Skip to contentSkip to conversation

drm/i915: Use ilk_init_lp_watermarks() on BDW

PATCH In review [email protected] 4 participants 9 messages May 19, 2015 – May 22, 2015

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

PATCH 1/3 drm/i915: Use ilk_init_lp_watermarks() on BDW

From: Ville Syrjälä <[email protected]>

We're not using ilk_init_lp_watermarks() on BDW for some reason. Probably due to the BDW patches and the relevant WM patches landing roughlly at the same time. Fix it up.

---
 drivers/gpu/drm/i915/intel_pm.c | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)
drivers/gpu/drm/i915/intel_pm.c +1 -3
index ce1d079..206bd41 100644
@@ -6166,9 +6166,7 @@ static void broadwell_init_clock_gating(struct drm_device *dev)
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	enum pipe pipe;
 
-	I915_WRITE(WM3_LP_ILK, 0);
-	I915_WRITE(WM2_LP_ILK, 0);
-	I915_WRITE(WM1_LP_ILK, 0);
+	ilk_init_lp_watermarks(dev);
 
 	/* WaSwitchSolVfFArbitrationPriority:bdw */
 	I915_WRITE(GAM_ECOCHK, I915_READ(GAM_ECOCHK) | HSW_ECOCHK_ARB_PRIO_SOL);
Signature · 6 lines
2.0.5

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

PATCH 2/3 drm/i915: Move WaProgramL3SqcReg1Default:bdw to init_clock_gating()

From: Ville Syrjälä <[email protected]>

GEN8_L3SQCREG1 isn't saved in the context (verified by going through a context dump), and so we shouldn't be using the ring w/a code to initialize it. Also Bspec explicitly talks about MMIO and writing it with the CPU.

Additionally there's another w/a WaTempDisableDOPClkGating:bdw which tells us to disable DOP clock gating around the GEN8_L3SQCREG1 write to make sure everyone notices the change. So let's do that as well.

---
 drivers/gpu/drm/i915/intel_pm.c         | 10 ++++++++++
 drivers/gpu/drm/i915/intel_ringbuffer.c |  3 ---
 2 files changed, 10 insertions(+), 3 deletions(-)
drivers/gpu/drm/i915/intel_pm.c +10 -0
index 206bd41..5ec56b6 100644
@@ -6165,6 +6165,7 @@ static void broadwell_init_clock_gating(struct drm_device *dev)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	enum pipe pipe;
+	uint32_t misccpctl;
 
 	ilk_init_lp_watermarks(dev);
 
@@ -6195,6 +6196,15 @@ static void broadwell_init_clock_gating(struct drm_device *dev)
 	I915_WRITE(GEN8_UCGCTL6, I915_READ(GEN8_UCGCTL6) |
 		   GEN8_SDEUNIT_CLOCK_GATE_DISABLE);
 
+	/*
+	 * WaProgramL3SqcReg1Default:bdw
+	 * WaTempDisableDOPClkGating:bdw
+	 */
+	misccpctl = I915_READ(GEN7_MISCCPCTL);
+	I915_WRITE(GEN7_MISCCPCTL, misccpctl & ~GEN7_DOP_CLOCK_GATE_ENABLE);
+	I915_WRITE(GEN8_L3SQCREG1, BDW_WA_L3SQCREG1_DEFAULT);
+	I915_WRITE(GEN7_MISCCPCTL, misccpctl);
+
 	lpt_init_clock_gating(dev);
 }
 
drivers/gpu/drm/i915/intel_ringbuffer.c +0 -3
index 9b96ed7..50cdd67 100644
@@ -853,9 +853,6 @@ static int bdw_init_workarounds(struct intel_engine_cs *ring)
 			    GEN6_WIZ_HASHING_MASK,
 			    GEN6_WIZ_HASHING_16x4);
 
-	/* WaProgramL3SqcReg1Default:bdw */
-	WA_WRITE(GEN8_L3SQCREG1, BDW_WA_L3SQCREG1_DEFAULT);
-
 	return 0;
 }
 
Signature · 6 lines
2.0.5

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Reviewed-by on v1 2/3
#2 lore raw

Re: [PATCH 2/3] drm/i915: Move WaProgramL3SqcReg1Default:bdw to init_clock_gating()

Quoted 52 lines · On 05/19/2015 10:32 AM, [email protected] wrote:

From: Ville Syrjälä <[email protected]>

GEN8_L3SQCREG1 isn't saved in the context (verified by going through a context dump), and so we shouldn't be using the ring w/a code to initialize it. Also Bspec explicitly talks about MMIO and writing it with the CPU.

Additionally there's another w/a WaTempDisableDOPClkGating:bdw which tells us to disable DOP clock gating around the GEN8_L3SQCREG1 write to make sure everyone notices the change. So let's do that as well.

---
 drivers/gpu/drm/i915/intel_pm.c         | 10 ++++++++++
 drivers/gpu/drm/i915/intel_ringbuffer.c |  3 ---
 2 files changed, 10 insertions(+), 3 deletions(-)
drivers/gpu/drm/i915/intel_pm.c +10 -0
index 206bd41..5ec56b6 100644
@@ -6165,6 +6165,7 @@ static void broadwell_init_clock_gating(struct drm_device *dev)
 {
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	enum pipe pipe;
+	uint32_t misccpctl;
 
 	ilk_init_lp_watermarks(dev);
 
@@ -6195,6 +6196,15 @@ static void broadwell_init_clock_gating(struct drm_device *dev)
 	I915_WRITE(GEN8_UCGCTL6, I915_READ(GEN8_UCGCTL6) |
 		   GEN8_SDEUNIT_CLOCK_GATE_DISABLE);
 
+	/*
+	 * WaProgramL3SqcReg1Default:bdw
+	 * WaTempDisableDOPClkGating:bdw
+	 */
+	misccpctl = I915_READ(GEN7_MISCCPCTL);
+	I915_WRITE(GEN7_MISCCPCTL, misccpctl & ~GEN7_DOP_CLOCK_GATE_ENABLE);
+	I915_WRITE(GEN8_L3SQCREG1, BDW_WA_L3SQCREG1_DEFAULT);
+	I915_WRITE(GEN7_MISCCPCTL, misccpctl);
+
 	lpt_init_clock_gating(dev);
 }
 
drivers/gpu/drm/i915/intel_ringbuffer.c +0 -0
index 9b96ed7..50cdd67 100644

drivers/gpu/drm/i915/intel_ringbuffer.c static int bdw_init_workarounds(struct intel_engine_cs *ring) · line 858

Quoted 9 lines
drivers/gpu/drm/i915/intel_ringbuffer.c +0 -3
@@ -853,9 +853,6 @@ static int bdw_init_workarounds(struct intel_engine_cs *ring)
 			    GEN6_WIZ_HASHING_MASK,
 			    GEN6_WIZ_HASHING_16x4);
 
-	/* WaProgramL3SqcReg1Default:bdw */
-	WA_WRITE(GEN8_L3SQCREG1, BDW_WA_L3SQCREG1_DEFAULT);
-
 	return 0;
 }
 

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

#3 lore raw

PATCH 3/3 drm/i915: Enable GTT caching on gen8

In reply to [email protected]

From: Ville Syrjälä <[email protected]>

GTT caching was disabled by default on gen8 due to not working with big pages. Some information suggests that it got fixed, but still GTT caching has been left disabled by default. Or could be it just meant that the default was changed to off, and hence the problem got solved.

Enable GTT caching in the hopes of some performance increase. Whether or not the big pages issue has been fixed is irrelevant at this stage since we don't use big pages.

This gives me a 1-2% improvement in xonotic on my BSW. Haven't tried BDW, but supposedly it has larger TLBs so might not benefit as much. On HSW GTT caching is enabled by default.

---
 drivers/gpu/drm/i915/i915_reg.h |  2 ++
 drivers/gpu/drm/i915/intel_pm.c | 13 +++++++++++++
 2 files changed, 15 insertions(+)
drivers/gpu/drm/i915/i915_reg.h +2 -0
index 84af255..90640d5 100644
@@ -1461,6 +1461,8 @@ enum skl_disp_power_wells {
 #define RING_HWS_PGA(base)	((base)+0x80)
 #define RING_HWS_PGA_GEN6(base)	((base)+0x2080)
 
+#define HSW_GTT_CACHE_EN	0x4024
+#define   GTT_CACHE_EN_ALL	0xF0007FFF
 #define GEN7_WR_WATERMARK	0x4028
 #define GEN7_GFX_PRIO_CTRL	0x402C
 #define ARB_MODE		0x4030
drivers/gpu/drm/i915/intel_pm.c +13 -0
index 5ec56b6..58517a50 100644
@@ -6205,6 +6205,13 @@ static void broadwell_init_clock_gating(struct drm_device *dev)
 	I915_WRITE(GEN8_L3SQCREG1, BDW_WA_L3SQCREG1_DEFAULT);
 	I915_WRITE(GEN7_MISCCPCTL, misccpctl);
 
+	/*
+	 * WaGttCachingOffByDefault:bdw
+	 * GTT cache may not work with big pages, so if those
+	 * are ever enabled GTT cache may need to be disabled.
+	 */
+	I915_WRITE(HSW_GTT_CACHE_EN, GTT_CACHE_EN_ALL);
+
 	lpt_init_clock_gating(dev);
 }
 
@@ -6480,6 +6487,12 @@ static void cherryview_init_clock_gating(struct drm_device *dev)
 	/* WaDisableSDEUnitClockGating:chv */
 	I915_WRITE(GEN8_UCGCTL6, I915_READ(GEN8_UCGCTL6) |
 		   GEN8_SDEUNIT_CLOCK_GATE_DISABLE);
+
+	/*
+	 * GTT cache may not work with big pages, so if those
+	 * are ever enabled GTT cache may need to be disabled.
+	 */
+	I915_WRITE(HSW_GTT_CACHE_EN, GTT_CACHE_EN_ALL);
 }
 
 static void g4x_init_clock_gating(struct drm_device *dev)
Signature · 6 lines
2.0.5

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Tested-by on v1 3/3
#4 lore raw

Re: [PATCH 3/3] drm/i915: Enable GTT caching on gen8

Task id: 6434 -------------------------------------Summary-------------------------------------
Platform Delta drm-intel-nightly Series Applied

PNV                                  234/234              234/234
ILK                                  262/262              262/262
SNB                 -1              282/282              281/282
IVB                                  300/300              300/300
BYT                                  254/254              254/254
BDW                                  275/275              275/275

-------------------------------------Detailed-------------------------------------
Platform Test drm-intel-nightly Series Applied
SNB igt@pm_rpm@dpms-mode-unset-non-lpsp DMESG_WARN(11)PASS(1) DMESG_WARN(1)
(dmesg patch applied)WARNING:at_drivers/gpu/drm/i915/intel_uncore.c:#assert_device_not_suspended[i915]()@WARNING:.* at .* assert_device_not_suspended+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

Reviewed-by on v1 3/3
#5 lore raw

Re: [PATCH 3/3] drm/i915: Enable GTT caching on gen8

In reply to [email protected]
Quoted 57 lines · On 05/19/2015 10:32 AM, [email protected] wrote:

From: Ville Syrjälä <[email protected]>

GTT caching was disabled by default on gen8 due to not working with big pages. Some information suggests that it got fixed, but still GTT caching has been left disabled by default. Or could be it just meant that the default was changed to off, and hence the problem got solved.

Enable GTT caching in the hopes of some performance increase. Whether or not the big pages issue has been fixed is irrelevant at this stage since we don't use big pages.

This gives me a 1-2% improvement in xonotic on my BSW. Haven't tried BDW, but supposedly it has larger TLBs so might not benefit as much. On HSW GTT caching is enabled by default.

---
 drivers/gpu/drm/i915/i915_reg.h |  2 ++
 drivers/gpu/drm/i915/intel_pm.c | 13 +++++++++++++
 2 files changed, 15 insertions(+)
drivers/gpu/drm/i915/i915_reg.h +2 -0
index 84af255..90640d5 100644
@@ -1461,6 +1461,8 @@ enum skl_disp_power_wells {
 #define RING_HWS_PGA(base)	((base)+0x80)
 #define RING_HWS_PGA_GEN6(base)	((base)+0x2080)
 
+#define HSW_GTT_CACHE_EN	0x4024
+#define   GTT_CACHE_EN_ALL	0xF0007FFF
 #define GEN7_WR_WATERMARK	0x4028
 #define GEN7_GFX_PRIO_CTRL	0x402C
 #define ARB_MODE		0x4030
drivers/gpu/drm/i915/intel_pm.c +7 -0
index 5ec56b6..58517a50 100644
@@ -6205,6 +6205,13 @@ static void broadwell_init_clock_gating(struct drm_device *dev)
 	I915_WRITE(GEN8_L3SQCREG1, BDW_WA_L3SQCREG1_DEFAULT);
 	I915_WRITE(GEN7_MISCCPCTL, misccpctl);
 
+	/*
+	 * WaGttCachingOffByDefault:bdw
+	 * GTT cache may not work with big pages, so if those
+	 * are ever enabled GTT cache may need to be disabled.
+	 */
+	I915_WRITE(HSW_GTT_CACHE_EN, GTT_CACHE_EN_ALL);
+
 	lpt_init_clock_gating(dev);
 }
 
@@ -6480,6 +6487,12 @@ static void cherryview_init_clock_gating(struct drm_device *dev)
 	/* WaDisableSDEUnitClockGating:chv */
 	I915_WRITE(GEN8_UCGCTL6, I915_READ(GEN8_UCGCTL6) |

drivers/gpu/drm/i915/intel_pm.c static void cherryview_init_clock_gating(struct drm_device *dev) · line 6498

Quoted 10 lines
drivers/gpu/drm/i915/intel_pm.c +6 -0
 		   GEN8_SDEUNIT_CLOCK_GATE_DISABLE);
+
+	/*
+	 * GTT cache may not work with big pages, so if those
+	 * are ever enabled GTT cache may need to be disabled.
+	 */
+	I915_WRITE(HSW_GTT_CACHE_EN, GTT_CACHE_EN_ALL);
 }
 
 static void g4x_init_clock_gating(struct drm_device *dev)

Looks ok to me; I guess testing will be the real review here.

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

#6 lore raw

Re: [PATCH 3/3] drm/i915: Enable GTT caching on gen8

Quoted 66 lines · Jesse Barnes
Quoted 65 lines · On 05/19/2015 10:32 AM, [email protected] wrote:

From: Ville Syrjälä <[email protected]>

GTT caching was disabled by default on gen8 due to not working with big pages. Some information suggests that it got fixed, but still GTT caching has been left disabled by default. Or could be it just meant that the default was changed to off, and hence the problem got solved.

Enable GTT caching in the hopes of some performance increase. Whether or not the big pages issue has been fixed is irrelevant at this stage since we don't use big pages.

This gives me a 1-2% improvement in xonotic on my BSW. Haven't tried BDW, but supposedly it has larger TLBs so might not benefit as much. On HSW GTT caching is enabled by default.

---
 drivers/gpu/drm/i915/i915_reg.h |  2 ++
 drivers/gpu/drm/i915/intel_pm.c | 13 +++++++++++++
 2 files changed, 15 insertions(+)
drivers/gpu/drm/i915/i915_reg.h +2 -0
index 84af255..90640d5 100644
@@ -1461,6 +1461,8 @@ enum skl_disp_power_wells {
 #define RING_HWS_PGA(base)	((base)+0x80)
 #define RING_HWS_PGA_GEN6(base)	((base)+0x2080)
 
+#define HSW_GTT_CACHE_EN	0x4024
+#define   GTT_CACHE_EN_ALL	0xF0007FFF
 #define GEN7_WR_WATERMARK	0x4028
 #define GEN7_GFX_PRIO_CTRL	0x402C
 #define ARB_MODE		0x4030
drivers/gpu/drm/i915/intel_pm.c +13 -0
index 5ec56b6..58517a50 100644
@@ -6205,6 +6205,13 @@ static void broadwell_init_clock_gating(struct drm_device *dev)
 	I915_WRITE(GEN8_L3SQCREG1, BDW_WA_L3SQCREG1_DEFAULT);
 	I915_WRITE(GEN7_MISCCPCTL, misccpctl);
 
+	/*
+	 * WaGttCachingOffByDefault:bdw
+	 * GTT cache may not work with big pages, so if those
+	 * are ever enabled GTT cache may need to be disabled.
+	 */
+	I915_WRITE(HSW_GTT_CACHE_EN, GTT_CACHE_EN_ALL);
+
 	lpt_init_clock_gating(dev);
 }
 
@@ -6480,6 +6487,12 @@ static void cherryview_init_clock_gating(struct drm_device *dev)
 	/* WaDisableSDEUnitClockGating:chv */
 	I915_WRITE(GEN8_UCGCTL6, I915_READ(GEN8_UCGCTL6) |
 		   GEN8_SDEUNIT_CLOCK_GATE_DISABLE);
+
+	/*
+	 * GTT cache may not work with big pages, so if those
+	 * are ever enabled GTT cache may need to be disabled.
+	 */
+	I915_WRITE(HSW_GTT_CACHE_EN, GTT_CACHE_EN_ALL);
 }
 
Quoted 4 lines

drivers/gpu/drm/i915/intel_pm.c static void cherryview_init_clock_gating(struct drm_device *dev) · line 6498

Quoted 2 lines
drivers/gpu/drm/i915/intel_pm.c +0 -0
 static void g4x_init_clock_gating(struct drm_device *dev)

Looks ok to me; I guess testing will be the real review here.

Just aside: If you think the real test for a patch is the real world imo an important part of the review work is to make sure we do have that testing coverage. We have a few igts that specifically exercise gtt tlb issues (from previous generations), so I think we're covered here.

Quoted 1 line

Merged all three, thanks for patches&review.
-Daniel

Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#7 lore raw

Re: [PATCH 3/3] drm/i915: Enable GTT caching on gen8

Quoted 72 lines · Daniel Vetter
Quoted 66 lines · Jesse Barnes
Quoted 65 lines · On 05/19/2015 10:32 AM, [email protected] wrote:

From: Ville Syrjälä <[email protected]>

GTT caching was disabled by default on gen8 due to not working with big pages. Some information suggests that it got fixed, but still GTT caching has been left disabled by default. Or could be it just meant that the default was changed to off, and hence the problem got solved.

Enable GTT caching in the hopes of some performance increase. Whether or not the big pages issue has been fixed is irrelevant at this stage since we don't use big pages.

This gives me a 1-2% improvement in xonotic on my BSW. Haven't tried BDW, but supposedly it has larger TLBs so might not benefit as much. On HSW GTT caching is enabled by default.

---
 drivers/gpu/drm/i915/i915_reg.h |  2 ++
 drivers/gpu/drm/i915/intel_pm.c | 13 +++++++++++++
 2 files changed, 15 insertions(+)
drivers/gpu/drm/i915/i915_reg.h +2 -0
index 84af255..90640d5 100644
@@ -1461,6 +1461,8 @@ enum skl_disp_power_wells {
 #define RING_HWS_PGA(base)	((base)+0x80)
 #define RING_HWS_PGA_GEN6(base)	((base)+0x2080)
 
+#define HSW_GTT_CACHE_EN	0x4024
+#define   GTT_CACHE_EN_ALL	0xF0007FFF
 #define GEN7_WR_WATERMARK	0x4028
 #define GEN7_GFX_PRIO_CTRL	0x402C
 #define ARB_MODE		0x4030
drivers/gpu/drm/i915/intel_pm.c +13 -0
index 5ec56b6..58517a50 100644
@@ -6205,6 +6205,13 @@ static void broadwell_init_clock_gating(struct drm_device *dev)
 	I915_WRITE(GEN8_L3SQCREG1, BDW_WA_L3SQCREG1_DEFAULT);
 	I915_WRITE(GEN7_MISCCPCTL, misccpctl);
 
+	/*
+	 * WaGttCachingOffByDefault:bdw
+	 * GTT cache may not work with big pages, so if those
+	 * are ever enabled GTT cache may need to be disabled.
+	 */
+	I915_WRITE(HSW_GTT_CACHE_EN, GTT_CACHE_EN_ALL);
+
 	lpt_init_clock_gating(dev);
 }
 
@@ -6480,6 +6487,12 @@ static void cherryview_init_clock_gating(struct drm_device *dev)
 	/* WaDisableSDEUnitClockGating:chv */
 	I915_WRITE(GEN8_UCGCTL6, I915_READ(GEN8_UCGCTL6) |
 		   GEN8_SDEUNIT_CLOCK_GATE_DISABLE);
+
+	/*
+	 * GTT cache may not work with big pages, so if those
+	 * are ever enabled GTT cache may need to be disabled.
+	 */
+	I915_WRITE(HSW_GTT_CACHE_EN, GTT_CACHE_EN_ALL);
 }
 
Quoted 4 lines

drivers/gpu/drm/i915/intel_pm.c static void cherryview_init_clock_gating(struct drm_device *dev) · line 6498

Quoted 2 lines
drivers/gpu/drm/i915/intel_pm.c +0 -0
 static void g4x_init_clock_gating(struct drm_device *dev)

Looks ok to me; I guess testing will be the real review here.

Quoted 8 lines

Just aside: If you think the real test for a patch is the real world imo an important part of the review work is to make sure we do have that testing coverage. We have a few igts that specifically exercise gtt tlb issues (from previous generations), so I think we're covered here.

Quoted 1 line

Merged all three, thanks for patches&review.

Yeah I don't think it's something that requires a dedicated test, just lots of coverage with our existing stuff to sniff out problems.

Thanks,
Jesse

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

Reviewed-by on v1 1/3
#8 lore raw
In reply to [email protected]
Quoted 17 lines · On 05/19/2015 10:32 AM, [email protected] wrote:

From: Ville Syrjälä <[email protected]>

We're not using ilk_init_lp_watermarks() on BDW for some reason. Probably due to the BDW patches and the relevant WM patches landing roughlly at the same time. Fix it up.

---
 drivers/gpu/drm/i915/intel_pm.c | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)
drivers/gpu/drm/i915/intel_pm.c +0 -0
index ce1d079..206bd41 100644

drivers/gpu/drm/i915/intel_pm.c static void broadwell_init_clock_gating(struct drm_device *dev) · line 6172

Quoted 10 lines
drivers/gpu/drm/i915/intel_pm.c +1 -3
@@ -6166,9 +6166,7 @@ static void broadwell_init_clock_gating(struct drm_device *dev)
 	struct drm_i915_private *dev_priv = dev->dev_private;
 	enum pipe pipe;
 
-	I915_WRITE(WM3_LP_ILK, 0);
-	I915_WRITE(WM2_LP_ILK, 0);
-	I915_WRITE(WM1_LP_ILK, 0);
+	ilk_init_lp_watermarks(dev);
 
 	/* WaSwitchSolVfFArbitrationPriority:bdw */
 	I915_WRITE(GAM_ECOCHK, I915_READ(GAM_ECOCHK) | HSW_ECOCHK_ARB_PRIO_SOL);

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