From commit e95433c73a11759203af1cae5958f998c9673370, workload status setting
was changed to only capture on error path, but we need to set it properly in
normal path too, otherwise we'll fail to complete workload which could lead
guest VM vGPU reset.
Cc: Chris Wilson
== Series Details ==
Series: drm/i915: Set the Z inversion overlap field
URL : https://patchwork.freedesktop.org/series/14704/
State : warning
== Summary ==
Series 14704v1 drm/i915: Set the Z inversion overlap field
https://patchwork.freedesktop.org/api/1.0/series/14704/revisions/1/mbox/
Dual link Z-inversion overlap field is present
in MIPI_CTRL register unlike the older platforms,
hence setting the same in this patch.
Signed-off-by: Deepak M
---
drivers/gpu/drm/i915/i915_reg.h | 2 ++
drivers/gpu/drm/i915/intel_dsi.c | 17 +
2 files
On Tue, Nov 01, 2016 at 12:34:37PM -0700, Manasi Navare wrote:
> On Tue, Nov 01, 2016 at 09:16:28PM +0200, Jani Nikula wrote:
> > On Tue, 01 Nov 2016, Manasi Navare wrote:
> > > On Tue, Nov 01, 2016 at 10:49:14AM +0200, Jani Nikula wrote:
> > >> On Sat, 29 Oct 2016,
You can easily run out of GTT space with the current fixed allocation
of 64MB. Bump it to 128MB to avoid sporadic page-fault errors with
the simulator.
---
tools/aubdump.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/tools/aubdump.c b/tools/aubdump.c
index
Gen8+ have 64 bit GTT entries, so we need to allocate twice as much
space for the GTT table in order to cover the same number of GTT
pages. Fixes sporadic page-fault crashes on the simulator.
---
tools/aubdump.c | 21 -
1 file changed, 16 insertions(+), 5 deletions(-)
diff
>-Original Message-
>From: Intel-gfx [mailto:intel-gfx-boun...@lists.freedesktop.org] On Behalf Of
>Lionel Landwerlin
>Sent: Tuesday, November 01, 2016 11:15 AM
>To: intel-gfx@lists.freedesktop.org
>Subject: [Intel-gfx] [PATCH v3 i-g-t 1/2] tools: intel_aubdump: pass
>configuration
>-Original Message-
>From: Landwerlin, Lionel G
>Sent: Tuesday, November 01, 2016 11:15 AM
>To: intel-gfx@lists.freedesktop.org
>Cc: Landwerlin, Lionel G ; Gandikota, Sirisha
>
>Subject: [PATCH v4 i-g-t 2/2] aubdump: add --command
On Tue, Nov 01, 2016 at 06:43:03PM +0200, Mika Kuoppala wrote:
> Create new file for hangcheck specific code, intel_hangcheck.c,
> and move all related code in it.
>
> v2: s/intel_engine_hangcheck/intel_engine (Chris)
>
> No functional changes.
>
> Cc: Chris Wilson
>
On Tue, 01 Nov 2016, "Pandiyan, Dhinakaran"
wrote:
> On Tue, 2016-11-01 at 21:35 +0200, Jani Nikula wrote:
>> On Tue, 01 Nov 2016, Dhinakaran Pandiyan
>> wrote:
>> > According to BSpec, cdclk for BDW has to be not less than 432 MHz
On Tue, 2016-11-01 at 21:35 +0200, Jani Nikula wrote:
> On Tue, 01 Nov 2016, Dhinakaran Pandiyan
> wrote:
> > According to BSpec, cdclk for BDW has to be not less than 432 MHz with DP
> > audio enabled, port width x4, and link rate HBR2 (5.4 GHz). With cdclk less
>
On Tue, 01 Nov 2016, Dhinakaran Pandiyan wrote:
> According to BSpec, cdclk for BDW has to be not less than 432 MHz with DP
> audio enabled, port width x4, and link rate HBR2 (5.4 GHz). With cdclk less
> than 432 MHz, enabling audio leads to pipe FIFO underruns and
On Tue, Nov 01, 2016 at 09:16:28PM +0200, Jani Nikula wrote:
> On Tue, 01 Nov 2016, Manasi Navare wrote:
> > On Tue, Nov 01, 2016 at 10:49:14AM +0200, Jani Nikula wrote:
> >> On Sat, 29 Oct 2016, Manasi Navare wrote:
> >> > If link training
On Tue, 01 Nov 2016, Dhinakaran Pandiyan wrote:
> According to BSpec, cdclk for BDW has to be not less than 432 MHz with DP
> audio enabled, port width x4, and link rate HBR2 (5.4 GHz). With cdclk less
> than 432 MHz, enabling audio leads to pipe FIFO underruns and
On Tue, 01 Nov 2016, Manasi Navare wrote:
> On Tue, Nov 01, 2016 at 10:49:14AM +0200, Jani Nikula wrote:
>> On Sat, 29 Oct 2016, Manasi Navare wrote:
>> > If link training at a link rate optimal for a particular
>> > mode fails during
== Series Details ==
Series: series starting with [v4,1/2] drm/i915/dp: BDW cdclk fix for DP audio
URL : https://patchwork.freedesktop.org/series/14688/
State : warning
== Summary ==
Series 14688v1 Series without cover letter
According to BSpec, cdclk for BDW has to be not less than 432 MHz with DP
audio enabled, port width x4, and link rate HBR2 (5.4 GHz). With cdclk less
than 432 MHz, enabling audio leads to pipe FIFO underruns and displays
cycling on/off.
From BSpec:
"Display» BDW-SKL» dpr» [Register] DP_TP_CTL
According to BSpec, cdclk for BDW has to be not less than 432 MHz with DP
audio enabled, port width x4, and link rate HBR2 (5.4 GHz). With cdclk less
than 432 MHz, enabling audio leads to pipe FIFO underruns and displays
cycling on/off.
Let's apply this work around to GEN9 platforms too, as it
== Series Details ==
Series: drm/i915: Move hangcheck code out from i915_irq.c
URL : https://patchwork.freedesktop.org/series/14685/
State : warning
== Summary ==
Series 14685v1 drm/i915: Move hangcheck code out from i915_irq.c
This makes parsing options less complicated and easier to extend.
v2: Fix device id parsing (atoi -> sscanf) (Sirisha)
Combine with previous commit moving init function (Sirisha)
v3: Fix behavior change between bash 4.3 & 4.4 in <<< with \n characters
(Lionel)
Signed-off-by: Lionel
This comes handy if you want to look at your application output without
having to save it into a file. For example, use this with aubinator from
Mesa :
$ intel_aubdump -c '/path/to/aubinator --gen=hsw' my_gl_app
v2: Fix handling empty command line option
v3: Fix command line concatenation
>-Original Message-
>From: Mcgee, Jeff
>Sent: Tuesday, November 1, 2016 11:22 AM
>To: Srivatsa, Anusha
>Cc: intel-gfx@lists.freedesktop.org
>Subject: Re: [PATCH] drm/i915/huc: Update the construction of file path for HuC
>similar to that of GuC
>
>On Mon, Oct
On Mon, Oct 31, 2016 at 03:24:53PM -0700, Anusha Srivatsa wrote:
> Update the file construction and specifying the required version
> similar to that of GuC.Add an extra field for the build number.
> Adopted the approach used in
> https://patchwork.freedesktop.org/patch/104355/
>
> Cc: Jeff
== Series Details ==
Series: series starting with [1/2] drm/i915: Move hangcheck code out from
i915_irq.c
URL : https://patchwork.freedesktop.org/series/14682/
State : warning
== Summary ==
Series 14682v1 Series without cover letter
On Tue, Nov 1, 2016 at 2:57 PM, Chris Wilson
wrote:
> On Fri, Oct 28, 2016 at 03:14:29AM +0100, Robert Bragg wrote:
> > This adds 'compute', 'compute extended', 'memory reads', 'memory writes'
> > and 'sampler balance' metric sets for Haswell.
> >
> > The code is auto
Create new file for hangcheck specific code, intel_hangcheck.c,
and move all related code in it.
v2: s/intel_engine_hangcheck/intel_engine (Chris)
No functional changes.
Cc: Chris Wilson
Cc: Joonas Lahtinen
Cc: Tvrtko Ursulin
On Tue, Nov 01, 2016 at 04:36:34PM +, Tvrtko Ursulin wrote:
>
> Merged to dinq, thanks for the review!
Almost as bad as me in forgetting to add the r-b. dim save us! :)
-Chris
--
Chris Wilson, Intel Open Source Technology Centre
___
Intel-gfx
Chris Wilson writes:
> On Tue, Nov 01, 2016 at 06:03:21PM +0200, Mika Kuoppala wrote:
>> Create new file for hangcheck specific code, intel_engine_hangcheck.c,
>> and move all the related code in it.
>>
>> No functional changes.
>>
>> Cc: Chris Wilson
On 01/11/2016 16:16, Patchwork wrote:
== Series Details ==
Series: drm/i915: Allow shrinking of userptr objects once again (rev2)
URL : https://patchwork.freedesktop.org/series/14675/
State : warning
== Summary ==
Series 14675v2 drm/i915: Allow shrinking of userptr objects once again
Patchwork writes:
> == Series Details ==
>
> Series: series starting with [1/2] drm/i915/gtt: Fix pte clear range (rev3)
> URL : https://patchwork.freedesktop.org/series/14620/
> State : warning
>
> == Summary ==
>
> Series 14620v3 Series without cover letter
On Tue, Nov 01, 2016 at 10:49:14AM +0200, Jani Nikula wrote:
> On Sat, 29 Oct 2016, Manasi Navare wrote:
> > If link training at a link rate optimal for a particular
> > mode fails during modeset's atomic commit phase, then we
> > let the modeset complete and then
On Tue, Nov 01, 2016 at 06:03:21PM +0200, Mika Kuoppala wrote:
> Create new file for hangcheck specific code, intel_engine_hangcheck.c,
> and move all the related code in it.
>
> No functional changes.
>
> Cc: Chris Wilson
> Cc: Joonas Lahtinen
On Tue, Nov 01, 2016 at 06:03:22PM +0200, Mika Kuoppala wrote:
> We have a place already for error handling and error
> state capture, i915_gpu_error.c. Move code to more
> appropriate file.
i915_gpu_error.c is conditionally compiled iff we want error capture. We
need reset handling irrespective
== Series Details ==
Series: drm/i915: Allow shrinking of userptr objects once again (rev2)
URL : https://patchwork.freedesktop.org/series/14675/
State : warning
== Summary ==
Series 14675v2 drm/i915: Allow shrinking of userptr objects once again
Create new file for hangcheck specific code, intel_engine_hangcheck.c,
and move all the related code in it.
No functional changes.
Cc: Chris Wilson
Cc: Joonas Lahtinen
Cc: Tvrtko Ursulin
Signed-off-by: Mika
We have a place already for error handling and error
state capture, i915_gpu_error.c. Move code to more
appropriate file.
No functional changes.
Cc: Chris Wilson
Cc: Joonas Lahtinen
Cc: Tvrtko Ursulin
== Series Details ==
Series: series starting with [1/2] drm/i915/gtt: Fix pte clear range (rev3)
URL : https://patchwork.freedesktop.org/series/14620/
State : warning
== Summary ==
Series 14620v3 Series without cover letter
Hi,
On 02/03/16 14:00, Tomeu Vizoso wrote:
igt_create_bo_with_dimensions() is intended to abstract differences
between drivers in buffer object creation.
The driver-specific ioctls will be called if the driver that is being
tested can satisfy the needs of the calling subtest, or it will be
If we define drm_compat_ioctl NULL on CONFIG_COMPAT=n, we don't have to
check for the config everywhere.
Signed-off-by: Jani Nikula
---
Just an idea on top of Patrik's patch.
---
drivers/gpu/drm/arc/arcpgu_drv.c| 2 --
drivers/gpu/drm/arm/hdlcd_drv.c
On Tue, Nov 01, 2016 at 02:44:10PM +, Tvrtko Ursulin wrote:
> From: Tvrtko Ursulin
>
> Commit 1bec9b0bda3d ("drm/i915/shrinker: Only shmemfs objects
> are backed by swap") stopped considering the userptr objects
> in shrinker callbacks.
>
> Restore that so idle
On Fri, Oct 28, 2016 at 03:14:29AM +0100, Robert Bragg wrote:
> This adds 'compute', 'compute extended', 'memory reads', 'memory writes'
> and 'sampler balance' metric sets for Haswell.
>
> The code is auto generated from an XML description of metric sets,
> currently maintained in gputop, ref:
>
On Mon, Oct 31, 2016 at 08:56:34PM +, Chris Wilson wrote:
> On Mon, Oct 31, 2016 at 10:36:59PM +0200, ville.syrj...@linux.intel.com wrote:
> > From: Ville Syrjälä
> >
> > I pretty much just wanted to store struct intel_crtc * instead
> > of struct drm_crtc * in
From: Tvrtko Ursulin
Commit 1bec9b0bda3d ("drm/i915/shrinker: Only shmemfs objects
are backed by swap") stopped considering the userptr objects
in shrinker callbacks.
Restore that so idle userptr objects can be discarded in order
to free up memory.
One change further
On Tue, Nov 01, 2016 at 02:29:39PM +, Tvrtko Ursulin wrote:
>
> On 01/11/2016 13:52, Chris Wilson wrote:
> >On Tue, Nov 01, 2016 at 01:28:06PM +, Tvrtko Ursulin wrote:
> >>From: Tvrtko Ursulin
> >>
> >>Commit 1bec9b0bda3d ("drm/i915/shrinker: Only shmemfs
On 01/11/2016 13:52, Chris Wilson wrote:
On Tue, Nov 01, 2016 at 01:28:06PM +, Tvrtko Ursulin wrote:
From: Tvrtko Ursulin
Commit 1bec9b0bda3d ("drm/i915/shrinker: Only shmemfs objects
are backed by swap") stopped considering the userptr objects
in shrinker
On Mon, Oct 31, 2016 at 03:28:08PM +, Matthew Auld wrote:
> On 30 October 2016 at 13:28, Chris Wilson wrote:
> > For legacy contexts we employ an optimisation to only flush the context
> > when binding into the global GTT. This avoids stalling onthe GPU when
> >
On Tue, Nov 01, 2016 at 01:46:00PM -, Patchwork wrote:
> == Series Details ==
>
> Series: drm/i915: A game of OCD dominoes
> URL : https://patchwork.freedesktop.org/series/14634/
> State : warning
>
> == Summary ==
>
> Series 14634v1 drm/i915: A game of OCD dominoes
>
On Tue, Nov 01, 2016 at 01:28:06PM +, Tvrtko Ursulin wrote:
> From: Tvrtko Ursulin
>
> Commit 1bec9b0bda3d ("drm/i915/shrinker: Only shmemfs objects
> are backed by swap") stopped considering the userptr objects
> in shrinker callbacks.
>
> Restore that so idle
== Series Details ==
Series: drm/i915: A game of OCD dominoes
URL : https://patchwork.freedesktop.org/series/14634/
State : warning
== Summary ==
Series 14634v1 drm/i915: A game of OCD dominoes
https://patchwork.freedesktop.org/api/1.0/series/14634/revisions/1/mbox/
Test kms_cursor_legacy:
On Tue, Nov 01, 2016 at 02:34:00PM +0100, Maarten Lankhorst wrote:
> Op 01-11-16 om 14:09 schreef Ville Syrjälä:
> > On Mon, Oct 17, 2016 at 02:37:00PM +0200, Maarten Lankhorst wrote:
> >> Add for_each_(old)(new)_(plane,connector,crtc)_in_state iterators to
> >> replace the old
Op 01-11-16 om 14:09 schreef Ville Syrjälä:
> On Mon, Oct 17, 2016 at 02:37:00PM +0200, Maarten Lankhorst wrote:
>> Add for_each_(old)(new)_(plane,connector,crtc)_in_state iterators to
>> replace the old for_each_xxx_in_state ones. This is useful for >1 flip
>> depth and getting rid of all
Comparing pte index to a number of entries is wrong
when clearing a range of pte entries. Use end marker
of 'one past' to correctly point adequate number of
ptes to the scratch page.
v2: assert early instead of warning late (Chris)
v3: removed consts (Joonas)
Fixes: d209b9c3cd28 ("drm/i915/gtt:
From: Tvrtko Ursulin
Commit 1bec9b0bda3d ("drm/i915/shrinker: Only shmemfs objects
are backed by swap") stopped considering the userptr objects
in shrinker callbacks.
Restore that so idle userptr objects can be discarded in order
to free up memory.
One change further
== Series Details ==
Series: drm/i915: Improve lockdep tracking for obj->mm.lock
URL : https://patchwork.freedesktop.org/series/14672/
State : warning
== Summary ==
Series 14672v1 drm/i915: Improve lockdep tracking for obj->mm.lock
On Mon, Oct 17, 2016 at 02:37:00PM +0200, Maarten Lankhorst wrote:
> Add for_each_(old)(new)_(plane,connector,crtc)_in_state iterators to
> replace the old for_each_xxx_in_state ones. This is useful for >1 flip
> depth and getting rid of all xxx->state dereferences.
>
> Signed-off-by: Maarten
On Wed, Oct 19, 2016 at 03:55:38PM +0200, Maarten Lankhorst wrote:
> This gets rid of a warning that the connectors are used without locking
> when doing a nonblocking modeset.
>
> Signed-off-by: Maarten Lankhorst
Reviewed-by: Ville Syrjälä
On Tue, Nov 01, 2016 at 02:40:35PM +0200, Ville Syrjälä wrote:
> On Tue, Nov 01, 2016 at 09:57:43AM +0100, Maarten Lankhorst wrote:
> > Op 28-10-16 om 18:59 schreef ville.syrj...@linux.intel.com:
> > > From: Ville Syrjälä
> > >
> > > When we end up not recomputing
== Series Details ==
Series: drm/i915: Store the vma in an rbtree under the object
URL : https://patchwork.freedesktop.org/series/14671/
State : warning
== Summary ==
Series 14671v1 drm/i915: Store the vma in an rbtree under the object
On Tue, Nov 01, 2016 at 10:07:51AM +0100, Paul Bolle wrote:
> On Tue, 2016-11-01 at 09:57 +0100, Maarten Lankhorst wrote:
> > Otherwise looks sane, I have a similar patch in my tree. I didn't
> > submit it yet but the fix was similar. Except for the
> > dev_cdclk stuff.
> >
> > With the last
On Tue, Nov 01, 2016 at 09:57:43AM +0100, Maarten Lankhorst wrote:
> Op 28-10-16 om 18:59 schreef ville.syrj...@linux.intel.com:
> > From: Ville Syrjälä
> >
> > When we end up not recomputing the cdclk, we need to populate
> > intel_state->cdclk with the
Em Ter, 2016-11-01 às 12:04 +0100, Maarten Lankhorst escreveu:
> Using for_each_intel_plane_on_crtc will allow us to find all
> allocations
> that may have changed, not just the one added by the atomic state.
>
> This will print changes to plane allocations for crtc's when some
> planes are not
On ti, 2016-11-01 at 12:11 +, Chris Wilson wrote:
> The shrinker may appear to recurse into obj->mm.lock as the shrinker may
> be called from a direct reclaim path whilst handling get_pages. We
> filter out recursing on the same obj->mm.lock by inspecting
> obj->mm.pages, but we do want to
== Series Details ==
Series: drm/i915/bxt: Don't set OCL2_LDOFUSE_PWR_DIS bit in phy init sequence
URL : https://patchwork.freedesktop.org/series/14669/
State : warning
== Summary ==
Series 14669v1 drm/i915/bxt: Don't set OCL2_LDOFUSE_PWR_DIS bit in phy init
sequence
The shrinker may appear to recurse into obj->mm.lock as the shrinker may
be called from a direct reclaim path whilst handling get_pages. We
filter out recursing on the same obj->mm.lock by inspecting
obj->mm.pages, but we do want to take the lock on a second object in
order to reap their pages.
With full-ppgtt one of the main bottlenecks is the lookup of the VMA
underneath the object. For execbuf there is merit in having a very fast
direct lookup of ctx:handle to the vma using a hashtree, but that still
leaves a large number of other lookups. One way to speed up the lookup
would be to
On ti, 2016-11-01 at 13:11 +0200, Ander Conselvan de Oliveira wrote:
> Hardware engineers confirmed that writing to it has no effect, as implied by
> the FIXME comment.
>
> Cc: Imre Deak
> Signed-off-by: Ander Conselvan de Oliveira
>
On Mon, Oct 31, 2016 at 09:16:10PM -, Patchwork wrote:
> == Series Details ==
>
> Series: drm/i915: A game of OCD dominoes
> URL : https://patchwork.freedesktop.org/series/14634/
> State : warning
>
> == Summary ==
>
> Series 14634v1 drm/i915: A game of OCD dominoes
>
On Tue, Nov 01, 2016 at 10:46:07AM -, Patchwork wrote:
> == Series Details ==
>
> Series: series starting with [CI,1/2] drm/i915: Avoid accessing
> request->timeline outside of its lifetime
> URL : https://patchwork.freedesktop.org/series/14665/
> State : warning
>
> == Summary ==
>
>
Since the modeline may differ from actual hardware timings, do not rely
upon it but instead measure the actual and verify that it does not
change across the various flip/vblank configurations.
Signed-off-by: Chris Wilson
---
tests/kms_flip.c | 33
Signed-off-by: Chris Wilson
---
tests/kms_flip.c | 24
1 file changed, 12 insertions(+), 12 deletions(-)
diff --git a/tests/kms_flip.c b/tests/kms_flip.c
index 9829b35..74754d1 100644
--- a/tests/kms_flip.c
+++ b/tests/kms_flip.c
@@ -574,7
Using for_each_intel_plane_on_crtc will allow us to find all allocations
that may have changed, not just the one added by the atomic state.
This will print changes to plane allocations for crtc's when some
planes are not added to the atomic state.
Changes since v1:
- Rephrase commit message.
Hardware engineers confirmed that writing to it has no effect, as implied by
the FIXME comment.
Cc: Imre Deak
Signed-off-by: Ander Conselvan de Oliveira
---
drivers/gpu/drm/i915/intel_dpio_phy.c | 16
1 file changed,
== Series Details ==
Series: series starting with [CI,1/2] drm/i915: Avoid accessing
request->timeline outside of its lifetime
URL : https://patchwork.freedesktop.org/series/14665/
State : warning
== Summary ==
Series 14665v1 Series without cover letter
On Tue, Nov 01, 2016 at 12:22:45PM +0200, Mika Kuoppala wrote:
> Joonas Lahtinen writes:
> >> @@ -735,8 +737,8 @@ static bool gen8_ppgtt_clear_pt(struct
> >> i915_address_space *vm,
> >>
> >> pt_vaddr = kmap_px(pt);
> >>
> >> - for (pte = pte_start; pte <
Joonas Lahtinen writes:
> On ma, 2016-10-31 at 17:55 +0200, Mika Kuoppala wrote:
>> @@ -712,13 +712,13 @@ static int gen8_48b_mm_switch(struct i915_hw_ppgtt
>> *ppgtt,
>> */
>> static bool gen8_ppgtt_clear_pt(struct i915_address_space *vm,
>>
Whilst waiting on a request, we may do so without holding any locks or
any guards beyond a reference to the request. In order to avoid taking
locks within request deallocation, we drop references to its timeline
(via the context and ppgtt) upon retirement. We should avoid chasing
such pointers
If we have a tiled object and an unknown CPU swizzle pattern, we pin the
pages to prevent the object from being swapped out (and us corrupting
the contents as we do not know the access pattern and so cannot convert
it to linear and back to tiled on reuse). This requires us to remember
to drop the
On Tue, Nov 01, 2016 at 09:43:53AM +, Tvrtko Ursulin wrote:
>
> On 31/10/2016 10:26, Chris Wilson wrote:
> >With full-ppgtt one of the main bottlenecks is the lookup of the VMA
> >underneath the object. For execbuf there is merit in having a very fast
> >direct lookup of ctx:handle to the vma
On Mon, Oct 31, 2016 at 08:56:34PM +, Chris Wilson wrote:
> On Mon, Oct 31, 2016 at 10:36:59PM +0200, ville.syrj...@linux.intel.com wrote:
> > From: Ville Syrjälä
> >
> > I pretty much just wanted to store struct intel_crtc * instead
> > of struct drm_crtc * in
On Tue, Nov 01, 2016 at 09:20:33AM +, Tvrtko Ursulin wrote:
>
> On 01/11/2016 09:06, Chris Wilson wrote:
> >On Tue, Nov 01, 2016 at 08:54:09AM +, Tvrtko Ursulin wrote:
> >>
> >>On 01/11/2016 08:50, Chris Wilson wrote:
> >>>On Tue, Nov 01, 2016 at 08:41:01AM +, Tvrtko Ursulin wrote:
>
On 31/10/2016 10:26, Chris Wilson wrote:
With full-ppgtt one of the main bottlenecks is the lookup of the VMA
underneath the object. For execbuf there is merit in having a very fast
direct lookup of ctx:handle to the vma using a hashtree, but that still
leaves a large number of other lookups.
On Tue, Nov 01, 2016 at 09:20:19AM -, Patchwork wrote:
> == Series Details ==
>
> Series: series starting with [CI,1/3] drm/i915: Use the full hammer when
> shutting down the rcu tasks
> URL : https://patchwork.freedesktop.org/series/14660/
> State : warning
>
> == Summary ==
>
> Series
On 01/11/2016 09:06, Chris Wilson wrote:
On Tue, Nov 01, 2016 at 08:54:09AM +, Tvrtko Ursulin wrote:
On 01/11/2016 08:50, Chris Wilson wrote:
On Tue, Nov 01, 2016 at 08:41:01AM +, Tvrtko Ursulin wrote:
On 31/10/2016 10:26, Chris Wilson wrote:
With full-ppgtt one of the main
== Series Details ==
Series: series starting with [CI,1/3] drm/i915: Use the full hammer when
shutting down the rcu tasks
URL : https://patchwork.freedesktop.org/series/14660/
State : warning
== Summary ==
Series 14660v1 Series without cover letter
On ti, 2016-11-01 at 09:15 +, Chris Wilson wrote:
> On Tue, Nov 01, 2016 at 11:10:31AM +0200, Joonas Lahtinen wrote:
> >
> > On ma, 2016-10-31 at 12:40 +, Chris Wilson wrote:
> > As discussed in IRC, I think we rather should add a lock class to
> > obj->mm, to avoid chasing these in the
On Tue, Nov 01, 2016 at 11:10:31AM +0200, Joonas Lahtinen wrote:
> On ma, 2016-10-31 at 12:40 +, Chris Wilson wrote:
> As discussed in IRC, I think we rather should add a lock class to
> obj->mm, to avoid chasing these in the future.
All ears :) The complicated one isn't obj->mm, but the
On ma, 2016-10-31 at 12:40 +, Chris Wilson wrote:
> As we may allocate from within the obj->mm.lock we may enter the
> shrinker for direct reclaim. Operating on the current object is
> prevented by checking for obj->mm.pages (which is only set as the last
> operation in the allocation path).
On Tue, 2016-11-01 at 09:57 +0100, Maarten Lankhorst wrote:
> Otherwise looks sane, I have a similar patch in my tree. I didn't
> submit it yet but the fix was similar. Except for the
> dev_cdclk stuff.
>
> With the last dev_cdclk assignment removed:
>
> Reviewed-by: Maarten Lankhorst
On Tue, Nov 01, 2016 at 08:54:09AM +, Tvrtko Ursulin wrote:
>
> On 01/11/2016 08:50, Chris Wilson wrote:
> >On Tue, Nov 01, 2016 at 08:41:01AM +, Tvrtko Ursulin wrote:
> >>
> >>On 31/10/2016 10:26, Chris Wilson wrote:
> >>>With full-ppgtt one of the main bottlenecks is the lookup of the
On Tue, 2016-11-01 at 09:25 +0100, Maarten Lankhorst wrote:
> Op 20-10-16 om 11:27 schreef Mika Kahola:
> >
> > This is a testcase with multiple planes. The idea here is the
> > following
> >
> > - draw a uniform frame with blue color
> > - grab crc for reference
> > - put planes randomly on
Op 28-10-16 om 18:59 schreef ville.syrj...@linux.intel.com:
> From: Ville Syrjälä
>
> When we end up not recomputing the cdclk, we need to populate
> intel_state->cdclk with the "atomic_cdclk_freq" instead of the
> current cdclk_freq. When no pipes are active, the
On 01/11/2016 08:50, Chris Wilson wrote:
On Tue, Nov 01, 2016 at 08:41:01AM +, Tvrtko Ursulin wrote:
On 31/10/2016 10:26, Chris Wilson wrote:
With full-ppgtt one of the main bottlenecks is the lookup of the VMA
underneath the object. For execbuf there is merit in having a very fast
On 01/11/2016 08:48, Chris Wilson wrote:
On Tue, Nov 01, 2016 at 08:39:14AM +, Tvrtko Ursulin wrote:
On 31/10/2016 10:26, Chris Wilson wrote:
If we have a tiled object and an unknown CPU swizzle pattern, we pin the
pages to prevent the object from being swapped out (and us corrupting
the
On Sat, 29 Oct 2016, Manasi Navare wrote:
> If link training at a link rate optimal for a particular
> mode fails during modeset's atomic commit phase, then we
> let the modeset complete and then retry. We save the link rate
> value at which link training failed, update
During shrinking, we walk over the list of objects searching for
victims. Any that are not removed are put back into the global list.
Currently, they are put back in order (at the front) which means they
will be first to be scanned again. If we instead move them to the rear
of the list, we will
In the shrinker, we can safely remove an empty object (obj->mm.pages ==
NULL) after having discarded the pages because we are holding the
struct_mutex.
Signed-off-by: Chris Wilson
Reviewed-by: Tvrtko Ursulin
---
To flush all call_rcu() tasks (here from i915_gem_free_object()) we need
to call rcu_barrier() (not synchronize_rcu()). If we don't then we may
still have objects being freed as we continue to teardown the driver -
in particular, the recently released rings may race with the memory
manager
On Tue, Nov 01, 2016 at 08:39:14AM +, Tvrtko Ursulin wrote:
>
> On 31/10/2016 10:26, Chris Wilson wrote:
> >If we have a tiled object and an unknown CPU swizzle pattern, we pin the
> >pages to prevent the object from being swapped out (and us corrupting
> >the contents as we do not know the
On 31/10/2016 21:03, Chris Wilson wrote:
On Mon, Oct 31, 2016 at 05:35:50PM +, Tvrtko Ursulin wrote:
On 31/10/2016 10:26, Chris Wilson wrote:
Whilst waiting on a request, we may do so without holding any locks or
any guards beyond a reference to the request. In order to avoid taking
On 31/10/2016 10:26, Chris Wilson wrote:
With full-ppgtt one of the main bottlenecks is the lookup of the VMA
underneath the object. For execbuf there is merit in having a very fast
direct lookup of ctx:handle to the vma using a hashtree, but that still
leaves a large number of other lookups.
1 - 100 of 107 matches
Mail list logo