drm/scheduler: improve job distribution with multiple queues

This patch uses score based logic to select a new rq for better
loadbalance between multiple rq/scheds instead of num_jobs.

Below are test results after running amdgpu_test from mesa drm

Before this patch:

sched_name num of many times it got scheduled
========= ==================================
sdma0 314
sdma1 32
comp_1.0.0 56
comp_1.0.1 0
comp_1.1.0 0
comp_1.1.1 0
comp_1.2.0 0
comp_1.2.1 0
comp_1.3.0 0
comp_1.3.1 0
After this patch:

sched_name num of many times it got scheduled
========= ==================================
sdma0 216
sdma1 185
comp_1.0.0 39
comp_1.0.1 9
comp_1.1.0 12
comp_1.1.1 0
comp_1.2.0 12
comp_1.2.1 0
comp_1.3.0 12
comp_1.3.1 0

Signed-off-by: Nirmoy Das <[e-mail]>
Reviewed-by: Christian König <[e-mail]>
Signed-off-by: Alex Deucher <[e-mail]>

drivers/gpu/drm/scheduler/sched_entity.c

···
 drm_sched_entity_get_free_sched(struct drm_sched_entity *entity)
 {
 	struct drm_sched_rq *rq = NULL;
-	unsigned int min_jobs = UINT_MAX, num_jobs;
+	unsigned int min_score = UINT_MAX, num_score;
 	int i;
 
 	for (i = 0; i < entity->num_sched_list; ++i) {
···
 			continue;
 		}
 
-		num_jobs = atomic_read(&sched->num_jobs);
-		if (num_jobs < min_jobs) {
-			min_jobs = num_jobs;
+		num_score = atomic_read(&sched->score);
+		if (num_score < min_score) {
+			min_score = num_score;
 			rq = &entity->sched_list[i]->sched_rq[entity->priority];
 		}
 	}
···
 	bool first;
 
 	trace_drm_sched_job(sched_job, entity);
-	atomic_inc(&entity->rq->sched->num_jobs);
+	atomic_inc(&entity->rq->sched->score);
 	WRITE_ONCE(entity->last_user, current->group_leader);
 	first = spsc_queue_push(&entity->job_queue, &sched_job->queue_node);
 

drivers/gpu/drm/scheduler/sched_main.c

···
 	if (!list_empty(&entity->list))
 		return;
 	spin_lock(&rq->lock);
+	atomic_inc(&rq->sched->score);
 	list_add_tail(&entity->list, &rq->entities);
 	spin_unlock(&rq->lock);
 }
···
 	if (list_empty(&entity->list))
 		return;
 	spin_lock(&rq->lock);
+	atomic_dec(&rq->sched->score);
 	list_del_init(&entity->list);
 	if (rq->current_entity == entity)
 		rq->current_entity = NULL;
···
 	struct drm_gpu_scheduler *sched = s_fence->sched;
 
 	atomic_dec(&sched->hw_rq_count);
-	atomic_dec(&sched->num_jobs);
+	atomic_dec(&sched->score);
 
 	trace_drm_sched_process_job(s_fence);
 
···
 	spin_lock_init(&sched->job_list_lock);
 	atomic_set(&sched->hw_rq_count, 0);
 	INIT_DELAYED_WORK(&sched->work_tdr, drm_sched_job_timedout);
-	atomic_set(&sched->num_jobs, 0);
+	atomic_set(&sched->score, 0);
 	atomic64_set(&sched->job_id_count, 0);
 
 	/* Each scheduler will run on a seperate kernel thread */

include/drm/gpu_scheduler.h

···
  * @job_list_lock: lock to protect the ring_mirror_list.
  * @hang_limit: once the hangs by a job crosses this limit then it is marked
  *              guilty and it will be considered for scheduling further.
- * @num_jobs: the number of jobs in queue in the scheduler
+ * @score: score to help loadbalancer pick a idle sched
  * @ready: marks if the underlying HW is ready to work
  * @free_guilty: A hit to time out handler to free the guilty job.
  *
···
 	struct list_head		ring_mirror_list;
 	spinlock_t			job_list_lock;
 	int				hang_limit;
-	atomic_t                        num_jobs;
-	bool			ready;
+	atomic_t                        score;
+	bool				ready;
 	bool				free_guilty;
 };