drm/scheduler: rework entity creation
Subsystems (8)
PATCH v2 1/4 drm/scheduler: rework entity creation
Entity currently keeps a copy of run_queue list and modify it in drm_sched_entity_set_priority(). Entities shouldn't modify run_queue list. Use drm_gpu_scheduler list instead of drm_sched_rq list in drm_sched_entity struct. In this way we can select a runqueue based on entity/ctx's priority for a drm scheduler.
--- drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c | 7 ++- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 8 ++- drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c | 7 ++- drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c | 7 ++- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 14 +++-- drivers/gpu/drm/etnaviv/etnaviv_drv.c | 7 ++- drivers/gpu/drm/lima/lima_sched.c | 5 +- drivers/gpu/drm/panfrost/panfrost_job.c | 8 ++- drivers/gpu/drm/scheduler/sched_entity.c | 74 ++++++++++-------------- drivers/gpu/drm/v3d/v3d_drv.c | 8 ++- include/drm/gpu_scheduler.h | 8 ++- 11 files changed, 78 insertions(+), 75 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c +4 -3
@@ -122,7 +122,7 @@ static int amdgpu_ctx_init(struct amdgpu_device *adev,for (i =0 ; i <AMDGPU_HW_IP_NUM ; ++i) {struct amdgpu_ring *rings[AMDGPU_MAX_RINGS ]; -struct drm_sched_rq *rqs[AMDGPU_MAX_RINGS ]; +struct drm_gpu_scheduler *sched_list[AMDGPU_MAX_RINGS ];unsigned num_rings =0 ;unsigned num_rqs =0 ;
@@ -181,12 +181,13 @@ static int amdgpu_ctx_init(struct amdgpu_device *adev,if (!rings[j]->adev)continue ; - rqs[num_rqs++] = &rings[j]->sched.sched_rq[priority]; + sched_list[num_rqs++] = &rings[j]->sched; }for (j =0 ; j < amdgpu_ctx_num_entities[i]; ++j) r =drm_sched_entity_init (&ctx->entities[i][j].entity, - rqs, num_rqs, &ctx->guilty); + priority, sched_list, + num_rqs, &ctx->guilty);if (r) goto error_cleanup_entities; }
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +5 -3
@@ -1954,11 +1954,13 @@ void amdgpu_ttm_set_buffer_funcs_status(struct amdgpu_device *adev, bool enable)if (enable) {struct amdgpu_ring *ring; -struct drm_sched_rq *rq; +struct drm_gpu_scheduler *sched; ring = adev->mman.buffer_funcs_ring; - rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_KERNEL ]; - r =drm_sched_entity_init (&adev->mman.entity, &rq,1 ,NULL ); + sched = &ring->sched; + r =drm_sched_entity_init (&adev->mman.entity, +DRM_SCHED_PRIORITY_KERNEL , &sched, +1 ,NULL );if (r) {DRM_ERROR ("Failed setting up TTM BO move entity (%d)\n" , r);
drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c +4 -3
@@ -330,12 +330,13 @@ int amdgpu_uvd_sw_fini(struct amdgpu_device *adev)int amdgpu_uvd_entity_init (struct amdgpu_device *adev) {struct amdgpu_ring *ring; -struct drm_sched_rq *rq; +struct drm_gpu_scheduler *sched;int r; ring = &adev->uvd.inst[0 ].ring; - rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL ]; - r =drm_sched_entity_init (&adev->uvd.entity, &rq,1 ,NULL ); + sched = &ring->sched; + r =drm_sched_entity_init (&adev->uvd.entity,DRM_SCHED_PRIORITY_NORMAL , + &sched,1 ,NULL );if (r) {DRM_ERROR ("Failed setting up UVD kernel entity.\n" );return r;
drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c +4 -3
@@ -240,12 +240,13 @@ int amdgpu_vce_sw_fini(struct amdgpu_device *adev)int amdgpu_vce_entity_init (struct amdgpu_device *adev) {struct amdgpu_ring *ring; -struct drm_sched_rq *rq; +struct drm_gpu_scheduler *sched;int r; ring = &adev->vce.ring[0 ]; - rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL ]; - r =drm_sched_entity_init (&adev->vce.entity, &rq,1 ,NULL ); + sched = &ring->sched; + r =drm_sched_entity_init (&adev->vce.entity,DRM_SCHED_PRIORITY_NORMAL , + &sched,1 ,NULL );if (r !=0 ) {DRM_ERROR ("Failed setting up VCE run queue.\n" );return r;
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +10 -4
@@ -2744,6 +2744,7 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, {struct amdgpu_bo_param bp;struct amdgpu_bo *root; +struct drm_gpu_scheduler *sched_list[AMDGPU_MAX_RINGS ];int r, i; vm->va =RB_ROOT_CACHED ;
@@ -2757,14 +2758,19 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,spin_lock_init (&vm->invalidated_lock);INIT_LIST_HEAD (&vm->freed); +for (i =0 ; i < adev->vm_manager.vm_pte_num_rqs; i++) + sched_list[i] = adev->vm_manager.vm_pte_rqs[i]->sched; +/* create scheduler entities for page table updates */ - r =drm_sched_entity_init (&vm->direct, adev->vm_manager.vm_pte_rqs, - adev->vm_manager.vm_pte_num_rqs,NULL ); + r =drm_sched_entity_init (&vm->direct,DRM_SCHED_PRIORITY_NORMAL , + sched_list, adev->vm_manager.vm_pte_num_rqs, +NULL );if (r)return r; - r =drm_sched_entity_init (&vm->delayed, adev->vm_manager.vm_pte_rqs, - adev->vm_manager.vm_pte_num_rqs,NULL ); + r =drm_sched_entity_init (&vm->delayed,DRM_SCHED_PRIORITY_NORMAL , + sched_list, adev->vm_manager.vm_pte_num_rqs, +NULL );if (r) goto error_free_direct;
drivers/gpu/drm/etnaviv/etnaviv_drv.c +4 -3
@@ -65,12 +65,13 @@ static int etnaviv_open(struct drm_device *dev, struct drm_file *file)for (i =0 ; i <ETNA_MAX_PIPES ; i++) {struct etnaviv_gpu *gpu = priv->gpu[i]; -struct drm_sched_rq *rq; +struct drm_gpu_scheduler *sched;if (gpu) { - rq = &gpu->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL ]; + sched = &gpu->sched;drm_sched_entity_init (&ctx->sched_entity[i], - &rq,1 ,NULL ); +DRM_SCHED_PRIORITY_NORMAL , &sched, +1 ,NULL ); } }
drivers/gpu/drm/lima/lima_sched.c +3 -2
@@ -159,9 +159,10 @@ int lima_sched_context_init(struct lima_sched_pipe *pipe,struct lima_sched_context *context, atomic_t *guilty) { - struct drm_sched_rq *rq = pipe->base.sched_rq + DRM_SCHED_PRIORITY_NORMAL; + struct drm_gpu_scheduler *sched = &pipe->base; -return drm_sched_entity_init (&context->base, &rq,1 , guilty); +return drm_sched_entity_init (&context->base,DRM_SCHED_PRIORITY_NORMAL , + &sched,1 , guilty); } void lima_sched_context_fini(struct lima_sched_pipe *pipe,
drivers/gpu/drm/panfrost/panfrost_job.c +5 -3
@@ -542,12 +542,14 @@ int panfrost_job_open(struct panfrost_file_priv *panfrost_priv) {struct panfrost_device *pfdev = panfrost_priv->pfdev;struct panfrost_job_slot *js = pfdev->js; -struct drm_sched_rq *rq; +struct drm_gpu_scheduler *sched;int ret, i;for (i =0 ; i <NUM_JOB_SLOTS ; i++) { - rq = &js->queue[i].sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL ]; - ret =drm_sched_entity_init (&panfrost_priv->sched_entity[i], &rq,1 ,NULL ); + sched = &js->queue[i].sched; + ret =drm_sched_entity_init (&panfrost_priv->sched_entity[i], +DRM_SCHED_PRIORITY_NORMAL , &sched, +1 ,NULL );if (WARN_ON (ret))return ret; }
drivers/gpu/drm/scheduler/sched_entity.c +29 -45
148 lines not on this page · raw message
drivers/gpu/drm/v3d/v3d_drv.c +5 -3
@@ -140,7 +140,7 @@ v3d_open(struct drm_device *dev, struct drm_file *file) {struct v3d_dev *v3d =to_v3d_dev (dev);struct v3d_file_priv *v3d_priv; -struct drm_sched_rq *rq; +struct drm_gpu_scheduler *sched;int i; v3d_priv =kzalloc (sizeof (*v3d_priv),GFP_KERNEL );
@@ -150,8 +150,10 @@ v3d_open(struct drm_device *dev, struct drm_file *file) v3d_priv->v3d = v3d;for (i =0 ; i <V3D_MAX_QUEUES ; i++) { - rq = &v3d->queue[i].sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL ]; -drm_sched_entity_init (&v3d_priv->sched_entity[i], &rq,1 ,NULL ); + sched = &v3d->queue[i].sched; +drm_sched_entity_init (&v3d_priv->sched_entity[i], +DRM_SCHED_PRIORITY_NORMAL , &sched, +1 ,NULL ); } file->driver_priv = v3d_priv;
include/drm/gpu_scheduler.h +5 -3
@@ -81,8 +81,9 @@ enum drm_sched_priority {struct drm_sched_entity {struct list_head list;struct drm_sched_rq *rq; -struct drm_sched_rq **rq_list; -unsigned int num_rq_list; +unsigned int num_sched_list; +struct drm_gpu_scheduler **sched_list; +enum drm_sched_priority priority;spinlock_t rq_lock;struct spsc_queue job_queue;
@@ -312,7 +313,8 @@ void drm_sched_rq_remove_entity(struct drm_sched_rq *rq, struct drm_sched_entity *entity);int drm_sched_entity_init (struct drm_sched_entity *entity, -struct drm_sched_rq **rq_list, +enum drm_sched_priority priority, +struct drm_gpu_scheduler **sched_list,unsigned int num_rq_list,atomic_t *guilty);long drm_sched_entity_flush (struct drm_sched_entity *entity,long timeout);
--
2.24.0
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
PATCH 2/4 drm/amdgpu: replace vm_pte's run-queue list with drm gpu scheds list
drm_sched_entity_init() takes drm gpu scheduler list instead of drm_sched_rq list. This makes conversion of drm_sched_rq list to drm gpu scheduler list unnecessary
--- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 11 ++++------- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 4 ++-- drivers/gpu/drm/amd/amdgpu/cik_sdma.c | 8 +++----- drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c | 8 +++----- drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c | 8 +++----- drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c | 5 ++--- drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c | 8 +++----- drivers/gpu/drm/amd/amdgpu/si_dma.c | 8 +++----- 9 files changed, 24 insertions(+), 38 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +1 -1
@@ -2785,7 +2785,7 @@ 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_rqs =0 ; + adev->vm_manager.vm_pte_num_scheds =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 +4 -7
@@ -2744,7 +2744,6 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, {struct amdgpu_bo_param bp;struct amdgpu_bo *root; -struct drm_gpu_scheduler *sched_list[AMDGPU_MAX_RINGS ];int r, i; vm->va =RB_ROOT_CACHED ;
@@ -2758,19 +2757,17 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,spin_lock_init (&vm->invalidated_lock);INIT_LIST_HEAD (&vm->freed); -for (i =0 ; i < adev->vm_manager.vm_pte_num_rqs; i++) - sched_list[i] = adev->vm_manager.vm_pte_rqs[i]->sched;/* create scheduler entities for page table updates */ r =drm_sched_entity_init (&vm->direct,DRM_SCHED_PRIORITY_NORMAL , - sched_list, adev->vm_manager.vm_pte_num_rqs, -NULL ); + adev->vm_manager.vm_pte_scheds, + adev->vm_manager.vm_pte_num_scheds,NULL );if (r)return r; r =drm_sched_entity_init (&vm->delayed,DRM_SCHED_PRIORITY_NORMAL , - sched_list, adev->vm_manager.vm_pte_num_rqs, -NULL ); + adev->vm_manager.vm_pte_scheds, + adev->vm_manager.vm_pte_num_scheds,NULL );if (r) goto error_free_direct;
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +2 -2
@@ -330,8 +330,8 @@ struct amdgpu_vm_manager {u64 vram_base_offset;/* vm pte handling */ const struct amdgpu_vm_pte_funcs *vm_pte_funcs; -struct drm_sched_rq *vm_pte_rqs[AMDGPU_MAX_RINGS ]; -unsigned vm_pte_num_rqs; +struct drm_gpu_scheduler *vm_pte_scheds[AMDGPU_MAX_RINGS ]; +unsigned vm_pte_num_scheds;struct amdgpu_ring *page_fault;/* partial resident texture handling */
drivers/gpu/drm/amd/amdgpu/cik_sdma.c +3 -5
@@ -1373,16 +1373,14 @@ 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; adev->vm_manager.vm_pte_funcs = &cik_sdma_vm_pte_funcs;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_scheds[i] = + &adev->sdma.instance[i].ring.sched; } - adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances; + adev->vm_manager.vm_pte_num_scheds = adev->sdma.num_instances; } const struct amdgpu_ip_block_version cik_sdma_ip_block =
drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c +3 -5
@@ -1261,16 +1261,14 @@ 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; adev->vm_manager.vm_pte_funcs = &sdma_v2_4_vm_pte_funcs;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_scheds[i] = + &adev->sdma.instance[i].ring.sched; } - adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances; + adev->vm_manager.vm_pte_num_scheds = adev->sdma.num_instances; } const struct amdgpu_ip_block_version sdma_v2_4_ip_block =
drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c +3 -5
@@ -1699,16 +1699,14 @@ 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; adev->vm_manager.vm_pte_funcs = &sdma_v3_0_vm_pte_funcs;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_scheds[i] = + &adev->sdma.instance[i].ring.sched; } - adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances; + adev->vm_manager.vm_pte_num_scheds = adev->sdma.num_instances; } const struct amdgpu_ip_block_version sdma_v3_0_ip_block =
drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c +2 -3
@@ -2411,10 +2411,9 @@ static void sdma_v4_0_set_vm_pte_funcs(struct amdgpu_device *adev) sched = &adev->sdma.instance[i].page.sched; else 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_scheds[i] = sched; } - adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances; + adev->vm_manager.vm_pte_num_scheds = adev->sdma.num_instances; }const struct amdgpu_ip_block_version sdma_v4_0_ip_block = {
drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c +3 -5
@@ -1723,17 +1723,15 @@ static const struct amdgpu_vm_pte_funcs sdma_v5_0_vm_pte_funcs = {static void sdma_v5_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_v5_0_vm_pte_funcs;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_scheds[i] = + &adev->sdma.instance[i].ring.sched; } - adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances; + adev->vm_manager.vm_pte_num_scheds = adev->sdma.num_instances; } }
drivers/gpu/drm/amd/amdgpu/si_dma.c +3 -5
@@ -835,16 +835,14 @@ 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; adev->vm_manager.vm_pte_funcs = &si_dma_vm_pte_funcs;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_scheds[i] = + &adev->sdma.instance[i].ring.sched; } - adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances; + adev->vm_manager.vm_pte_num_scheds = adev->sdma.num_instances; } const struct amdgpu_ip_block_version si_dma_ip_block =
Signature · 6 lines
2.24.0 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
PATCH v2 3/4 amd/amdgpu: add sched array to IPs with multiple run-queues
In reply to Nirmoy DasThis sched array can be passed on to entity creation routine instead of manually creating such sched array on every context creation.
--- drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c | 113 +++++++++++++-------- drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h | 3 + drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 2 + drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h | 4 + drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.h | 2 + drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h | 2 + drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h | 9 +- 7 files changed, 89 insertions(+), 46 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c +69 -44
@@ -74,7 +74,7 @@ static int amdgpu_ctx_init(struct amdgpu_device *adev, struct amdgpu_ctx *ctx) {unsigned num_entities =amdgpu_ctx_total_num_entities (); -unsigned i, j, k; +unsigned i, j;int r;if (priority <0 || priority >=DRM_SCHED_PRIORITY_MAX )
@@ -121,73 +121,56 @@ static int amdgpu_ctx_init(struct amdgpu_device *adev, ctx->override_priority =DRM_SCHED_PRIORITY_UNSET ;for (i =0 ; i <AMDGPU_HW_IP_NUM ; ++i) { -struct amdgpu_ring *rings[AMDGPU_MAX_RINGS ]; -struct drm_gpu_scheduler *sched_list[AMDGPU_MAX_RINGS ]; -unsigned num_rings =0 ; -unsigned num_rqs =0 ; +struct drm_gpu_scheduler **scheds; +struct drm_gpu_scheduler *sched; +unsigned num_scheds =0 ;switch (i) {case AMDGPU_HW_IP_GFX : - rings[0 ] = &adev->gfx.gfx_ring[0 ]; - num_rings =1 ; + scheds = adev->gfx.gfx_sched; + num_scheds =1 ;break ;case AMDGPU_HW_IP_COMPUTE : -for (j =0 ; j < adev->gfx.num_compute_rings; ++j) - rings[j] = &adev->gfx.compute_ring[j]; - num_rings = adev->gfx.num_compute_rings; + scheds = adev->gfx.compute_sched; + num_scheds = adev->gfx.num_compute_sched;break ;case AMDGPU_HW_IP_DMA : -for (j =0 ; j < adev->sdma.num_instances; ++j) - rings[j] = &adev->sdma.instance[j].ring; - num_rings = adev->sdma.num_instances; + scheds = adev->sdma.sdma_sched; + num_scheds = adev->sdma.num_sdma_sched;break ;case AMDGPU_HW_IP_UVD : - rings[0 ] = &adev->uvd.inst[0 ].ring; - num_rings =1 ; + sched = &adev->uvd.inst[0 ].ring.sched; + scheds = &sched; + num_scheds =1 ;break ;case AMDGPU_HW_IP_VCE : - rings[0 ] = &adev->vce.ring[0 ]; - num_rings =1 ; + sched = &adev->vce.ring[0 ].sched; + scheds = &sched; + num_scheds =1 ;break ;case AMDGPU_HW_IP_UVD_ENC : - rings[0 ] = &adev->uvd.inst[0 ].ring_enc[0 ]; - num_rings =1 ; + sched = &adev->uvd.inst[0 ].ring_enc[0 ].sched; + scheds = &sched; + num_scheds =1 ;break ;case AMDGPU_HW_IP_VCN_DEC : -for (j =0 ; j < adev->vcn.num_vcn_inst; ++j) { -if (adev->vcn.harvest_config & (1 << j)) -continue ; - rings[num_rings++] = &adev->vcn.inst[j].ring_dec; - } + scheds = adev->vcn.vcn_dec_sched; + num_scheds = adev->vcn.num_vcn_dec_sched;break ;case AMDGPU_HW_IP_VCN_ENC : -for (j =0 ; j < adev->vcn.num_vcn_inst; ++j) { -if (adev->vcn.harvest_config & (1 << j)) -continue ; -for (k =0 ; k < adev->vcn.num_enc_rings; ++k) - rings[num_rings++] = &adev->vcn.inst[j].ring_enc[k]; - } + scheds = adev->vcn.vcn_enc_sched; + num_scheds = adev->vcn.num_vcn_enc_sched;break ;case AMDGPU_HW_IP_VCN_JPEG : -for (j =0 ; j < adev->jpeg.num_jpeg_inst; ++j) { -if (adev->vcn.harvest_config & (1 << j)) -continue ; - rings[num_rings++] = &adev->jpeg.inst[j].ring_dec; - } + scheds = adev->jpeg.jpeg_sched; + num_scheds = adev->jpeg.num_jpeg_sched;break ; } -for (j =0 ; j < num_rings; ++j) { -if (!rings[j]->adev) -continue ; - - sched_list[num_rqs++] = &rings[j]->sched; - } -for (j =0 ; j < amdgpu_ctx_num_entities[i]; ++j) r =drm_sched_entity_init (&ctx->entities[i][j].entity, - priority, sched_list, - num_rqs, &ctx->guilty); + priority, scheds, + num_scheds, &ctx->guilty);if (r) goto error_cleanup_entities; }
@@ -631,3 +614,45 @@ void amdgpu_ctx_mgr_fini(struct amdgpu_ctx_mgr *mgr)idr_destroy (&mgr->ctx_handles);mutex_destroy (&mgr->lock); } + +void amdgpu_ctx_init_sched (struct amdgpu_device *adev) +{ +int i, j; + +for (i =0 ; i < adev->gfx.num_gfx_rings; i++) { + adev->gfx.gfx_sched[i] = &adev->gfx.gfx_ring[i].sched; + adev->gfx.num_gfx_sched++; + } + +for (i =0 ; i < adev->gfx.num_compute_rings; i++) { + adev->gfx.compute_sched[i] = &adev->gfx.compute_ring[i].sched; + adev->gfx.num_compute_sched++; + } + +for (i =0 ; i < adev->sdma.num_instances; i++) { + adev->sdma.sdma_sched[i] = &adev->sdma.instance[i].ring.sched; + adev->sdma.num_sdma_sched++; + } + +for (i =0 ; i < adev->vcn.num_vcn_inst; ++i) { +if (adev->vcn.harvest_config & (1 << i)) +continue ; + adev->vcn.vcn_dec_sched[adev->vcn.num_vcn_dec_sched++] = + &adev->vcn.inst[i].ring_dec.sched; + } + +for (i =0 ; i < adev->vcn.num_vcn_inst; ++i) { +if (adev->vcn.harvest_config & (1 << i)) +continue ; +for (j =0 ; j < adev->vcn.num_enc_rings; ++j) + adev->vcn.vcn_enc_sched[adev->vcn.num_vcn_enc_sched++] = + &adev->vcn.inst[i].ring_enc[j].sched; + } + +for (i =0 ; i < adev->jpeg.num_jpeg_inst; ++i) { +if (adev->jpeg.harvest_config & (1 << i)) +continue ; + adev->jpeg.jpeg_sched[adev->jpeg.num_jpeg_sched++] = + &adev->jpeg.inst[i].ring_dec.sched; + } +}
drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h +3 -0
@@ -87,4 +87,7 @@ void amdgpu_ctx_mgr_entity_fini(struct amdgpu_ctx_mgr *mgr);long amdgpu_ctx_mgr_entity_flush (struct amdgpu_ctx_mgr *mgr,long timeout);void amdgpu_ctx_mgr_fini (struct amdgpu_ctx_mgr *mgr); +void amdgpu_ctx_init_sched (struct amdgpu_device *adev); + +#endif
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +2 -0
@@ -3029,6 +3029,8 @@ int amdgpu_device_init(struct amdgpu_device *adev, goto failed; } +amdgpu_ctx_init_sched (adev); + adev->accel_working = true;amdgpu_vm_check_compute_bug (adev);
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h +4 -0
@@ -269,8 +269,12 @@ struct amdgpu_gfx {bool me_fw_write_wait;bool cp_fw_write_wait;struct amdgpu_ring gfx_ring[AMDGPU_MAX_GFX_RINGS ]; +struct drm_gpu_scheduler *gfx_sched[AMDGPU_MAX_GFX_RINGS ]; +uint32_t num_gfx_sched;unsigned num_gfx_rings;struct amdgpu_ring compute_ring[AMDGPU_MAX_COMPUTE_RINGS ]; +struct drm_gpu_scheduler *compute_sched[AMDGPU_MAX_COMPUTE_RINGS ]; +uint32_t num_compute_sched;unsigned num_compute_rings;struct amdgpu_irq_src eop_irq;struct amdgpu_irq_src priv_reg_irq;
drivers/gpu/drm/amd/amdgpu/amdgpu_jpeg.h +2 -0
@@ -43,6 +43,8 @@ struct amdgpu_jpeg {uint8_t num_jpeg_inst;struct amdgpu_jpeg_inst inst[AMDGPU_MAX_JPEG_INSTANCES ];struct amdgpu_jpeg_reg internal; +struct drm_gpu_scheduler *jpeg_sched[AMDGPU_MAX_JPEG_INSTANCES ]; +uint32_t num_jpeg_sched;unsigned harvest_config;struct delayed_work idle_work;enum amd_powergating_state cur_state;
drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h +2 -0
@@ -52,6 +52,8 @@ struct amdgpu_sdma_instance {struct amdgpu_sdma {struct amdgpu_sdma_instance instance[AMDGPU_MAX_SDMA_INSTANCES ]; +struct drm_gpu_scheduler *sdma_sched[AMDGPU_MAX_SDMA_INSTANCES ]; +uint32_t num_sdma_sched;struct amdgpu_irq_src trap_irq;struct amdgpu_irq_src illegal_inst_irq;struct amdgpu_irq_src ecc_irq;
drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h +7 -2
@@ -31,6 +31,7 @@#define AMDGPU_VCN_MAX_ENC_RINGS 3#define AMDGPU_MAX_VCN_INSTANCES 2 +#define AMDGPU_MAX_VCN_ENC_RINGS AMDGPU_VCN_MAX_ENC_RINGS * AMDGPU_MAX_VCN_INSTANCES#define AMDGPU_VCN_HARVEST_VCN0 (1 << 0)#define AMDGPU_VCN_HARVEST_VCN1 (1 << 1)
@@ -186,8 +187,12 @@ struct amdgpu_vcn {uint32_t *dpg_sram_curr_addr;uint8_t num_vcn_inst; -struct amdgpu_vcn_inst inst[AMDGPU_MAX_VCN_INSTANCES ]; -struct amdgpu_vcn_reg internal; +struct amdgpu_vcn_inst inst[AMDGPU_MAX_VCN_INSTANCES ]; +struct amdgpu_vcn_reg internal; +struct drm_gpu_scheduler *vcn_enc_sched[AMDGPU_MAX_VCN_ENC_RINGS ]; +struct drm_gpu_scheduler *vcn_dec_sched[AMDGPU_MAX_VCN_INSTANCES ]; +uint32_t num_vcn_enc_sched; +uint32_t num_vcn_dec_sched;unsigned harvest_config; int (*pause_dpg_mode)(struct amdgpu_device *adev,
--
2.24.0
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
PATCH 4/4 drm/scheduler: do not keep a copy of sched list
In reply to Nirmoy Dasentity should not keep copy and maintain sched list for itself.
--- drivers/gpu/drm/scheduler/sched_entity.c | 19 ++++--------------- 1 file changed, 4 insertions(+), 15 deletions(-)
drivers/gpu/drm/scheduler/sched_entity.c +4 -15
@@ -56,8 +56,6 @@ int drm_sched_entity_init(struct drm_sched_entity *entity, unsigned int num_sched_list, atomic_t *guilty) { - int i; -if (!(entity && sched_list && (num_sched_list ==0 || sched_list[0 ])))return -EINVAL ;
@@ -67,22 +65,14 @@ int drm_sched_entity_init(struct drm_sched_entity *entity, entity->guilty = guilty; entity->num_sched_list = num_sched_list; entity->priority = priority; - entity->sched_list =kcalloc (num_sched_list, -sizeof (struct drm_gpu_scheduler *),GFP_KERNEL ); + entity->sched_list = num_sched_list >1 ? sched_list :NULL ; + entity->last_scheduled =NULL ; -if (!entity->sched_list) -return -ENOMEM ; +if (num_sched_list) + entity->rq = &sched_list[0 ]->sched_rq[entity->priority];init_completion (&entity->entity_idle); -for (i =0 ; i < num_sched_list; i++) - entity->sched_list[i] = sched_list[i]; - -if (num_sched_list) - entity->rq = &entity->sched_list[0 ]->sched_rq[entity->priority]; - - entity->last_scheduled =NULL ; -spin_lock_init (&entity->rq_lock);spsc_queue_init (&entity->job_queue);
@@ -312,7 +302,6 @@ void drm_sched_entity_fini(struct drm_sched_entity *entity)dma_fence_put (entity->last_scheduled); entity->last_scheduled =NULL ; -kfree (entity->sched_list); }EXPORT_SYMBOL (drm_sched_entity_fini);
Signature · 6 lines
2.24.0 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx