[RFC PATCH 1/2] drm/sched: implement drm_sched_entity_num_jobs

[RFC PATCH 1/2] drm/sched: implement drm_sched_entity_num_jobs

[Nirmoy Das]

Implement drm_sched_entity_num_jobs() so that drm drivers can
query number of jobs in an entity.

Signed-off-by: Nirmoy Das <nirmoy.das@amd.com>
---
 drivers/gpu/drm/scheduler/sched_entity.c | 15 +++++++++++++++
 include/drm/gpu_scheduler.h              |  1 +
 2 files changed, 16 insertions(+)

diff --git a/drivers/gpu/drm/scheduler/sched_entity.c b/drivers/gpu/drm/scheduler/sched_entity.c
index 90fd9c30ae5a..dfe8216f2d52 100644
--- a/drivers/gpu/drm/scheduler/sched_entity.c
+++ b/drivers/gpu/drm/scheduler/sched_entity.c
@@ -119,6 +119,21 @@ static bool drm_sched_entity_is_idle(struct drm_sched_entity *entity)
 	return false;
 }
 
+/**
+ * drm_sched_entity_num_job - Get number of jobs in the entity
+ *
+ * @entity: scheduler entity
+ *
+ * Returns number of jobs in the entity
+ */
+int drm_sched_entity_num_jobs(struct drm_sched_entity *entity)
+{
+	if (drm_sched_entity_is_idle(entity))
+		return 0;
+
+	return spsc_queue_count(&entity->job_queue);
+}
+EXPORT_SYMBOL(drm_sched_entity_num_jobs);
 /**
  * drm_sched_entity_is_ready - Check if entity is ready
  *
diff --git a/include/drm/gpu_scheduler.h b/include/drm/gpu_scheduler.h
index d8972836d248..b5ceff75cbbe 100644
--- a/include/drm/gpu_scheduler.h
+++ b/include/drm/gpu_scheduler.h
@@ -341,5 +341,6 @@ void drm_sched_fence_finished(struct drm_sched_fence *fence);
 unsigned long drm_sched_suspend_timeout(struct drm_gpu_scheduler *sched);
 void drm_sched_resume_timeout(struct drm_gpu_scheduler *sched,
 		                unsigned long remaining);
+int drm_sched_entity_num_jobs(struct drm_sched_entity *entity);
 
 #endif
-- 
2.25.0

_______________________________________________
amd-gfx mailing list
amd-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

[RFC PATCH 2/2] drm/amdgpu: disable gpu load balancer for vcn jobs

[Nirmoy Das]

VCN HW  doesn't support dynamic load balance on multiple
instances for a context. This modifies the entity's sched_list
to a sched_list consist of only one drm scheduler.

Signed-off-by: Nirmoy Das <nirmoy.das@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c  |  1 +
 drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c | 25 +++++++++++++++++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h |  2 ++
 drivers/gpu/drm/amd/amdgpu/amdgpu_job.c |  1 +
 4 files changed, 29 insertions(+)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
index 8304d0c87899..00032093d8a9 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
@@ -1257,6 +1257,7 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p,
 	priority = job->base.s_priority;
 	drm_sched_entity_push_job(&job->base, entity);
 
+	amdgpu_ctx_limit_load_balance(entity);
 	amdgpu_vm_move_to_lru_tail(p->adev, &fpriv->vm);
 
 	ttm_eu_fence_buffer_objects(&p->ticket, &p->validated, p->fence);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c
index fa575bdc03c8..57b49188306d 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c
@@ -139,6 +139,7 @@ static int amdgpu_ctx_init_entity(struct amdgpu_ctx *ctx, const u32 hw_ip, const
 	if (r)
 		goto error_free_entity;
 
+	entity->hw_ip = hw_ip;
 	ctx->entities[hw_ip][ring] = entity;
 	return 0;
 
@@ -559,6 +560,30 @@ void amdgpu_ctx_priority_override(struct amdgpu_ctx *ctx,
 	}
 }
 
+static void limit_vcn_load_balance(struct amdgpu_ctx_entity *centity)
+{
+	struct drm_gpu_scheduler **scheds = &centity->entity.rq->sched;
+
+	if (drm_sched_entity_num_jobs(&centity->entity) == 1)
+		drm_sched_entity_modify_sched(&centity->entity, scheds, 1);
+
+}
+
+void amdgpu_ctx_limit_load_balance(struct drm_sched_entity *entity)
+{
+	struct amdgpu_ctx_entity *centity = to_amdgpu_ctx_entity(entity);
+
+	if (!centity)
+		return;
+
+	switch (centity->hw_ip) {
+	case AMDGPU_HW_IP_VCN_DEC:
+	case AMDGPU_HW_IP_VCN_ENC:
+		limit_vcn_load_balance(centity);
+	}
+
+}
+
 int amdgpu_ctx_wait_prev_fence(struct amdgpu_ctx *ctx,
 			       struct drm_sched_entity *entity)
 {
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h
index de490f183af2..d52d8d562d77 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h
@@ -33,6 +33,7 @@ struct amdgpu_fpriv;
 
 struct amdgpu_ctx_entity {
 	uint64_t		sequence;
+	uint32_t		hw_ip;
 	struct drm_sched_entity	entity;
 	struct dma_fence	*fences[];
 };
@@ -90,5 +91,6 @@ void amdgpu_ctx_mgr_fini(struct amdgpu_ctx_mgr *mgr);
 
 void amdgpu_ctx_init_sched(struct amdgpu_device *adev);
 
+void amdgpu_ctx_limit_load_balance(struct drm_sched_entity *entity);
 
 #endif
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
index 4981e443a884..955d12bc89ae 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
@@ -154,6 +154,7 @@ int amdgpu_job_submit(struct amdgpu_job *job, struct drm_sched_entity *entity,
 	amdgpu_job_free_resources(job);
 	priority = job->base.s_priority;
 	drm_sched_entity_push_job(&job->base, entity);
+	amdgpu_ctx_limit_load_balance(entity);
 
 	return 0;
 }
-- 
2.25.0

_______________________________________________
amd-gfx mailing list
amd-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

Re: [RFC PATCH 1/2] drm/sched: implement drm_sched_entity_num_jobs

[Andrey Grodzovsky]

On 3/11/20 1:18 PM, Nirmoy Das wrote:
> Implement drm_sched_entity_num_jobs() so that drm drivers can
> query number of jobs in an entity.
>
> Signed-off-by: Nirmoy Das <nirmoy.das@amd.com>
> ---
>   drivers/gpu/drm/scheduler/sched_entity.c | 15 +++++++++++++++
>   include/drm/gpu_scheduler.h              |  1 +
>   2 files changed, 16 insertions(+)
>
> diff --git a/drivers/gpu/drm/scheduler/sched_entity.c b/drivers/gpu/drm/scheduler/sched_entity.c
> index 90fd9c30ae5a..dfe8216f2d52 100644
> --- a/drivers/gpu/drm/scheduler/sched_entity.c
> +++ b/drivers/gpu/drm/scheduler/sched_entity.c
> @@ -119,6 +119,21 @@ static bool drm_sched_entity_is_idle(struct drm_sched_entity *entity)
>   	return false;
>   }
>   
> +/**
> + * drm_sched_entity_num_job - Get number of jobs in the entity
> + *
> + * @entity: scheduler entity
> + *
> + * Returns number of jobs in the entity
> + */
> +int drm_sched_entity_num_jobs(struct drm_sched_entity *entity)
> +{
> +	if (drm_sched_entity_is_idle(entity))
> +		return 0;
> +
> +	return spsc_queue_count(&entity->job_queue);
> +}


What about the jobs which already have been dequeued from job_queue and 
are in progress in the HW ring but yet to complete - don't they count ?

Andrey



> +EXPORT_SYMBOL(drm_sched_entity_num_jobs);
>   /**
>    * drm_sched_entity_is_ready - Check if entity is ready
>    *
> diff --git a/include/drm/gpu_scheduler.h b/include/drm/gpu_scheduler.h
> index d8972836d248..b5ceff75cbbe 100644
> --- a/include/drm/gpu_scheduler.h
> +++ b/include/drm/gpu_scheduler.h
> @@ -341,5 +341,6 @@ void drm_sched_fence_finished(struct drm_sched_fence *fence);
>   unsigned long drm_sched_suspend_timeout(struct drm_gpu_scheduler *sched);
>   void drm_sched_resume_timeout(struct drm_gpu_scheduler *sched,
>   		                unsigned long remaining);
> +int drm_sched_entity_num_jobs(struct drm_sched_entity *entity);
>   
>   #endif
_______________________________________________
amd-gfx mailing list
amd-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

Re: [RFC PATCH 1/2] drm/sched: implement drm_sched_entity_num_jobs

[Christian König]

Am 11.03.20 um 18:58 schrieb Nirmoy:
>
> On 3/11/20 6:23 PM, Andrey Grodzovsky wrote:
>>
>> On 3/11/20 1:18 PM, Nirmoy Das wrote:
>>> Implement drm_sched_entity_num_jobs() so that drm drivers can
>>> query number of jobs in an entity.
>>>
>>> Signed-off-by: Nirmoy Das <nirmoy.das@amd.com>
>>> ---
>>>   drivers/gpu/drm/scheduler/sched_entity.c | 15 +++++++++++++++
>>>   include/drm/gpu_scheduler.h              |  1 +
>>>   2 files changed, 16 insertions(+)
>>>
>>> diff --git a/drivers/gpu/drm/scheduler/sched_entity.c 
>>> b/drivers/gpu/drm/scheduler/sched_entity.c
>>> index 90fd9c30ae5a..dfe8216f2d52 100644
>>> --- a/drivers/gpu/drm/scheduler/sched_entity.c
>>> +++ b/drivers/gpu/drm/scheduler/sched_entity.c
>>> @@ -119,6 +119,21 @@ static bool drm_sched_entity_is_idle(struct 
>>> drm_sched_entity *entity)
>>>       return false;
>>>   }
>>>   +/**
>>> + * drm_sched_entity_num_job - Get number of jobs in the entity
>
>
> typo s/drm_sched_entity_num_job/drm_sched_entity_num_jobs
>
>
>>> + *
>>> + * @entity: scheduler entity
>>> + *
>>> + * Returns number of jobs in the entity
>>> + */
>>> +int drm_sched_entity_num_jobs(struct drm_sched_entity *entity)
>>> +{
>>> +    if (drm_sched_entity_is_idle(entity))
>>> +        return 0;
>>> +
>>> +    return spsc_queue_count(&entity->job_queue);
>>> +}
>>
>>
>> What about the jobs which already have been dequeued from job_queue 
>> and are in progress in the HW ring but yet to complete - don't they 
>> count ?
>
> Hi Andrey,
>
> I am thinking in terms of software side of the counting because for an 
> entity once a job dequeued, that job is completely lost.
>
> I might be wrong here that's why tagged RFC :)

My question is rather what do we need that for in the first place?

Thanks,
Christian.

>
>
> Regards,
>
> Nirmoy
>
>
>>
>> Andrey
>>
>>
>>
>>> +EXPORT_SYMBOL(drm_sched_entity_num_jobs);
>>>   /**
>>>    * drm_sched_entity_is_ready - Check if entity is ready
>>>    *
>>> diff --git a/include/drm/gpu_scheduler.h b/include/drm/gpu_scheduler.h
>>> index d8972836d248..b5ceff75cbbe 100644
>>> --- a/include/drm/gpu_scheduler.h
>>> +++ b/include/drm/gpu_scheduler.h
>>> @@ -341,5 +341,6 @@ void drm_sched_fence_finished(struct 
>>> drm_sched_fence *fence);
>>>   unsigned long drm_sched_suspend_timeout(struct drm_gpu_scheduler 
>>> *sched);
>>>   void drm_sched_resume_timeout(struct drm_gpu_scheduler *sched,
>>>                           unsigned long remaining);
>>> +int drm_sched_entity_num_jobs(struct drm_sched_entity *entity);
>>>     #endif
> _______________________________________________
> amd-gfx mailing list
> amd-gfx@lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/amd-gfx

_______________________________________________
amd-gfx mailing list
amd-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

Re: [RFC PATCH 1/2] drm/sched: implement drm_sched_entity_num_jobs

[Nirmoy]

On 3/11/20 6:23 PM, Andrey Grodzovsky wrote:
>
> On 3/11/20 1:18 PM, Nirmoy Das wrote:
>> Implement drm_sched_entity_num_jobs() so that drm drivers can
>> query number of jobs in an entity.
>>
>> Signed-off-by: Nirmoy Das <nirmoy.das@amd.com>
>> ---
>>   drivers/gpu/drm/scheduler/sched_entity.c | 15 +++++++++++++++
>>   include/drm/gpu_scheduler.h              |  1 +
>>   2 files changed, 16 insertions(+)
>>
>> diff --git a/drivers/gpu/drm/scheduler/sched_entity.c 
>> b/drivers/gpu/drm/scheduler/sched_entity.c
>> index 90fd9c30ae5a..dfe8216f2d52 100644
>> --- a/drivers/gpu/drm/scheduler/sched_entity.c
>> +++ b/drivers/gpu/drm/scheduler/sched_entity.c
>> @@ -119,6 +119,21 @@ static bool drm_sched_entity_is_idle(struct 
>> drm_sched_entity *entity)
>>       return false;
>>   }
>>   +/**
>> + * drm_sched_entity_num_job - Get number of jobs in the entity


typo s/drm_sched_entity_num_job/drm_sched_entity_num_jobs


>> + *
>> + * @entity: scheduler entity
>> + *
>> + * Returns number of jobs in the entity
>> + */
>> +int drm_sched_entity_num_jobs(struct drm_sched_entity *entity)
>> +{
>> +    if (drm_sched_entity_is_idle(entity))
>> +        return 0;
>> +
>> +    return spsc_queue_count(&entity->job_queue);
>> +}
>
>
> What about the jobs which already have been dequeued from job_queue 
> and are in progress in the HW ring but yet to complete - don't they 
> count ?

Hi Andrey,

I am thinking in terms of software side of the counting because for an 
entity once a job dequeued, that job is completely lost.

I might be wrong here that's why tagged RFC :)


Regards,

Nirmoy


>
> Andrey
>
>
>
>> +EXPORT_SYMBOL(drm_sched_entity_num_jobs);
>>   /**
>>    * drm_sched_entity_is_ready - Check if entity is ready
>>    *
>> diff --git a/include/drm/gpu_scheduler.h b/include/drm/gpu_scheduler.h
>> index d8972836d248..b5ceff75cbbe 100644
>> --- a/include/drm/gpu_scheduler.h
>> +++ b/include/drm/gpu_scheduler.h
>> @@ -341,5 +341,6 @@ void drm_sched_fence_finished(struct 
>> drm_sched_fence *fence);
>>   unsigned long drm_sched_suspend_timeout(struct drm_gpu_scheduler 
>> *sched);
>>   void drm_sched_resume_timeout(struct drm_gpu_scheduler *sched,
>>                           unsigned long remaining);
>> +int drm_sched_entity_num_jobs(struct drm_sched_entity *entity);
>>     #endif
_______________________________________________
amd-gfx mailing list
amd-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

Re: [RFC PATCH 2/2] drm/amdgpu: disable gpu load balancer for vcn jobs

[Christian König]

Am 11.03.20 um 18:18 schrieb Nirmoy Das:
> VCN HW  doesn't support dynamic load balance on multiple
> instances for a context. This modifies the entity's sched_list
> to a sched_list consist of only one drm scheduler.
>
> Signed-off-by: Nirmoy Das <nirmoy.das@amd.com>
> ---
>   drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c  |  1 +
>   drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c | 25 +++++++++++++++++++++++++
>   drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h |  2 ++
>   drivers/gpu/drm/amd/amdgpu/amdgpu_job.c |  1 +
>   4 files changed, 29 insertions(+)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
> index 8304d0c87899..00032093d8a9 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
> @@ -1257,6 +1257,7 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p,
>   	priority = job->base.s_priority;
>   	drm_sched_entity_push_job(&job->base, entity);
>   
> +	amdgpu_ctx_limit_load_balance(entity);
>   	amdgpu_vm_move_to_lru_tail(p->adev, &fpriv->vm);
>   
>   	ttm_eu_fence_buffer_objects(&p->ticket, &p->validated, p->fence);
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c
> index fa575bdc03c8..57b49188306d 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c
> @@ -139,6 +139,7 @@ static int amdgpu_ctx_init_entity(struct amdgpu_ctx *ctx, const u32 hw_ip, const
>   	if (r)
>   		goto error_free_entity;
>   
> +	entity->hw_ip = hw_ip;
>   	ctx->entities[hw_ip][ring] = entity;
>   	return 0;
>   
> @@ -559,6 +560,30 @@ void amdgpu_ctx_priority_override(struct amdgpu_ctx *ctx,
>   	}
>   }
>   
> +static void limit_vcn_load_balance(struct amdgpu_ctx_entity *centity)
> +{
> +	struct drm_gpu_scheduler **scheds = &centity->entity.rq->sched;
> +
> +	if (drm_sched_entity_num_jobs(&centity->entity) == 1)

That check doesn't work correctly, the job might actually already be 
processed when we hit here.

> +		drm_sched_entity_modify_sched(&centity->entity, scheds, 1);

Just always update the scheduler here.

> +
> +}
> +
> +void amdgpu_ctx_limit_load_balance(struct drm_sched_entity *entity)
> +{
> +	struct amdgpu_ctx_entity *centity = to_amdgpu_ctx_entity(entity);
> +
> +	if (!centity)
> +		return;

That check looks superfluous to me.

> +
> +	switch (centity->hw_ip) {

Better get the ring from entity->rq->sched instead.

> +	case AMDGPU_HW_IP_VCN_DEC:
> +	case AMDGPU_HW_IP_VCN_ENC:

Maybe better to make that a flag in the ring functions, but this way 
works as well.

Regards,
Christian.

> +		limit_vcn_load_balance(centity);
> +	}
> +
> +}
> +
>   int amdgpu_ctx_wait_prev_fence(struct amdgpu_ctx *ctx,
>   			       struct drm_sched_entity *entity)
>   {
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h
> index de490f183af2..d52d8d562d77 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h
> @@ -33,6 +33,7 @@ struct amdgpu_fpriv;
>   
>   struct amdgpu_ctx_entity {
>   	uint64_t		sequence;
> +	uint32_t		hw_ip;
>   	struct drm_sched_entity	entity;
>   	struct dma_fence	*fences[];
>   };
> @@ -90,5 +91,6 @@ void amdgpu_ctx_mgr_fini(struct amdgpu_ctx_mgr *mgr);
>   
>   void amdgpu_ctx_init_sched(struct amdgpu_device *adev);
>   
> +void amdgpu_ctx_limit_load_balance(struct drm_sched_entity *entity);
>   
>   #endif
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
> index 4981e443a884..955d12bc89ae 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
> @@ -154,6 +154,7 @@ int amdgpu_job_submit(struct amdgpu_job *job, struct drm_sched_entity *entity,
>   	amdgpu_job_free_resources(job);
>   	priority = job->base.s_priority;
>   	drm_sched_entity_push_job(&job->base, entity);
> +	amdgpu_ctx_limit_load_balance(entity);
>   
>   	return 0;
>   }

_______________________________________________
amd-gfx mailing list
amd-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/amd-gfx

Re: [RFC PATCH 2/2] drm/amdgpu: disable gpu load balancer for vcn jobs

[Nirmoy]

On 3/11/20 7:03 PM, Christian König wrote:
> Am 11.03.20 um 18:18 schrieb Nirmoy Das:
>> VCN HW  doesn't support dynamic load balance on multiple
>> instances for a context. This modifies the entity's sched_list
>> to a sched_list consist of only one drm scheduler.
>>
>> Signed-off-by: Nirmoy Das <nirmoy.das@amd.com>
>> ---
>>   drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c  |  1 +
>>   drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c | 25 +++++++++++++++++++++++++
>>   drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h |  2 ++
>>   drivers/gpu/drm/amd/amdgpu/amdgpu_job.c |  1 +
>>   4 files changed, 29 insertions(+)
>>
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c 
>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
>> index 8304d0c87899..00032093d8a9 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
>> @@ -1257,6 +1257,7 @@ static int amdgpu_cs_submit(struct 
>> amdgpu_cs_parser *p,
>>       priority = job->base.s_priority;
>>       drm_sched_entity_push_job(&job->base, entity);
>>   +    amdgpu_ctx_limit_load_balance(entity);
>>       amdgpu_vm_move_to_lru_tail(p->adev, &fpriv->vm);
>>         ttm_eu_fence_buffer_objects(&p->ticket, &p->validated, 
>> p->fence);
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c 
>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c
>> index fa575bdc03c8..57b49188306d 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c
>> @@ -139,6 +139,7 @@ static int amdgpu_ctx_init_entity(struct 
>> amdgpu_ctx *ctx, const u32 hw_ip, const
>>       if (r)
>>           goto error_free_entity;
>>   +    entity->hw_ip = hw_ip;
>>       ctx->entities[hw_ip][ring] = entity;
>>       return 0;
>>   @@ -559,6 +560,30 @@ void amdgpu_ctx_priority_override(struct 
>> amdgpu_ctx *ctx,
>>       }
>>   }
>>   +static void limit_vcn_load_balance(struct amdgpu_ctx_entity *centity)
>> +{
>> +    struct drm_gpu_scheduler **scheds = &centity->entity.rq->sched;
>> +
>> +    if (drm_sched_entity_num_jobs(&centity->entity) == 1)
>
> That check doesn't work correctly, the job might actually already be 
> processed when we hit here.


Okay now I know what Andrey meant. I will resend a updated patch.


Thanks,

Nirmoy


>
>> + drm_sched_entity_modify_sched(&centity->entity, scheds, 1);
>
> Just always update the scheduler here.
>
>> +
>> +}
>> +
>> +void amdgpu_ctx_limit_load_balance(struct drm_sched_entity *entity)
>> +{
>> +    struct amdgpu_ctx_entity *centity = to_amdgpu_ctx_entity(entity);
>> +
>> +    if (!centity)
>> +        return;
>
> That check looks superfluous to me.
>
>> +
>> +    switch (centity->hw_ip) {
>
> Better get the ring from entity->rq->sched instead.
>
>> +    case AMDGPU_HW_IP_VCN_DEC:
>> +    case AMDGPU_HW_IP_VCN_ENC:
>
> Maybe better to make that a flag in the ring functions, but this way 
> works as well.
>
> Regards,
> Christian.
>
>> +        limit_vcn_load_balance(centity);
>> +    }
>> +
>> +}
>> +
>>   int amdgpu_ctx_wait_prev_fence(struct amdgpu_ctx *ctx,
>>                      struct drm_sched_entity *entity)
>>   {
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h 
>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h
>> index de490f183af2..d52d8d562d77 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h
>> @@ -33,6 +33,7 @@ struct amdgpu_fpriv;
>>     struct amdgpu_ctx_entity {
>>       uint64_t        sequence;
>> +    uint32_t        hw_ip;
>>       struct drm_sched_entity    entity;
>>       struct dma_fence    *fences[];
>>   };
>> @@ -90,5 +91,6 @@ void amdgpu_ctx_mgr_fini(struct amdgpu_ctx_mgr *mgr);
>>     void amdgpu_ctx_init_sched(struct amdgpu_device *adev);
>>   +void amdgpu_ctx_limit_load_balance(struct drm_sched_entity *entity);
>>     #endif
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c 
>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
>> index 4981e443a884..955d12bc89ae 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
>> @@ -154,6 +154,7 @@ int amdgpu_job_submit(struct amdgpu_job *job, 
>> struct drm_sched_entity *entity,
>>       amdgpu_job_free_resources(job);
>>       priority = job->base.s_priority;
>>       drm_sched_entity_push_job(&job->base, entity);
>> +    amdgpu_ctx_limit_load_balance(entity);
>>         return 0;
>>   }
>
_______________________________________________
amd-gfx mailing list
amd-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/amd-gfx