Skip to contentSkip to conversation

drm/i915: Use true PPGTT in Gen8 if execlists are enabled

PATCH Superseded Michel Thierry 3 participants 11 messages Dec 11, 2014 – Dec 16, 2014

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

PATCH 1/1 drm/i915: Use true PPGTT in Gen8 if execlists are enabled

When execlists submission is enabled, try full ppgtt by default.

Note, this patch considers that execlist support has been enabled by default on Gen8.

---
 drivers/gpu/drm/i915/i915_gem_gtt.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)
drivers/gpu/drm/i915/i915_gem_gtt.c +6 -3
index 171f6ea..4ed3904 100644
@@ -40,8 +40,8 @@ static int sanitize_enable_ppgtt(struct drm_device *dev, int enable_ppgtt)
 
 	has_aliasing_ppgtt = INTEL_INFO(dev)->gen >= 6;
 	has_full_ppgtt = INTEL_INFO(dev)->gen >= 7;
-	if (IS_GEN8(dev))
-		has_full_ppgtt = false; /* XXX why? */
+	if (IS_GEN8(dev) && !i915.enable_execlists)
+		has_full_ppgtt = false; /* Only enforce with execlists */
 
 	/*
 	 * We don't allow disabling PPGTT for gen9+ as it's a requirement for
@@ -72,7 +72,10 @@ static int sanitize_enable_ppgtt(struct drm_device *dev, int enable_ppgtt)
 		return 0;
 	}
 
-	return has_aliasing_ppgtt ? 1 : 0;
+	if (INTEL_INFO(dev)->gen < 8)
+		return has_aliasing_ppgtt ? 1 : 0;
+	else
+		return has_full_ppgtt ? 2 : has_aliasing_ppgtt ? 1 : 0;
 }
 
 
Signature · 6 lines
2.1.1

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Tested-by on v1 1/1
#1 lore raw

-------------------------------------Summary-------------------------------------
Platform Delta drm-intel-nightly Series Applied
PNV 364/364 364/364
ILK +1-1 364/366 364/366

SNB                                  448/450              448/450
IVB                                  497/498              497/498
BYT                                  289/289              289/289
HSW                                  563/564              563/564
BDW                                  417/417              417/417

-------------------------------------Detailed-------------------------------------
Platform Test drm-intel-nightly Series Applied
*ILK igt_gem_fenced_exec_thrash_no-spare-fences-busy PASS(2, M26M37) DMESG_WARN(1, M37)
ILK igt_kms_flip_wf_vblank-ts-check DMESG_WARN(7, M26)PASS(23, M26M37) PASS(1, M37)
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

#2 lore raw
In reply to Michel Thierry
Quoted 15 lines · Michel Thierry

When execlists submission is enabled, try full ppgtt by default.

Note, this patch considers that execlist support has been enabled by default on Gen8.

---
 drivers/gpu/drm/i915/i915_gem_gtt.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)
drivers/gpu/drm/i915/i915_gem_gtt.c +0 -0
index 171f6ea..4ed3904 100644

drivers/gpu/drm/i915/i915_gem_gtt.c static int sanitize_enable_ppgtt(struct drm_device *dev, int enable_ppgtt) · line 44

Quoted 7 lines
drivers/gpu/drm/i915/i915_gem_gtt.c +2 -2
@@ -40,8 +40,8 @@ static int sanitize_enable_ppgtt(struct drm_device *dev, int enable_ppgtt)
 
 	has_aliasing_ppgtt = INTEL_INFO(dev)->gen >= 6;
 	has_full_ppgtt = INTEL_INFO(dev)->gen >= 7;
-	if (IS_GEN8(dev))
-		has_full_ppgtt = false; /* XXX why? */
+	if (IS_GEN8(dev) && !i915.enable_execlists)
+		has_full_ppgtt = false; /* Only enforce with execlists */

Imo this has outlived it's usefulness - enable_ppgtt is an unsafe parameter so everyone setting it themselves gets what they need. Afair this was just because of the execlist depency on gen8 for ppgtt. -Daniel

Quoted 22 lines
drivers/gpu/drm/i915/i915_gem_gtt.c +4 -1
 	/*
 	 * We don't allow disabling PPGTT for gen9+ as it's a requirement for
@@ -72,7 +72,10 @@ static int sanitize_enable_ppgtt(struct drm_device *dev, int enable_ppgtt)
 		return 0;
 	}
 
-	return has_aliasing_ppgtt ? 1 : 0;
+	if (INTEL_INFO(dev)->gen < 8)
+		return has_aliasing_ppgtt ? 1 : 0;
+	else
+		return has_full_ppgtt ? 2 : has_aliasing_ppgtt ? 1 : 0;
 }
 
 
Signature · 6 lines
2.1.1

_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#3 lore raw
Quoted 17 lines · Daniel Vetter
Quoted 16 lines · Michel Thierry

When execlists submission is enabled, try full ppgtt by default.

Note, this patch considers that execlist support has been enabled by default on Gen8.

---
  drivers/gpu/drm/i915/i915_gem_gtt.c | 9 ++++++---
  1 file changed, 6 insertions(+), 3 deletions(-)
drivers/gpu/drm/i915/i915_gem_gtt.c +0 -0
index 171f6ea..4ed3904 100644
@@ -40,8 +40,8 @@ static int sanitize_enable_ppgtt(struct drm_device *dev, int enable_ppgtt)
 
Quoted 10 lines

drivers/gpu/drm/i915/i915_gem_gtt.c static int sanitize_enable_ppgtt(struct drm_device *dev, int enable_ppgtt) · line 44

Quoted 6 lines
drivers/gpu/drm/i915/i915_gem_gtt.c +2 -2
  	has_aliasing_ppgtt = INTEL_INFO(dev)->gen >= 6;
  	has_full_ppgtt = INTEL_INFO(dev)->gen >= 7;
-	if (IS_GEN8(dev))
-		has_full_ppgtt = false; /* XXX why? */
+	if (IS_GEN8(dev) && !i915.enable_execlists)
+		has_full_ppgtt = false; /* Only enforce with execlists */

Imo this has outlived it's usefulness - enable_ppgtt is an unsafe parameter so everyone setting it themselves gets what they need. Afair this was just because of the execlist depency on gen8 for ppgtt. -Daniel

Not sure if I'm following you up on this...
The aim was to change the default value to _full_ only when execlists are also enable (after Thomas' patch). In gen8, we don't want to have full ppgtt with legacy ring submission.

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

#4 lore raw
Quoted 27 lines · Michel Thierry
Quoted 11 lines · Daniel Vetter
Quoted 10 lines · Michel Thierry

When execlists submission is enabled, try full ppgtt by default.

Note, this patch considers that execlist support has been enabled by default on Gen8.

---
 drivers/gpu/drm/i915/i915_gem_gtt.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)
Quoted 15 lines

drivers/gpu/drm/i915/i915_gem_gtt.c static int sanitize_enable_ppgtt(struct drm_device *dev, int enable_ppgtt) · line 43

Quoted 11 lines
drivers/gpu/drm/i915/i915_gem_gtt.c +2 -2
index 171f6ea..4ed3904 100644
@@ -40,8 +40,8 @@ static int sanitize_enable_ppgtt(struct drm_device *dev, int enable_ppgtt)
 	has_aliasing_ppgtt = INTEL_INFO(dev)->gen >= 6;
 	has_full_ppgtt = INTEL_INFO(dev)->gen >= 7;
-	if (IS_GEN8(dev))
-		has_full_ppgtt = false; /* XXX why? */
+	if (IS_GEN8(dev) && !i915.enable_execlists)
+		has_full_ppgtt = false; /* Only enforce with execlists */

Imo this has outlived it's usefulness - enable_ppgtt is an unsafe parameter so everyone setting it themselves gets what they need. Afair this was just because of the execlist depency on gen8 for ppgtt. -Daniel

Quoted 4 lines

Not sure if I'm following you up on this...
The aim was to change the default value to _full_ only when execlists are also enable (after Thomas' patch). In gen8, we don't want to have full ppgtt with legacy ring submission.

Yeah I gotten confused. The comment is a bit misleading thought, maybe "Full ppgtt needs execlist since otherwise the ctx switch can hang"? -Daniel

Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#5 lore raw
Quoted 28 lines · Daniel Vetter
Quoted 27 lines · Michel Thierry
Quoted 26 lines · Daniel Vetter
Quoted 15 lines · Michel Thierry

When execlists submission is enabled, try full ppgtt by default.

Note, this patch considers that execlist support has been enabled by default on Gen8.

---
 drivers/gpu/drm/i915/i915_gem_gtt.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)
drivers/gpu/drm/i915/i915_gem_gtt.c +0 -0
index 171f6ea..4ed3904 100644

drivers/gpu/drm/i915/i915_gem_gtt.c static int sanitize_enable_ppgtt(struct drm_device *dev, int enable_ppgtt) · line 43

Quoted 6 lines
drivers/gpu/drm/i915/i915_gem_gtt.c +2 -2
@@ -40,8 +40,8 @@ static int sanitize_enable_ppgtt(struct drm_device *dev, int enable_ppgtt)
 	has_aliasing_ppgtt = INTEL_INFO(dev)->gen >= 6;
 	has_full_ppgtt = INTEL_INFO(dev)->gen >= 7;
-	if (IS_GEN8(dev))
-		has_full_ppgtt = false; /* XXX why? */
+	if (IS_GEN8(dev) && !i915.enable_execlists)
+		has_full_ppgtt = false; /* Only enforce with execlists */

Imo this has outlived it's usefulness - enable_ppgtt is an unsafe parameter so everyone setting it themselves gets what they need. Afair this was just because of the execlist depency on gen8 for ppgtt. -Daniel

Quoted 7 lines
Quoted 4 lines

Not sure if I'm following you up on this...
The aim was to change the default value to _full_ only when execlists are also enable (after Thomas' patch). In gen8, we don't want to have full ppgtt with legacy ring submission.

Yeah I gotten confused. The comment is a bit misleading thought, maybe "Full ppgtt needs execlist since otherwise the ctx switch can hang"?

Actually I wasnt' confused: gen8+ has full ppgtt, like gen7. It's broken without execlist though, so what we should do is just change the default below. Otherwise if a user explicitly sets execlist=0 and ppgtt=2 we won't obey that, which doesn't make sense - for isolating bugs this might be a valid testcase (even when we know there's problems). So your patch should do two things:
- Generally allow full ppgtt on gen8 - the current restriction doesn't make sense.
- Switch the default to full ppgtt when execlist is enabled on gen8+, leave the default to aliasing/disabled everywhere else as is.
-Daniel

Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#6 lore raw
Quoted 35 lines · Daniel Vetter
Quoted 28 lines · Daniel Vetter
Quoted 27 lines · Michel Thierry
Quoted 26 lines · Daniel Vetter
Quoted 15 lines · Michel Thierry

When execlists submission is enabled, try full ppgtt by default.

Note, this patch considers that execlist support has been enabled by default on Gen8.

---
  drivers/gpu/drm/i915/i915_gem_gtt.c | 9 ++++++---
  1 file changed, 6 insertions(+), 3 deletions(-)
drivers/gpu/drm/i915/i915_gem_gtt.c +0 -0
index 171f6ea..4ed3904 100644

drivers/gpu/drm/i915/i915_gem_gtt.c static int sanitize_enable_ppgtt(struct drm_device *dev, int enable_ppgtt) · line 43

Quoted 6 lines
drivers/gpu/drm/i915/i915_gem_gtt.c +2 -2
@@ -40,8 +40,8 @@ static int sanitize_enable_ppgtt(struct drm_device *dev, int enable_ppgtt)
  	has_aliasing_ppgtt = INTEL_INFO(dev)->gen >= 6;
  	has_full_ppgtt = INTEL_INFO(dev)->gen >= 7;
-	if (IS_GEN8(dev))
-		has_full_ppgtt = false; /* XXX why? */
+	if (IS_GEN8(dev) && !i915.enable_execlists)
+		has_full_ppgtt = false; /* Only enforce with execlists */

Imo this has outlived it's usefulness - enable_ppgtt is an unsafe parameter so everyone setting it themselves gets what they need. Afair this was just because of the execlist depency on gen8 for ppgtt. -Daniel

Quoted 6 lines
Quoted 4 lines

Not sure if I'm following you up on this...
The aim was to change the default value to _full_ only when execlists are also enable (after Thomas' patch). In gen8, we don't want to have full ppgtt with legacy ring submission.

Yeah I gotten confused. The comment is a bit misleading thought, maybe "Full ppgtt needs execlist since otherwise the ctx switch can hang"?

Quoted 11 lines

Actually I wasnt' confused: gen8+ has full ppgtt, like gen7. It's broken without execlist though, so what we should do is just change the default below. Otherwise if a user explicitly sets execlist=0 and ppgtt=2 we won't obey that, which doesn't make sense - for isolating bugs this might be a valid testcase (even when we know there's problems). So your patch should do two things:
- Generally allow full ppgtt on gen8 - the current restriction doesn't make sense.
- Switch the default to full ppgtt when execlist is enabled on gen8+, leave the default to aliasing/disabled everywhere else as is.
-Daniel

Thanks for clarifying. I'll send a new patch.

-Michel

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

#7 lore raw

PATCH v2 1/1 drm/i915: Use true PPGTT in Gen8+ when execlists are enabled

In reply to Michel Thierry

In Gen8+, full ppgtt needs execlist, otherwise the ctx switch can hang.

Also remove the current restriction, a user should be able to explicitly set ppgtt=2.

Note, this patch considers that execlist support has been enabled by default on Gen8.

v2: Remove non-default restriction and clarify comment (Daniel)

---
 drivers/gpu/drm/i915/i915_gem_gtt.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)
drivers/gpu/drm/i915/i915_gem_gtt.c +4 -3
index 171f6ea..b64323b 100644
@@ -40,8 +40,6 @@ static int sanitize_enable_ppgtt(struct drm_device *dev, int enable_ppgtt)
 
 	has_aliasing_ppgtt = INTEL_INFO(dev)->gen >= 6;
 	has_full_ppgtt = INTEL_INFO(dev)->gen >= 7;
-	if (IS_GEN8(dev))
-		has_full_ppgtt = false; /* XXX why? */
 
 	/*
 	 * We don't allow disabling PPGTT for gen9+ as it's a requirement for
@@ -72,7 +70,10 @@ static int sanitize_enable_ppgtt(struct drm_device *dev, int enable_ppgtt)
 		return 0;
 	}
 
-	return has_aliasing_ppgtt ? 1 : 0;
+	if (INTEL_INFO(dev)->gen >= 8 && i915.enable_execlists)
+		return 2;
+	else
+		return has_aliasing_ppgtt ? 1 : 0;
 }
 
 
Signature · 6 lines
2.1.1

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

Re: [PATCH v2] drm/i915: Use true PPGTT in Gen8+ when execlists are enabled

Quoted 9 lines · Michel Thierry

In Gen8+, full ppgtt needs execlist, otherwise the ctx switch can hang.

Also remove the current restriction, a user should be able to explicitly set ppgtt=2.

Note, this patch considers that execlist support has been enabled by default on Gen8.

v2: Remove non-default restriction and clarify comment (Daniel)

Missing git add since there's no no comment any more? -Daniel

Quoted 35 lines
---
 drivers/gpu/drm/i915/i915_gem_gtt.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)
drivers/gpu/drm/i915/i915_gem_gtt.c +4 -3
index 171f6ea..b64323b 100644
@@ -40,8 +40,6 @@ static int sanitize_enable_ppgtt(struct drm_device *dev, int enable_ppgtt)
 
 	has_aliasing_ppgtt = INTEL_INFO(dev)->gen >= 6;
 	has_full_ppgtt = INTEL_INFO(dev)->gen >= 7;
-	if (IS_GEN8(dev))
-		has_full_ppgtt = false; /* XXX why? */
 
 	/*
 	 * We don't allow disabling PPGTT for gen9+ as it's a requirement for
@@ -72,7 +70,10 @@ static int sanitize_enable_ppgtt(struct drm_device *dev, int enable_ppgtt)
 		return 0;
 	}
 
-	return has_aliasing_ppgtt ? 1 : 0;
+	if (INTEL_INFO(dev)->gen >= 8 && i915.enable_execlists)
+		return 2;
+	else
+		return has_aliasing_ppgtt ? 1 : 0;
 }
 
 
Signature · 1 line
2.1.1
Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
#9 lore raw

Re: [PATCH v2] drm/i915: Use true PPGTT in Gen8+ when execlists are enabled

Quoted 6 lines · Daniel Vetter
Quoted 5 lines · Michel Thierry

In Gen8+, full ppgtt needs execlist, otherwise the ctx switch can hang.

Also remove the current restriction, a user should be able to explicitly set ppgtt=2.

Quoted 6 lines
Quoted 4 lines

Note, this patch considers that execlist support has been enabled by default on Gen8.

v2: Remove non-default restriction and clarify comment (Daniel)

Missing git add since there's no no comment any more?

Ok, Michel clarified on irc that he meant commit message. I've fixed that up and merged the patch.
-Daniel

Quoted 42 lines

-Daniel

Quoted 35 lines
---
 drivers/gpu/drm/i915/i915_gem_gtt.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)
drivers/gpu/drm/i915/i915_gem_gtt.c +4 -3
index 171f6ea..b64323b 100644
@@ -40,8 +40,6 @@ static int sanitize_enable_ppgtt(struct drm_device *dev, int enable_ppgtt)
 
 	has_aliasing_ppgtt = INTEL_INFO(dev)->gen >= 6;
 	has_full_ppgtt = INTEL_INFO(dev)->gen >= 7;
-	if (IS_GEN8(dev))
-		has_full_ppgtt = false; /* XXX why? */
 
 	/*
 	 * We don't allow disabling PPGTT for gen9+ as it's a requirement for
@@ -72,7 +70,10 @@ static int sanitize_enable_ppgtt(struct drm_device *dev, int enable_ppgtt)
 		return 0;
 	}
 
-	return has_aliasing_ppgtt ? 1 : 0;
+	if (INTEL_INFO(dev)->gen >= 8 && i915.enable_execlists)
+		return 2;
+	else
+		return has_aliasing_ppgtt ? 1 : 0;
 }
 
 
Signature · 1 line
2.1.1
Signature · 3 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
Signature · 7 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Tested-by on v2 1/1
#10 lore raw

Re: [PATCH v2] drm/i915: Use true PPGTT in Gen8+ when execlists are enabled

In reply to Michel Thierry

-------------------------------------Summary-------------------------------------
Platform Delta drm-intel-nightly Series Applied
PNV 366/366 366/366
ILK +5 360/366 365/366

SNB                                  448/450              448/450
IVB                                  497/498              497/498
BYT                                  289/289              289/289
HSW                                  563/564              563/564
BDW                                  417/417              417/417
-------------------------------------Detailed-------------------------------------
Platform  Test                                drm-intel-nightly          Series Applied
 ILK  igt_drv_suspend_fence-restore-untiled      DMESG_WARN(1, M26)PASS(7, M37M26)      PASS(1, M37)
 ILK  igt_kms_flip_bcs-flip-vs-modeset-interruptible      DMESG_WARN(1, M26)PASS(7, M37M26)      PASS(1, M37)
 ILK  igt_kms_flip_busy-flip-interruptible      DMESG_WARN(2, M26)PASS(6, M37M26)      PASS(1, M37)
 ILK  igt_kms_flip_flip-vs-rmfb-interruptible      DMESG_WARN(1, M26)PASS(7, M37M26)      PASS(1, M37)
 ILK  igt_kms_flip_rcs-flip-vs-dpms      DMESG_WARN(1, M26)PASS(6, M37M26)      PASS(1, M37)

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