second patch, better to reset "train_set_valid" for long pulse HPD in the second patch as well. HPD is not enabled for eDP as of now.

So, you mean that the second patch could be written as independent of the first patch? In that way you could choose which patch you apply and you wouldn't have to apply the both patches to get the DP case covered.

nope, i was referring to the following change alone.

(quoted diff) +2 -0
@@ -4822,6 +4843,8 @@ intel_dp_hpd_pulse(struct intel_digital_port *intel_dig_port, bool long_hpd)
  	intel_display_power_get(dev_priv, power_domain);
 
  	if (long_hpd) {
+		/* indicate that we need to restart link training */
+		intel_dp->train_set_valid = false;
 
 
Quoted 2 lines

to be specific, if 2nd patch is to enable this optimization for DP, the above change will belong here as well. eDP never has HPD enabled so it will never be hit for eDP only scenario.

Ok, now I'm following you. I rephrase the patch set so the HPD is taken into consideration on this second patch as well.

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

Tested-by on v3 2/2
#9 lore raw
In reply to Mika Kahola

Task id: 6274 -------------------------------------Summary-------------------------------------
Platform Delta drm-intel-nightly Series Applied

PNV                                  276/276              276/276
ILK                                  301/301              301/301
SNB                                  318/318              318/318
IVB                                  345/345              345/345

BYT -1 285/285 284/285
BDW 321/321 321/321
-------------------------------------Detailed-------------------------------------
Platform Test drm-intel-nightly Series Applied
*BYT igt@kms_setmode@invalid-clone-exclusive-crtc PASS(7) DMESG_WARN(1)PASS(1)
(dmesg patch applied)drm:check_crtc_state[i915]]*ERROR*mismatch_in_has_infoframe(expected#,found#)@mismatch in has_infoframe .* found WARNING:at_drivers/gpu/drm/i915/intel_display.c:#check_crtc_state[i915]()@WARNING:.* at .* check_crtc_state+0x
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