drm/amdgpu: use new scheduler load balancing for VMs

Instead of the fixed round robin use let the scheduler balance the load
of page table updates.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drivers/gpu/drm/amd/amdgpu/amdgpu_device.c

@@ -2348,7 +2348,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
 	adev->mman.buffer_funcs = NULL;
 	adev->mman.buffer_funcs_ring = NULL;
 	adev->vm_manager.vm_pte_funcs = NULL;
-	adev->vm_manager.vm_pte_num_rings = 0;
+	adev->vm_manager.vm_pte_num_rqs = 0;
 	adev->gmc.gmc_funcs = NULL;
 	adev->fence_context = dma_fence_context_alloc(AMDGPU_MAX_RINGS);
 	bitmap_zero(adev->gfx.pipe_reserve_bitmap, AMDGPU_MAX_COMPUTE_QUEUES);

drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c

@@ -2569,9 +2569,6 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	struct amdgpu_bo *root;
 	const unsigned align = min(AMDGPU_VM_PTB_ALIGN_SIZE,
 		AMDGPU_VM_PTE_COUNT(adev) * 8);
-	unsigned ring_instance;
-	struct amdgpu_ring *ring;
-	struct drm_sched_rq *rq;
 	unsigned long size;
 	uint64_t flags;
 	int r, i;
@@ -2587,12 +2584,8 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	INIT_LIST_HEAD(&vm->freed);
 
 	/* create scheduler entity for page table updates */
-
-	ring_instance = atomic_inc_return(&adev->vm_manager.vm_pte_next_ring);
-	ring_instance %= adev->vm_manager.vm_pte_num_rings;
-	ring = adev->vm_manager.vm_pte_rings[ring_instance];
-	rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_KERNEL];
-	r = drm_sched_entity_init(&vm->entity, &rq, 1, NULL);
+	r = drm_sched_entity_init(&vm->entity, adev->vm_manager.vm_pte_rqs,
+				  adev->vm_manager.vm_pte_num_rqs, NULL);
 	if (r)
 		return r;
 
@@ -2901,7 +2894,6 @@ void amdgpu_vm_manager_init(struct amdgpu_device *adev)
 	for (i = 0; i < AMDGPU_MAX_RINGS; ++i)
 		adev->vm_manager.seqno[i] = 0;
 
-	atomic_set(&adev->vm_manager.vm_pte_next_ring, 0);
 	spin_lock_init(&adev->vm_manager.prt_lock);
 	atomic_set(&adev->vm_manager.num_prt_users, 0);
 

drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h

@@ -266,9 +266,8 @@ struct amdgpu_vm_manager {
 	u64					vram_base_offset;
 	/* vm pte handling */
 	const struct amdgpu_vm_pte_funcs	*vm_pte_funcs;
-	struct amdgpu_ring                      *vm_pte_rings[AMDGPU_MAX_RINGS];
-	unsigned				vm_pte_num_rings;
-	atomic_t				vm_pte_next_ring;
+	struct drm_sched_rq			*vm_pte_rqs[AMDGPU_MAX_RINGS];
+	unsigned				vm_pte_num_rqs;
 
 	/* partial resident texture handling */
 	spinlock_t				prt_lock;

drivers/gpu/drm/amd/amdgpu/cik_sdma.c

@@ -1386,15 +1386,17 @@ static const struct amdgpu_vm_pte_funcs cik_sdma_vm_pte_funcs = {
 
 static void cik_sdma_set_vm_pte_funcs(struct amdgpu_device *adev)
 {
+	struct drm_gpu_scheduler *sched;
 	unsigned i;
 
 	if (adev->vm_manager.vm_pte_funcs == NULL) {
 		adev->vm_manager.vm_pte_funcs = &cik_sdma_vm_pte_funcs;
-		for (i = 0; i < adev->sdma.num_instances; i++)
-			adev->vm_manager.vm_pte_rings[i] =
-				&adev->sdma.instance[i].ring;
-
-		adev->vm_manager.vm_pte_num_rings = adev->sdma.num_instances;
+		for (i = 0; i < adev->sdma.num_instances; i++) {
+			sched = &adev->sdma.instance[i].ring.sched;
+			adev->vm_manager.vm_pte_rqs[i] =
+				&sched->sched_rq[DRM_SCHED_PRIORITY_KERNEL];
+		}
+		adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances;
 	}
 }
 

drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c

@@ -1312,15 +1312,17 @@ static const struct amdgpu_vm_pte_funcs sdma_v2_4_vm_pte_funcs = {
 
 static void sdma_v2_4_set_vm_pte_funcs(struct amdgpu_device *adev)
 {
+	struct drm_gpu_scheduler *sched;
 	unsigned i;
 
 	if (adev->vm_manager.vm_pte_funcs == NULL) {
 		adev->vm_manager.vm_pte_funcs = &sdma_v2_4_vm_pte_funcs;
-		for (i = 0; i < adev->sdma.num_instances; i++)
-			adev->vm_manager.vm_pte_rings[i] =
-				&adev->sdma.instance[i].ring;
-
-		adev->vm_manager.vm_pte_num_rings = adev->sdma.num_instances;
+		for (i = 0; i < adev->sdma.num_instances; i++) {
+			sched = &adev->sdma.instance[i].ring.sched;
+			adev->vm_manager.vm_pte_rqs[i] =
+				&sched->sched_rq[DRM_SCHED_PRIORITY_KERNEL];
+		}
+		adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances;
 	}
 }
 

drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c

@@ -1752,15 +1752,17 @@ static const struct amdgpu_vm_pte_funcs sdma_v3_0_vm_pte_funcs = {
 
 static void sdma_v3_0_set_vm_pte_funcs(struct amdgpu_device *adev)
 {
+	struct drm_gpu_scheduler *sched;
 	unsigned i;
 
 	if (adev->vm_manager.vm_pte_funcs == NULL) {
 		adev->vm_manager.vm_pte_funcs = &sdma_v3_0_vm_pte_funcs;
-		for (i = 0; i < adev->sdma.num_instances; i++)
-			adev->vm_manager.vm_pte_rings[i] =
-				&adev->sdma.instance[i].ring;
-
-		adev->vm_manager.vm_pte_num_rings = adev->sdma.num_instances;
+		for (i = 0; i < adev->sdma.num_instances; i++) {
+			sched = &adev->sdma.instance[i].ring.sched;
+			adev->vm_manager.vm_pte_rqs[i] =
+				&sched->sched_rq[DRM_SCHED_PRIORITY_KERNEL];
+		}
+		adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances;
 	}
 }
 

drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c

@@ -1796,15 +1796,17 @@ static const struct amdgpu_vm_pte_funcs sdma_v4_0_vm_pte_funcs = {
 
 static void sdma_v4_0_set_vm_pte_funcs(struct amdgpu_device *adev)
 {
+	struct drm_gpu_scheduler *sched;
 	unsigned i;
 
 	if (adev->vm_manager.vm_pte_funcs == NULL) {
 		adev->vm_manager.vm_pte_funcs = &sdma_v4_0_vm_pte_funcs;
-		for (i = 0; i < adev->sdma.num_instances; i++)
-			adev->vm_manager.vm_pte_rings[i] =
-				&adev->sdma.instance[i].ring;
-
-		adev->vm_manager.vm_pte_num_rings = adev->sdma.num_instances;
+		for (i = 0; i < adev->sdma.num_instances; i++) {
+			sched = &adev->sdma.instance[i].ring.sched;
+			adev->vm_manager.vm_pte_rqs[i] =
+				&sched->sched_rq[DRM_SCHED_PRIORITY_KERNEL];
+		}
+		adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances;
 	}
 }
 

drivers/gpu/drm/amd/amdgpu/si_dma.c

@@ -879,15 +879,17 @@ static const struct amdgpu_vm_pte_funcs si_dma_vm_pte_funcs = {
 
 static void si_dma_set_vm_pte_funcs(struct amdgpu_device *adev)
 {
+	struct drm_gpu_scheduler *sched;
 	unsigned i;
 
 	if (adev->vm_manager.vm_pte_funcs == NULL) {
 		adev->vm_manager.vm_pte_funcs = &si_dma_vm_pte_funcs;
-		for (i = 0; i < adev->sdma.num_instances; i++)
-			adev->vm_manager.vm_pte_rings[i] =
-				&adev->sdma.instance[i].ring;
-
-		adev->vm_manager.vm_pte_num_rings = adev->sdma.num_instances;
+		for (i = 0; i < adev->sdma.num_instances; i++) {
+			sched = &adev->sdma.instance[i].ring.sched;
+			adev->vm_manager.vm_pte_rqs[i] =
+				&sched->sched_rq[DRM_SCHED_PRIORITY_KERNEL];
+		}
+		adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances;
 	}
 }