attached patch removed function call + comment.
Pierre-Eric
From 4596e9040bde510fd974d500af69070341871b46 Mon Sep 17 00:00:00 2001
From: Pierre-Eric Pelloux-Prayer pell...@gmail.com
Date: Fri, 10 Jun 2011 15:45:01 +0200
Subject: [PATCH 2/2] r600g: remove useless call to u_upload_flush
No regressions
==
dst, input values are not overwritten before being used).
I've sent the accompagnying test case to piglit ML.
Pierre-Eric
From f0cd600074ef0a086742a4a0932b8d5880fc87b3 Mon Sep 17 00:00:00 2001
From: Pierre-Eric Pelloux-Prayer pell...@gmail.com
Date: Mon, 20 Jun 2011 11:25:14 +0200
Subject
u_upload_flush() is called each time r600_flush() is used.
There is a comment specifically saying :
/* XXX This shouldn't be really necessary, but removing it breaks some
tests.
* Needless buffer reallocations may significantly increase memory
consumption,
* so getting
Hi,
the attached patch should be an improvement over Vadim Girlin's patch
fixing LIT instruction for r600g (commit
2fe39b46e73aea37152777fe11d489e0b1bc3f92).
Instructions used in tgsi_lit have been reordered to always write to a
dst channel after the same channel in src has been read (so if src
Hi Christian,
Sorry I missed your original email. I just tested the patch and it works fine,
so feel free to add my Tested-by tag.
Pierre-Eric
On 09/03/2020 12:49, Christian König wrote:
> Pierre-eric, just a gentle ping on this? Could I get a tested-by?
>
> Ray can you ack or even review
Hi Bas,
2 small typos you may want to fix:
On 04/09/2020 18:07, Bas Nieuwenhuizen wrote:
> This adds modifiers for GFX9+ AMD GPUs.
>
> As the modifiers need a lot of parameters I split things out in
> getters and setters.
> - Advantage: simplifies the code a lot
> - Disadvantage: Makes it
Hi Daniel,
On 04/06/2020 10:12, Daniel Vetter wrote:
[...]
> @@ -6910,7 +6910,11 @@ static void amdgpu_dm_commit_planes(struct
> drm_atomic_state *state,
>* explicitly on fences instead
>* and in general should be called for
>* blocking commit to
d other drivers and it doesn't look like they end up triggering
> this case so I think this is safe to relax.
This patch fixes https://gitlab.freedesktop.org/drm/amd/-/issues/1379:
Tested-by: Pierre-Eric Pelloux-Prayer
Thanks!
P-E
>
> Signed-off-by: Bas Nieuwenhuizen
> Fixes: 816853f9dc40
On 09/03/2022 11:24, Christian König wrote:
> Am 09.03.22 um 11:10 schrieb Simon Ser:
>> On Wednesday, March 9th, 2022 at 10:56, Pierre-Eric Pelloux-Prayer
>> wrote:
>>
>>> Would it be possible to include the app parameters as well?
>>
Hi Shashank,
On 08/03/2022 19:04, Shashank Sharma wrote:
> From: Shashank Sharma
>
> This patch adds a new sysfs event, which will indicate
> the userland about a GPU reset, and can also provide
> some information like:
> - process ID of the process involved with the GPU reset
> - process name
Le 14/02/2024 à 13:09, Christian König a écrit :
Am 13.02.24 um 16:50 schrieb Pierre-Eric Pelloux-Prayer:
amdgpu_cs_ioctl already exists but serves a different
purpose.
amdgpu_cs_ioctl2 marks the beginning of the kernel processing of
the ioctl which is useful for tools to map which events
Le 14/02/2024 à 16:10, Steven Rostedt a écrit :
On Wed, 14 Feb 2024 13:00:16 +0100
Christian König wrote:
+DEFINE_EVENT(dma_fence_from, dma_fence_sync_to,
For a single event you should probably use TRACE_EVENT() instead of
declaring a class. A class is only used if you have multiple events
To inform tools about the relationship between the fences.
Signed-off-by: Pierre-Eric Pelloux-Prayer
---
drivers/dma-buf/dma-fence-chain.c | 4
1 file changed, 4 insertions(+)
diff --git a/drivers/dma-buf/dma-fence-chain.c
b/drivers/dma-buf/dma-fence-chain.c
index 9663ba1bb6ac
This new event can be used to trace where a given dma_fence is added
as a dependency of some other work.
I plan to use it in amdgpu.
Signed-off-by: Pierre-Eric Pelloux-Prayer
---
drivers/dma-buf/dma-fence.c | 1 +
include/trace/events/dma_fence.h | 34
2
of amdgpu_fence_sync.
Signed-off-by: Pierre-Eric Pelloux-Prayer
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 8
drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 14 +++---
drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c | 8 +++-
drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c
/patch/20240117184329.479554-1-pierre-eric.pelloux-pra...@amd.com/
Changes from V1:
* uses trace_dma_fence_sync_to from dma-fence-chain.c
* new amdgpu events
* new drm plane commit event
Pierre-Eric Pelloux-Prayer (6):
tracing, dma-buf: add a trace_dma_fence_sync_to event
dma-buf/fence-chain
[018] d.h3. 2571.915968: drm_vblank_event: crtc=0,
seq=153862, time=2571916377180, high-prec=true
Signed-off-by: Pierre-Eric Pelloux-Prayer
---
drivers/gpu/drm/drm_atomic_uapi.c | 19 +++
drivers/gpu/drm/drm_trace.h | 28 ++--
2 files changed, 45
Useful to identify why sdma jobs are submitted.
Signed-off-by: Pierre-Eric Pelloux-Prayer
---
drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 2 ++
drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h | 16
2 files changed, 18 insertions(+)
diff --git a/drivers/gpu/drm/amd/amdgpu
-by: Pierre-Eric Pelloux-Prayer
---
drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c| 2 ++
drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h | 12
2 files changed, 14 insertions(+)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
index 6830892383c3
=2571916093743, high-prec=true
-0 [018] d.h3. 2571.915968: drm_vblank_event: crtc=0,
seq=153862, time=2571916377180, high-prec=true
v2: fix unchecked memory allocation
Signed-off-by: Pierre-Eric Pelloux-Prayer
---
drivers/gpu/drm/drm_atomic_uapi.c | 21 +
drivers/gpu/drm
Useful to identify why sdma jobs are submitted.
v2: moved from amdgpu_bo_create to amdgpu_bo_fill
Signed-off-by: Pierre-Eric Pelloux-Prayer
---
drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h | 18 ++
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 2 ++
2 files changed, 20 insertions
to amdgpu_cs_start
Signed-off-by: Pierre-Eric Pelloux-Prayer
---
drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c| 2 ++
drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h | 12
2 files changed, 14 insertions(+)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
b/drivers/gpu/drm/amd/amdgpu
drm_sched_job_add_dependency adds dependencies so use the new
trace event.
Signed-off-by: Pierre-Eric Pelloux-Prayer
---
drivers/gpu/drm/scheduler/sched_main.c | 4
1 file changed, 4 insertions(+)
diff --git a/drivers/gpu/drm/scheduler/sched_main.c
b/drivers/gpu/drm/scheduler
.
With work queues, the source is now something like: "kworker/u64:0-15248"
(and the PID isn't stable), so the "timeline=gfx_0.0.0" attribute
isn't enough in multi-GPU setups.
This commit adds a dev=devname attribute to resolve this issue.
Signed-off-by: Pierre-Eric Pelloux-
attribute to the trace_amdgpu_sched_run_job event
* addressed review comments
Pierre-Eric Pelloux-Prayer (8):
tracing, dma-buf: add a trace_dma_fence_sync_to event
dma-buf/fence-chain: use trace_dma_fence_sync_to
amdgpu: use trace_dma_fence_sync_to in amdgpu_fence_sync
drm/amdgpu: add
This new event can be used to trace where a given dma_fence is added
as a dependency of some other work.
I plan to use it in amdgpu.
Signed-off-by: Pierre-Eric Pelloux-Prayer
---
drivers/dma-buf/dma-fence.c | 1 +
include/trace/events/dma_fence.h | 27 +++
2 files
issue
I have prototype code using this in UMR, as can be see here:
https://gitlab.freedesktop.org/tomstdenis/umr/-/merge_requests/37
v2: add a macro since every caller passes __func__ as the reason parameter
Signed-off-by: Pierre-Eric Pelloux-Prayer
---
drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c
To inform tools about the relationship between the fences.
Signed-off-by: Pierre-Eric Pelloux-Prayer
---
drivers/dma-buf/dma-fence-chain.c | 4
1 file changed, 4 insertions(+)
diff --git a/drivers/dma-buf/dma-fence-chain.c
b/drivers/dma-buf/dma-fence-chain.c
index 9663ba1bb6ac
of amdgpu_fence_sync.
Signed-off-by: Pierre-Eric Pelloux-Prayer
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 8
drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 14 +++---
drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c | 8 +++-
drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c
This new event can be used to trace where a given dma_fence is added
as a dependency of some other work.
I plan to use it in amdgpu.
Signed-off-by: Pierre-Eric Pelloux-Prayer
---
drivers/dma-buf/dma-fence.c | 1 +
include/trace/events/dma_fence.h | 34
2
Le 16/02/2024 à 17:24, Ville Syrjälä a écrit :
On Fri, Feb 16, 2024 at 04:09:55PM +0100, Pierre-Eric Pelloux-Prayer wrote:
With this and the dma_fence_used_as_dependency event, a tool can draw the
relationship between the compositing draw, the atomic commit, and vblank.
An example on a 2
-by: Pierre-Eric Pelloux-Prayer
---
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 2 +-
drivers/gpu/drm/etnaviv/etnaviv_sched.c| 2 +-
drivers/gpu/drm/imagination/pvr_queue.c| 2 +-
drivers/gpu/drm/lima/lima_sched.c | 2 +-
drivers/gpu/drm/msm/msm_ringbuffer.c | 2 +-
drivers
everything is vertically aligned to 'TP_fast_assign('. How to best fix it?
WARNING: Statements should start on a tabstop
#82: FILE: drivers/gpu/drm/scheduler/gpu_scheduler_trace.h:80:
+unsigned long idx;
v2: https://lists.freedesktop.org/archives/dri-devel/2024-February/441341.html
P
the same base event class,
the caller of trace_drm_run_job have to pass a dep_count of 0 (which
is ignored because dependencies are only considered at submit time).
Signed-off-by: Pierre-Eric Pelloux-Prayer
---
.../gpu/drm/scheduler/gpu_scheduler_trace.h | 58 ---
drivers/gpu/drm
the fence's context is already an
identifier of the job owner.
Signed-off-by: Pierre-Eric Pelloux-Prayer
---
.../gpu/drm/scheduler/gpu_scheduler_trace.h | 27 ++-
1 file changed, 15 insertions(+), 12 deletions(-)
diff --git a/drivers/gpu/drm/scheduler/gpu_scheduler_trace.
not
unique).
Signed-off-by: Pierre-Eric Pelloux-Prayer
---
drivers/gpu/drm/scheduler/gpu_scheduler_trace.h | 13 +++--
1 file changed, 11 insertions(+), 2 deletions(-)
diff --git a/drivers/gpu/drm/scheduler/gpu_scheduler_trace.h
b/drivers/gpu/drm/scheduler/gpu_scheduler_trace.h
index
Hi,
Le 06/06/2024 à 15:19, Steven Rostedt a écrit :
On Thu, 6 Jun 2024 15:06:24 +0200
Pierre-Eric Pelloux-Prayer wrote:
Print identifiers instead of pointers:
* "fence=%p" is replaced by "fence=(context:%llu, seqno:%lld)" to have a
coherent way to print the fence. A possi
Hi,
Le 06/06/2024 à 15:18, Christian König a écrit :
Am 06.06.24 um 15:06 schrieb Pierre-Eric Pelloux-Prayer:
When tracing is enabled, being able to identify which device is sending
events is useful; for this the next commit will extend events to include
drm_device::primary::index
v3: https://lists.freedesktop.org/archives/dri-devel/2024-June/456792.html
Changes since v3:
* trace device name instead of drm_device primary index
* no pointer deref in the TP_printk anymore. Instead the fence context/seqno
are saved in TP_fast_assign
Pierre-Eric Pelloux-Prayer (3):
drm
Until the switch from kthread to workqueue, a userspace application could
determine the source device from the pid of the thread sending the event.
With workqueues this is not possible anymore, so the event needs to contain
the dev_name() to identify the device.
Signed-off-by: Pierre-Eric
the fence's context is already an
identifier of the job owner.
For drm_sched_job_wait_dep, we also print the hardware exec context of
the fence that's initiating the wait (the scheduled fence ctx is not
relevant here, since it's not traced in other events).
Signed-off-by: Pierre-Eric Pelloux-Pray
the same base event class,
the caller of trace_drm_run_job have to pass a dep_count of 0 (which
is ignored because dependencies are only considered at submit time).
Signed-off-by: Pierre-Eric Pelloux-Prayer
---
.../gpu/drm/scheduler/gpu_scheduler_trace.h | 59 ---
drivers/gpu/drm
Hi,
Le 10/06/2024 à 18:31, Daniel Vetter a écrit :
> On Mon, Jun 10, 2024 at 03:26:53PM +0200, Pierre-Eric Pelloux-Prayer
wrote:
>> v3:
https://lists.freedesktop.org/archives/dri-devel/2024-June/456792.html
>>
>> Changes since v3:
>> * trace device name instead
43 matches
Mail list logo
