UbiquitousOS
/
OpenCloudOS-Kernel
mirror of https://gitee.com/OpenCloudOS/OpenCloudOS-Kernel.git
11
0
Fork 0
Code Issues Projects Releases Wiki Activity

drm/amdgpu: move explicit sync check into the CS 

This moves the memory allocation out of the critical code path.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Luben Tuikov <luben.tuikov@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20221014084641.128280-8-christian.koenig@amd.com
This commit is contained in:
Christian König 2022-09-29 13:05:56 +02:00
parent f7d66fb2ea
commit 1b2d5eda5a
4 changed files with 17 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
View File

@ -449,8 +449,19 @@ static int amdgpu_syncobj_lookup_and_add(struct amdgpu_cs_parser *p,
} }
r = amdgpu_sync_fence(&p->gang_leader->sync, fence); r = amdgpu_sync_fence(&p->gang_leader->sync, fence);
dma_fence_put(fence); if (r)
goto error;
/*
* When we have an explicit dependency it might be necessary to insert a
* pipeline sync to make sure that all caches etc are flushed and the
* next job actually sees the results from the previous one.
*/
if (fence->context == p->gang_leader->base.entity->fence_context)
r = amdgpu_sync_fence(&p->gang_leader->explicit_sync, fence);
error:
dma_fence_put(fence);
return r; return r;
} }

2
drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c
View File

@ -182,7 +182,7 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs,
need_ctx_switch = ring->current_ctx != fence_ctx; need_ctx_switch = ring->current_ctx != fence_ctx;
if (ring->funcs->emit_pipeline_sync && job && if (ring->funcs->emit_pipeline_sync && job &&
((tmp = amdgpu_sync_get_fence(&job->sched_sync)) || ((tmp = amdgpu_sync_get_fence(&job->explicit_sync)) ||
(amdgpu_sriov_vf(adev) && need_ctx_switch) || (amdgpu_sriov_vf(adev) && need_ctx_switch) ||
amdgpu_vm_need_pipeline_sync(ring, job))) { amdgpu_vm_need_pipeline_sync(ring, job))) {
need_pipe_sync = true; need_pipe_sync = true;

12
drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
View File

@ -108,7 +108,7 @@ int amdgpu_job_alloc(struct amdgpu_device *adev, struct amdgpu_vm *vm,
(*job)->vm = vm; (*job)->vm = vm;
amdgpu_sync_create(&(*job)->sync); amdgpu_sync_create(&(*job)->sync);
amdgpu_sync_create(&(*job)->sched_sync); amdgpu_sync_create(&(*job)->explicit_sync);
(*job)->vram_lost_counter = atomic_read(&adev->vram_lost_counter); (*job)->vram_lost_counter = atomic_read(&adev->vram_lost_counter);
(*job)->vm_pd_addr = AMDGPU_BO_INVALID_OFFSET; (*job)->vm_pd_addr = AMDGPU_BO_INVALID_OFFSET;
@ -176,7 +176,7 @@ static void amdgpu_job_free_cb(struct drm_sched_job *s_job)
drm_sched_job_cleanup(s_job); drm_sched_job_cleanup(s_job);
amdgpu_sync_free(&job->sync); amdgpu_sync_free(&job->sync);
amdgpu_sync_free(&job->sched_sync); amdgpu_sync_free(&job->explicit_sync);
dma_fence_put(&job->hw_fence); dma_fence_put(&job->hw_fence);
} }
@ -204,7 +204,7 @@ void amdgpu_job_free(struct amdgpu_job *job)
amdgpu_job_free_resources(job); amdgpu_job_free_resources(job);
amdgpu_sync_free(&job->sync); amdgpu_sync_free(&job->sync);
amdgpu_sync_free(&job->sched_sync); amdgpu_sync_free(&job->explicit_sync);
if (job->gang_submit != &job->base.s_fence->scheduled) if (job->gang_submit != &job->base.s_fence->scheduled)
dma_fence_put(job->gang_submit); dma_fence_put(job->gang_submit);
@ -251,12 +251,6 @@ amdgpu_job_dependency(struct drm_sched_job *sched_job,
int r; int r;
fence = amdgpu_sync_get_fence(&job->sync); fence = amdgpu_sync_get_fence(&job->sync);
if (fence && drm_sched_dependency_optimized(fence, s_entity)) {
r = amdgpu_sync_fence(&job->sched_sync, fence);
if (r)
DRM_ERROR("Error adding fence (%d)\n", r);
}
while (!fence && job->vm && !job->vmid) { while (!fence && job->vm && !job->vmid) {
r = amdgpu_vmid_grab(job->vm, ring, job, &fence); r = amdgpu_vmid_grab(job->vm, ring, job, &fence);
if (r) if (r)

2
drivers/gpu/drm/amd/amdgpu/amdgpu_job.h
View File

@ -48,7 +48,7 @@ struct amdgpu_job {
struct drm_sched_job base; struct drm_sched_job base;
struct amdgpu_vm *vm; struct amdgpu_vm *vm;
struct amdgpu_sync sync; struct amdgpu_sync sync;
struct amdgpu_sync sched_sync; struct amdgpu_sync explicit_sync;
struct dma_fence hw_fence; struct dma_fence hw_fence;
struct dma_fence *gang_submit; struct dma_fence *gang_submit;
uint32_t preamble_status; uint32_t preamble_status;