entity should not keep copy and maintain sched list for
itself.
Signed-off-by: Nirmoy Das
Reviewed-by: Christian König
---
drivers/gpu/drm/scheduler/sched_entity.c | 19 ---
1 file changed, 4 insertions(+), 15 deletions(-)
diff --git a/drivers/gpu/drm/scheduler/sched_entity.c
entity should not keep copy and maintain sched list for
itself.
Signed-off-by: Nirmoy Das
Reviewed-by: Christian König
---
drivers/gpu/drm/scheduler/sched_entity.c | 19 ---
1 file changed, 4 insertions(+), 15 deletions(-)
diff --git a/drivers/gpu/drm/scheduler/sched_entity.c
Am 10.12.19 um 19:17 schrieb Nirmoy Das:
entity should not keep copy and maintain sched list for
itself.
Signed-off-by: Nirmoy Das
Reviewed-by: Christian König
---
drivers/gpu/drm/scheduler/sched_entity.c | 19 ---
1 file changed, 4 insertions(+), 15 deletions(-)
diff
On 12/10/19 6:32 PM, Christian König wrote:
Maybe make this "num_sched_list > 1 ? sched_list : NULL" to avoid
accidentally dereferencing a stale pointer to the stack.
Do you mean "num_sched_list >= 1 ? sched_list : NULL"
No, the entity->sched_list field should be NULL when
entity should not keep copy and maintain sched list for
itself.
Signed-off-by: Nirmoy Das
---
drivers/gpu/drm/scheduler/sched_entity.c | 19 ---
1 file changed, 4 insertions(+), 15 deletions(-)
diff --git a/drivers/gpu/drm/scheduler/sched_entity.c
Am 10.12.19 um 16:08 schrieb Nirmoy:
I think amdgpu_ctx_init() should check for num_scheds and not call
drm_sched_entity_init()
if its zero.
Ah, that's where that came from. No that is intentionally this way, but
see below.
On 12/10/19 3:47 PM, Nirmoy wrote:
On 12/10/19 2:00 PM,
entity should not keep copy and maintain sched list for
itself.
Signed-off-by: Nirmoy Das
---
drivers/gpu/drm/scheduler/sched_entity.c | 10 +-
1 file changed, 1 insertion(+), 9 deletions(-)
diff --git a/drivers/gpu/drm/scheduler/sched_entity.c
I think amdgpu_ctx_init() should check for num_scheds and not call
drm_sched_entity_init()
if its zero.
On 12/10/19 3:47 PM, Nirmoy wrote:
On 12/10/19 2:00 PM, Christian König wrote:
Am 10.12.19 um 13:53 schrieb Nirmoy Das:
entity should not keep copy and maintain sched list for
itself.
On 12/10/19 2:00 PM, Christian König wrote:
Am 10.12.19 um 13:53 schrieb Nirmoy Das:
entity should not keep copy and maintain sched list for
itself.
Signed-off-by: Nirmoy Das
---
drivers/gpu/drm/scheduler/sched_entity.c | 10 +-
1 file changed, 1 insertion(+), 9 deletions(-)
diff
Am 10.12.19 um 14:00 schrieb Christian König:
Am 10.12.19 um 13:53 schrieb Nirmoy Das:
entity should not keep copy and maintain sched list for
itself.
Signed-off-by: Nirmoy Das
---
drivers/gpu/drm/scheduler/sched_entity.c | 10 +-
1 file changed, 1 insertion(+), 9 deletions(-)
Am 10.12.19 um 13:53 schrieb Nirmoy Das:
entity should not keep copy and maintain sched list for
itself.
Signed-off-by: Nirmoy Das
---
drivers/gpu/drm/scheduler/sched_entity.c | 10 +-
1 file changed, 1 insertion(+), 9 deletions(-)
diff --git
entity should not keep copy and maintain sched list for
itself.
Signed-off-by: Nirmoy Das
---
drivers/gpu/drm/scheduler/sched_entity.c | 10 +-
1 file changed, 1 insertion(+), 9 deletions(-)
diff --git a/drivers/gpu/drm/scheduler/sched_entity.c
Yeah, that won't work very well.
During bring-up we also often have the case that we can't correctly
initialize all engines, e.g. only the first SDMA comes up etc.
So better stick with the initial approach of constructing the scheduler
array for each engine type which needs it.
Regards,
I can see one issue with this. I am ignoring/removing changes from
commit 2a84e48e9712ea8591a10dd59d59ccab3d54efd6 drm/amdgpu: Only add rqs
for initialized rings.
I wonder if we can handle that differently.
Regards,
Nirmoy
On 12/9/19 2:56 PM, Nirmoy wrote:
Hi Christian,
I got a
Hi Christian,
I got a different idea, a bit more simple let me know what do you think
about it:
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
index 50bab33cba39..8de4de4f7a43 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
+++
Yes, you need to do this for the SDMA as well but in general that looks
like the idea I had in mind as well.
I would do it like this:
1. Change the special case when you only get one scheduler for an entity
to drop the pointer to the scheduler list.
This way we always use the same
On 12/6/19 8:41 PM, Christian König wrote:
Am 06.12.19 um 18:33 schrieb Nirmoy Das:
entity should not keep copy and maintain sched list for
itself.
That is a good step, but we need to take this further.
How about something like ?
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h
Am 06.12.19 um 18:33 schrieb Nirmoy Das:
entity should not keep copy and maintain sched list for
itself.
That is a good step, but we need to take this further.
The sched_list is static for the whole device and we shouldn't allocate
it inside the context at all.
Christian.
Signed-off-by:
entity should not keep copy and maintain sched list for
itself.
Signed-off-by: Nirmoy Das
---
drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c | 11 ---
drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h | 1 +
drivers/gpu/drm/scheduler/sched_entity.c | 12 +---
3 files changed, 10
19 matches
Mail list logo
