On Fri, 24 Nov 2017, Matthew Auld wrote:
> From: Joonas Lahtinen
Please CC the linux kernel mailinglist on patches related to x86. The
MAINTAINERS file says:
X86 ARCHITECTURE (32-BIT AND 64-BIT)
M: Thomas Gleixner
M: Ingo Molnar
== Series Details ==
Series: make stolen resource centric (rev3)
URL : https://patchwork.freedesktop.org/series/34256/
State : warning
== Summary ==
Test kms_frontbuffer_tracking:
Subgroup fbc-1p-primscrn-cur-indfb-draw-render:
skip -> PASS (shard-hsw)
Quoting Matthew Auld (2017-11-24 21:29:30)
> Keeps things consistent now that we make use of struct resource. This
> should keep us covered in case we ever get huge amounts of stolen
> memory.
>
> v2: bunch of missing conversions (Chris)
>
> Signed-off-by: Matthew Auld
>
Quoting Matthew Auld (2017-11-24 21:29:29)
> Signed-off-by: Matthew Auld
> Cc: Joonas Lahtinen
> Cc: Chris Wilson
> Cc: Paulo Zanoni
> ---
> diff --git
Quoting Matthew Auld (2017-11-24 21:29:28)
Always have a small explanation here.
"Now that we are using struct resource to track the stolen region, it is
more convenient if we track the reserved portion of that region in a
resource as well."
> v2: s/<= end + 1/< end/ (Chris)
>
> Signed-off-by:
Quoting Matthew Auld (2017-11-24 21:29:27)
> v2: check range_overflow when writing to 32b registers (Chris)
> pepper in some comments (Chris)
>
> Signed-off-by: Matthew Auld
> Cc: Joonas Lahtinen
> Cc: Chris Wilson
Quoting Matthew Auld (2017-11-24 21:29:26)
> In a later patch we introduce 'struct resource mappable', which means we
> will have both mappable.end and mappable_end, both meaning different
> things. Therefore to avoid any confusion prefer mappable_size for
> mappable_end, since it is effectively
Quoting Matthew Auld (2017-11-24 21:29:25)
> We duplicate the stolen discovery code in early-quirks and in i915,
> however now the stolen region is exported as a resource from
> early-quirks we can nuke the duplication.
>
> Signed-off-by: Matthew Auld
> Cc: Joonas
Quoting Matthew Auld (2017-11-24 21:29:22)
> From: Joonas Lahtinen
>
> In preparation for upcoming SKUs, allow more freedom in placement
> of the Intel graphics stolen memory by BIOS to full 64bit range.
>
> v2: export the stolen region as a resource
> fix
== Series Details ==
Series: make stolen resource centric (rev3)
URL : https://patchwork.freedesktop.org/series/34256/
State : success
== Summary ==
Series 34256v3 make stolen resource centric
https://patchwork.freedesktop.org/api/1.0/series/34256/revisions/3/mbox/
Test gem_exec_reloc:
Quoting Chris Wilson (2017-11-24 20:40:31)
> Quoting Tvrtko Ursulin (2017-11-24 17:13:31)
> > From: Tvrtko Ursulin
> >
> > Chris has discovered that RC6, RC6p and RC6pp counters are mutually
> > exclusive, and even that on some SNB SKUs you get RC6p increasing, and on
>
v2: check range_overflow when writing to 32b registers (Chris)
pepper in some comments (Chris)
Signed-off-by: Matthew Auld
Cc: Joonas Lahtinen
Cc: Chris Wilson
Cc: Paulo Zanoni
---
v2: s/<= end + 1/< end/ (Chris)
Signed-off-by: Matthew Auld
Cc: Joonas Lahtinen
Cc: Chris Wilson
Cc: Paulo Zanoni
---
drivers/gpu/drm/i915/i915_drv.h| 4
Signed-off-by: Matthew Auld
Cc: Joonas Lahtinen
Cc: Chris Wilson
Cc: Paulo Zanoni
---
drivers/gpu/drm/i915/gvt/gvt.h| 2 +-
drivers/gpu/drm/i915/i915_drv.c
We duplicate the stolen discovery code in early-quirks and in i915,
however now the stolen region is exported as a resource from
early-quirks we can nuke the duplication.
Signed-off-by: Matthew Auld
Cc: Joonas Lahtinen
Cc: Chris Wilson
In a later patch we introduce 'struct resource mappable', which means we
will have both mappable.end and mappable_end, both meaning different
things. Therefore to avoid any confusion prefer mappable_size for
mappable_end, since it is effectively the size of the resource.
Signed-off-by: Matthew
Keeps things consistent now that we make use of struct resource. This
should keep us covered in case we ever get huge amounts of stolen
memory.
v2: bunch of missing conversions (Chris)
Signed-off-by: Matthew Auld
Cc: Joonas Lahtinen
Cc:
Makes things much easier to follow.
Suggested-by: Ville Syrjälä
Signed-off-by: Matthew Auld
Cc: Joonas Lahtinen
Cc: Ville Syrjälä
Cc: Chris Wilson
From: Joonas Lahtinen
In preparation for upcoming SKUs, allow more freedom in placement
of the Intel graphics stolen memory by BIOS to full 64bit range.
v2: export the stolen region as a resource
fix u16 << 16 (Chris)
Signed-off-by: Joonas Lahtinen
Replace the magical +2, +9 etc. with +MB, which is far easier to read.
Suggested-by: Ville Syrjälä
Signed-off-by: Matthew Auld
Cc: Joonas Lahtinen
Cc: Ville Syrjälä
Cc: Chris
Continuation of Paulo' stolen series[1], addressing the feedback from Joonas and
Chris.
[1] https://patchwork.freedesktop.org/series/30923/
Joonas Lahtinen (1):
x86/early-quirks: Extend Intel graphics stolen memory placement to
64bit
Matthew Auld (8):
x86/early-quirks: replace the
Quoting Michal Wajdeczko (2017-11-24 17:02:39)
> We should be consistent on naming of similar definitions.
>
> Signed-off-by: Michal Wajdeczko
> Cc: Chris Wilson
> Cc: Joonas Lahtinen
> Cc: Sagar Arun Kamble
Quoting Tvrtko Ursulin (2017-11-24 17:13:31)
> From: Tvrtko Ursulin
>
> Chris has discovered that RC6, RC6p and RC6pp counters are mutually
> exclusive, and even that on some SNB SKUs you get RC6p increasing, and on
> the others RC6.
>
> Furthermore RC6p and RC6pp were
Quoting Tvrtko Ursulin (2017-11-24 17:16:18)
> From: Tvrtko Ursulin
>
> Since i915 PMU is removing separate RC6 counters and now aggregates all
> under a single one, catch up the test and intel-gpu-overlay with those
> changes.
>
> Signed-off-by: Tvrtko Ursulin
== Series Details ==
Series: intel/pmu: Catch-up with i915 RC6 aggregation changes
URL : https://patchwork.freedesktop.org/series/34371/
State : warning
== Summary ==
Test kms_frontbuffer_tracking:
Subgroup fbc-1p-pri-indfb-multidraw:
fail -> SKIP
== Series Details ==
Series: drm/i915/guc: Use consistent name for scratch register count
URL : https://patchwork.freedesktop.org/series/34369/
State : warning
== Summary ==
Test kms_frontbuffer_tracking:
Subgroup fbc-1p-offscren-pri-shrfb-draw-blt:
pass -> FAIL
== Series Details ==
Series: make stolen resource centric (rev2)
URL : https://patchwork.freedesktop.org/series/34256/
State : failure
== Summary ==
Test kms_frontbuffer_tracking:
Subgroup fbc-1p-offscren-pri-shrfb-draw-blt:
pass -> FAIL (shard-snb)
Quoting Matthew Auld (2017-11-24 17:56:34)
> On 24 November 2017 at 16:54, Chris Wilson wrote:
> > Quoting Matthew Auld (2017-11-24 16:42:57)
> >> Keeps things consistent now that we make use of struct resource. This
> >> should keep us covered in case we ever get huge
Quoting Chris Wilson (2017-11-24 18:19:03)
> Quoting Tvrtko Ursulin (2017-11-24 17:13:31)
> > From: Tvrtko Ursulin
> >
> > Chris has discovered that RC6, RC6p and RC6pp counters are mutually
> > exclusive, and even that on some SNB SKUs you get RC6p increasing, and on
>
== Series Details ==
Series: intel/pmu: Catch-up with i915 RC6 aggregation changes
URL : https://patchwork.freedesktop.org/series/34371/
State : success
== Summary ==
IGT patchset tested on top of latest successful build
c6577473df7117b7a6e030605df1e28cd0e55708 lib/igt_core: Move write_stderr
== Series Details ==
Series: series starting with [1/2] igt: Remove Android support
URL : https://patchwork.freedesktop.org/series/34365/
State : success
== Summary ==
Test kms_frontbuffer_tracking:
Subgroup fbc-1p-offscren-pri-shrfb-draw-blt:
fail -> PASS
Quoting Tvrtko Ursulin (2017-11-24 17:13:31)
> From: Tvrtko Ursulin
>
> Chris has discovered that RC6, RC6p and RC6pp counters are mutually
> exclusive, and even that on some SNB SKUs you get RC6p increasing, and on
> the others RC6.
>
> Furthermore RC6p and RC6pp were
Quoting Tvrtko Ursulin (2017-11-24 17:16:18)
> From: Tvrtko Ursulin
>
> Since i915 PMU is removing separate RC6 counters and now aggregates all
> under a single one, catch up the test and intel-gpu-overlay with those
> changes.
First run failed
(perf_pmu:1928)
== Series Details ==
Series: drm/i915/pmu: Aggregate all RC6 states into one counter
URL : https://patchwork.freedesktop.org/series/34370/
State : failure
== Summary ==
Series 34370v1 drm/i915/pmu: Aggregate all RC6 states into one counter
On 24 November 2017 at 16:54, Chris Wilson wrote:
> Quoting Matthew Auld (2017-11-24 16:42:57)
>> Keeps things consistent now that we make use of struct resource. This
>> should keep us covered in case we ever get huge amounts of stolen
>> memory.
>>
>> v2: bunch of
== Series Details ==
Series: drm/i915: Expose the busyspin durations for i915_wait_request
URL : https://patchwork.freedesktop.org/series/34364/
State : success
== Summary ==
Test gem_busy:
Subgroup close-race:
fail -> PASS (shard-snb) fdo#103829
Test
== Series Details ==
Series: drm/i915/guc: Use consistent name for scratch register count
URL : https://patchwork.freedesktop.org/series/34369/
State : success
== Summary ==
Series 34369v1 drm/i915/guc: Use consistent name for scratch register count
From: Tvrtko Ursulin
Since i915 PMU is removing separate RC6 counters and now aggregates all
under a single one, catch up the test and intel-gpu-overlay with those
changes.
Signed-off-by: Tvrtko Ursulin
---
lib/igt_perf.h | 4 +---
From: Tvrtko Ursulin
Chris has discovered that RC6, RC6p and RC6pp counters are mutually
exclusive, and even that on some SNB SKUs you get RC6p increasing, and on
the others RC6.
Furthermore RC6p and RC6pp were only present starting from GEN6 until,
GEN7, not including
Quoting Patchwork (2017-11-24 17:02:54)
> == Series Details ==
>
> Series: series starting with [1/2] drm/i915/guc: Advance over port[0] if set
> and not preempting
> URL : https://patchwork.freedesktop.org/series/34361/
> State : success
>
> == Summary ==
>
> Test kms_frontbuffer_tracking:
On Fri, Nov 24, 2017 at 04:42:51PM +, Matthew Auld wrote:
> Replace the magical +2, +9 etc. with +MB, which is far easier to read.
>
> Suggested-by: Ville Syrjälä
> Signed-off-by: Matthew Auld
> Cc: Joonas Lahtinen
== Series Details ==
Series: lib/igt_core: Move write_stderr out of LIBUNWIND ifdef
URL : https://patchwork.freedesktop.org/series/34349/
State : warning
== Summary ==
Test kms_flip:
Subgroup wf_vblank-vs-modeset-interruptible:
pass -> SKIP (shard-hsw)
On Thu, Nov 23, 2017 at 09:04:48PM +0200, Ville Syrjala wrote:
> From: Ville Syrjälä
>
> In order to guarantee that pipe_src_w/h matches the user mode h/vdisplay
> we must not adjust pipe_src_w to accommodate double wide/dual link.
> Instead just reject the mode
== Series Details ==
Series: series starting with [1/2] drm/i915/guc: Advance over port[0] if set
and not preempting
URL : https://patchwork.freedesktop.org/series/34361/
State : success
== Summary ==
Test kms_frontbuffer_tracking:
Subgroup fbc-1p-offscren-pri-shrfb-draw-render:
We should be consistent on naming of similar definitions.
Signed-off-by: Michal Wajdeczko
Cc: Chris Wilson
Cc: Joonas Lahtinen
Cc: Sagar Arun Kamble
---
== Series Details ==
Series: make stolen resource centric (rev2)
URL : https://patchwork.freedesktop.org/series/34256/
State : success
== Summary ==
Series 34256v2 make stolen resource centric
https://patchwork.freedesktop.org/api/1.0/series/34256/revisions/2/mbox/
fi-bdw-5557u total:289
On Thu, Nov 23, 2017 at 09:05:02PM +0200, Ville Syrjala wrote:
> From: Ville Syrjälä
>
> Move the plane clip rectangle handling into
> drm_atomic_helper_check_plane_state(). Drivers no longer
> have to worry about such mundane details.
>
> Cc: Laurent Pinchart
On 24 November 2017 at 11:27, Joonas Lahtinen
wrote:
> We seem to be missing some W/A for 2M pages and are getting
> a hit on raw read bandwidths (even 30%) even though the write
> bandwidths improve (even 10%).
>
> For now, disable THP, which is our only
Quoting Tvrtko Ursulin (2017-11-24 16:44:21)
>
> On 24/11/2017 14:54, Chris Wilson wrote:
> > An interesting discussion regarding "hybrid interrupt polling" for NVMe
> > came to the conclusion that the ideal busyspin before sleeping was half
> > of the expected request latency (and better if it
Quoting Matthew Auld (2017-11-24 16:42:57)
> Keeps things consistent now that we make use of struct resource. This
> should keep us covered in case we ever get huge amounts of stolen
> memory.
>
> v2: bunch of missing conversions (Chris)
>
> Signed-off-by: Matthew Auld
>
On 24/11/2017 14:54, Chris Wilson wrote:
An interesting discussion regarding "hybrid interrupt polling" for NVMe
came to the conclusion that the ideal busyspin before sleeping was half
of the expected request latency (and better if it was already halfway
through that request). This suggested
v2: s/<= end + 1/< end/ (Chris)
Signed-off-by: Matthew Auld
Cc: Joonas Lahtinen
Cc: Chris Wilson
Cc: Paulo Zanoni
---
drivers/gpu/drm/i915/i915_drv.h| 4
Replace the magical +2, +9 etc. with +MB, which is far easier to read.
Suggested-by: Ville Syrjälä
Signed-off-by: Matthew Auld
Cc: Joonas Lahtinen
Cc: Ville Syrjälä
Cc: Chris
Keeps things consistent now that we make use of struct resource. This
should keep us covered in case we ever get huge amounts of stolen
memory.
v2: bunch of missing conversions (Chris)
Signed-off-by: Matthew Auld
Cc: Joonas Lahtinen
Cc:
v2: check range_overflow when writing to 32b registers (Chris)
pepper in some comments (Chris)
Signed-off-by: Matthew Auld
Cc: Joonas Lahtinen
Cc: Chris Wilson
Cc: Paulo Zanoni
---
We duplicate the stolen discovery code in early-quirks and in i915,
however now the stolen region is exported as a resource from
early-quirks we can nuke the duplication.
Signed-off-by: Matthew Auld
Cc: Joonas Lahtinen
Cc: Chris Wilson
From: Joonas Lahtinen
In preparation for upcoming SKUs, allow more freedom in placement
of the Intel graphics stolen memory by BIOS to full 64bit range.
v2: export the stolen region as a resource
fix u16 << 16 (Chris)
Signed-off-by: Joonas Lahtinen
Signed-off-by: Matthew Auld
Cc: Joonas Lahtinen
Cc: Chris Wilson
Cc: Paulo Zanoni
---
drivers/gpu/drm/i915/gvt/gvt.h| 2 +-
drivers/gpu/drm/i915/i915_drv.c
In a later patch we introduce 'struct resource mappable', which means we
will have both mappable.end and mappable_end, both meaning different
things. Therefore to avoid any confusion prefer mappable_size for
mappable_end, since it is effectively the size of the resource.
Signed-off-by: Matthew
Continuation of Paulo' stolen series[1], addressing the feedback from Joonas and
Chris.
[1] https://patchwork.freedesktop.org/series/30923/
Joonas Lahtinen (1):
x86/early-quirks: Extend Intel graphics stolen memory placement to
64bit
Matthew Auld (7):
x86/early-quirks: replace the
== Series Details ==
Series: drm/i915/guc: Tidy ELSP port assignment
URL : https://patchwork.freedesktop.org/series/34358/
State : success
== Summary ==
Test kms_flip:
Subgroup modeset-vs-vblank-race-interruptible:
fail -> PASS (shard-hsw) fdo#103060
Test
== Series Details ==
Series: series starting with [CI,1/3] drm/i915/guc: Tidy ELSP port assignment
URL : https://patchwork.freedesktop.org/series/34360/
State : warning
== Summary ==
Test gem_busy:
Subgroup close-race:
pass -> FAIL (shard-snb) fdo#103829
On Fri, Nov 24, 2017 at 02:55:28PM +0200, Ville Syrjälä wrote:
> On Thu, Nov 23, 2017 at 08:50:41PM +, Chris Wilson wrote:
> > Quoting Ville Syrjala (2017-11-23 19:41:55)
> > > From: Ville Syrjälä
> > >
> > > We're supposed to examine msgs[i] and msgs[i+1] to
== Series Details ==
Series: series starting with [1/2] igt: Remove Android support
URL : https://patchwork.freedesktop.org/series/34365/
State : success
== Summary ==
IGT patchset tested on top of latest successful build
c6577473df7117b7a6e030605df1e28cd0e55708 lib/igt_core: Move
On Fri, Nov 24, 2017 at 04:35:10PM +0100, Maarten Lankhorst wrote:
> Op 24-11-17 om 16:03 schreef Daniel Vetter:
> > On Thu, Nov 23, 2017 at 01:26:14PM +0100, Maarten Lankhorst wrote:
> >> This was implemented correctly only on the atomic ioctl before, but
> >> it should really be working on all 3
Op 24-11-17 om 16:03 schreef Daniel Vetter:
> On Thu, Nov 23, 2017 at 01:26:14PM +0100, Maarten Lankhorst wrote:
>> This was implemented correctly only on the atomic ioctl before, but
>> it should really be working on all 3 ioctl's involved, so ensure we
>> always set crtc_id correctly with a
On Fri, Nov 24, 2017 at 04:08:28PM +0200, Ville Syrjälä wrote:
> On Fri, Nov 24, 2017 at 11:59:45AM +, Liviu Dudau wrote:
> > On Thu, Nov 23, 2017 at 09:05:02PM +0200, Ville Syrjala wrote:
> > > From: Ville Syrjälä
> >
> > Hi Ville,
> >
> > >
> > > Move the
== Series Details ==
Series: drm/i915: Expose the busyspin durations for i915_wait_request
URL : https://patchwork.freedesktop.org/series/34364/
State : success
== Summary ==
Series 34364v1 drm/i915: Expose the busyspin durations for i915_wait_request
This reverts commit d7d3f4e87b827152f00bdf89a67871736672b492
and gets rid of the config option from the meson.build.
It was needed only for the Android support.
Signed-off-by: Arkadiusz Hiler
---
configure.ac | 6 +-
lib/igt_aux.c | 35
This patch gets rid of the Android support, deleting all the hacks and
moving code around to the places it belongs.
Android build is not really maintained properly and rots rather fast.
With recent push for Meson here and Android going for Soong it will only
accelerate.
It's a good time to drop
On Fri, Nov 24, 2017 at 01:37:44PM +, Chris Wilson wrote:
> Our execlist emulation is intended to only use a maximum of 2 ports per
> engine, so as to not overflow the wq. (By knowing the limits, we can
> avoid having to handle the wq exhaustion.) However, upon adding
> preemption, we lost the
On Thu, Nov 23, 2017 at 02:13:08PM +0200, Jani Nikula wrote:
> I'm juggling too many things, and drm-misc maintenance is one that I
> keep dropping on the floor. Admit reality and remove myself as
> maintainer. This still leaves us with a nice team of three who are
> actually doing the drm-misc
Quoting Chris Wilson (2017-11-24 12:12:33)
> Quoting Joonas Lahtinen (2017-11-24 11:27:50)
> > We seem to be missing some W/A for 2M pages and are getting
> > a hit on raw read bandwidths (even 30%) even though the write
> > bandwidths improve (even 10%).
> >
> > For now, disable THP, which is
On Thu, Nov 23, 2017 at 01:26:14PM +0100, Maarten Lankhorst wrote:
> This was implemented correctly only on the atomic ioctl before, but
> it should really be working on all 3 ioctl's involved, so ensure we
> always set crtc_id correctly with a testcase.
>
> Signed-off-by: Maarten Lankhorst
== Series Details ==
Series: drm/i915: Disable THP until we have a read BW W/A
URL : https://patchwork.freedesktop.org/series/34355/
State : warning
== Summary ==
Warning: bzip CI_DRM_3381/shard-glkb6/results8.json.bz2 wasn't in correct JSON
format
Test perf:
Subgroup polling:
An interesting discussion regarding "hybrid interrupt polling" for NVMe
came to the conclusion that the ideal busyspin before sleeping was half
of the expected request latency (and better if it was already halfway
through that request). This suggested that we too should look again at
our tradeoff
== Series Details ==
Series: series starting with [1/2] drm/i915: Move engine->needs_cmd_parser to
engine->flags
URL : https://patchwork.freedesktop.org/series/34354/
State : success
== Summary ==
Warning: bzip CI_DRM_3381/shard-glkb6/results8.json.bz2 wasn't in correct JSON
format
Test
== Series Details ==
Series: lib/igt_core: Move write_stderr out of LIBUNWIND ifdef
URL : https://patchwork.freedesktop.org/series/34349/
State : success
== Summary ==
IGT patchset tested on top of latest successful build
2a1c1c099113022f091f9e9585be88c6c9cefec5 igt/perf_pmu: Recalibrate
== Series Details ==
Series: drm/i915/guc: Rename i915_guc_reg.h to intel_guc_reg.h (rev2)
URL : https://patchwork.freedesktop.org/series/34319/
State : success
== Summary ==
Warning: bzip CI_DRM_3381/shard-glkb6/results8.json.bz2 wasn't in correct JSON
format
Test kms_flip:
Subgroup
On Fri, Nov 24, 2017 at 03:32:22PM +0100, Philipp Zabel wrote:
> On Thu, 2017-11-23 at 21:04 +0200, Ville Syrjala wrote:
> > From: Ville Syrjälä
> >
> > Use drm_mode_get_hv_timing() to fill out the plane clip rectangle.
> >
> > No functional changes as the code
On Thu, 2017-11-23 at 21:04 +0200, Ville Syrjala wrote:
> From: Ville Syrjälä
>
> Use drm_mode_get_hv_timing() to fill out the plane clip rectangle.
>
> No functional changes as the code already uses crtc_state->mode
> to populate the clip, which is also what
On Thu, 2017-11-23 at 21:04 +0200, Ville Syrjala wrote:
> From: Ville Syrjälä
>
> Use drm_mode_get_hv_timing() to fill out the plane clip rectangle.
>
> Note that this replaces crtc_state->adjusted_mode usage with
> crtc_state->mode. The latter is the correct
== Series Details ==
Series: series starting with [1/2] drm/i915/guc: Advance over port[0] if set
and not preempting
URL : https://patchwork.freedesktop.org/series/34361/
State : success
== Summary ==
Series 34361v1 series starting with [1/2] drm/i915/guc: Advance over port[0] if
set and
Quoting Michał Winiarski (2017-11-24 12:37:56)
> Since we see the effects for GuC preeption, let's gather some evidence.
>
> (SKL)
> intel_guc_send_mmio latency: 100 rounds of gem_exec_latency --r '*-preemption'
>
> drm-tip:
> usecs : count distribution
> 0 -> 1
On Fri, Nov 24, 2017 at 11:59:45AM +, Liviu Dudau wrote:
> On Thu, Nov 23, 2017 at 09:05:02PM +0200, Ville Syrjala wrote:
> > From: Ville Syrjälä
>
> Hi Ville,
>
> >
> > Move the plane clip rectangle handling into
> > drm_atomic_helper_check_plane_state().
> -Original Message-
> From: Ville Syrjälä [mailto:ville.syrj...@linux.intel.com]
> Sent: Friday, November 24, 2017 3:50 PM
> To: Rantala, Valtteri
> Cc: intel-gfx@lists.freedesktop.org
> Subject: Re: [Intel-gfx] [PATCH v2] drm/i915/glk: Apply
>
On Fri, Nov 24, 2017 at 09:30:17AM +0200, Joonas Lahtinen wrote:
> @@ -426,14 +426,14 @@ static size_t __init chv_stolen_size(int num, int slot,
> int func)
>* 0x17 to 0x1d: 4MB increments start at 36MB
>*/
> if (gms < 0x11)
> - return (size_t)gms * MB(32);
> +
Chris Wilson writes:
> Quoting Chris Wilson (2017-11-24 13:37:44)
>> Our execlist emulation is intended to only use a maximum of 2 ports per
>> engine, so as to not overflow the wq. (By knowing the limits, we can
>> avoid having to handle the wq exhaustion.) However,
On Fri, Nov 24, 2017 at 10:37:12AM +0200, Valtteri Rantala wrote:
> Testing the texture read performance shows that the same tuning for
> the SQ credits is needed on GLK as on BXT/APL. This has been also
> confirmed by Altug from the HW team.
>
> V2: Rebase
>
> Signed-off-by: Valtteri Rantala
Chris Wilson writes:
> Our execlist emulation is intended to only use a maximum of 2 ports per
> engine, so as to not overflow the wq. (By knowing the limits, we can
> avoid having to handle the wq exhaustion.) However, upon adding
> preemption, we lost the skip over
Quoting Chris Wilson (2017-11-24 13:37:44)
> Our execlist emulation is intended to only use a maximum of 2 ports per
> engine, so as to not overflow the wq. (By knowing the limits, we can
> avoid having to handle the wq exhaustion.) However, upon adding
> preemption, we lost the skip over the
On Thu, Nov 23, 2017 at 08:47:45PM +, Chris Wilson wrote:
> Ironlake does support being able to saving and reloading context specific
> registers between contexts, providing isolation of the basic GPU state
> (as programmable by userspace). This allows userspace to assume that the
> GPU
On Thu, Nov 23, 2017 at 09:04:59PM +0200, Ville Syrjala wrote:
> From: Ville Syrjälä
>
> Use drm_mode_get_hv_timing() to fill out the plane clip rectangle.
>
> No functional changes as the code already uses crtc_state->mode
> to populate the clip, which is also
On Thu, Nov 23, 2017 at 09:04:47PM +0200, Ville Syrjala wrote:
> From: Ville Syrjälä
>
> This series first unifies all users of drm_atomic_helper_check_plane_state()
> to populate the clip rectangle with drm_mode_get_hv_timing(), and once
> everything is unified
Our execlist emulation is intended to only use a maximum of 2 ports per
engine, so as to not overflow the wq. (By knowing the limits, we can
avoid having to handle the wq exhaustion.) However, upon adding
preemption, we lost the skip over the first port if set for the
non-preemption path. Restore
From: Michał Winiarski
Also:
Revert "drm/i915/guc: Assert that we switch between known ggtt->invalidate
functions"
This reverts commit 04f7b24eccdfae680a36e9825fe0d61dcd5ed528.
---
drivers/gpu/drm/i915/i915_gem_gtt.c | 8 ++--
drivers/gpu/drm/i915/i915_params.h
== Series Details ==
Series: series starting with [CI,1/3] drm/i915/guc: Tidy ELSP port assignment
URL : https://patchwork.freedesktop.org/series/34360/
State : success
== Summary ==
Series 34360v1 series starting with [CI,1/3] drm/i915/guc: Tidy ELSP port
assignment
On Fri, Nov 24, 2017 at 02:26:09PM +0530, Sharma, Shashank wrote:
> Regards
>
> Shashank
>
>
> On 11/17/2017 6:19 PM, Ville Syrjälä wrote:
> > On Fri, Nov 17, 2017 at 05:50:11PM +0530, Sharma, Shashank wrote:
> >> Regards
> >>
> >> Shashank
> >>
> >>
> >> On 11/17/2017 5:05 PM, Ville Syrjälä
On 24/11/2017 10:38, Patchwork wrote:
== Series Details ==
Series: drm/i915/pmu: Stop averaging with the previous sample
URL : https://patchwork.freedesktop.org/series/34346/
State : success
== Summary ==
Series 34346v1 drm/i915/pmu: Stop averaging with the previous sample
Quoting Chris Wilson (2017-11-24 12:29:16)
> Since we know that the port is empty, we do not need to extract the
> count from the old request it and copy it over to the new request, or
> attempt to unref the NULL old request pointer.
>
> Signed-off-by: Chris Wilson
>
1 - 100 of 176 matches
Mail list logo