It turned out that not the last IB specified is the gang leader,
but instead the last job allocated.
This is a bit unfortunate and not very intuitive for the CS
interface, so try to fix this.
Signed-off-by: Christian König <christian.koenig@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20221115094206.6181-1-christian.koenig@amd.com
Tested-by: Timur Kristóf <timur.kristof@gmail.com>
Acked-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Fixes:
4624459c84d7 ("drm/amdgpu: add gang submit frontend v6")
return r;
++(num_ibs[r]);
return r;
++(num_ibs[r]);
+ p->gang_leader_idx = r;
if (ret)
goto free_all_kdata;
}
if (ret)
goto free_all_kdata;
}
- p->gang_leader = p->jobs[p->gang_size - 1];
+ p->gang_leader = p->jobs[p->gang_leader_idx];
if (p->ctx->vram_lost_counter != p->gang_leader->vram_lost_counter) {
ret = -ECANCELED;
if (p->ctx->vram_lost_counter != p->gang_leader->vram_lost_counter) {
ret = -ECANCELED;
- for (i = 0; i < p->gang_size - 1; ++i) {
+ for (i = 0; i < p->gang_size; ++i) {
+ if (p->jobs[i] == leader)
+ continue;
+
r = amdgpu_sync_clone(&leader->sync, &p->jobs[i]->sync);
if (r)
return r;
}
r = amdgpu_sync_clone(&leader->sync, &p->jobs[i]->sync);
if (r)
return r;
}
- r = amdgpu_ctx_wait_prev_fence(p->ctx, p->entities[p->gang_size - 1]);
+ r = amdgpu_ctx_wait_prev_fence(p->ctx, p->entities[p->gang_leader_idx]);
if (r && r != -ERESTARTSYS)
DRM_ERROR("amdgpu_ctx_wait_prev_fence failed.\n");
if (r && r != -ERESTARTSYS)
DRM_ERROR("amdgpu_ctx_wait_prev_fence failed.\n");
for (i = 0; i < p->gang_size; ++i)
drm_sched_job_arm(&p->jobs[i]->base);
for (i = 0; i < p->gang_size; ++i)
drm_sched_job_arm(&p->jobs[i]->base);
- for (i = 0; i < (p->gang_size - 1); ++i) {
+ for (i = 0; i < p->gang_size; ++i) {
+ if (p->jobs[i] == leader)
+ continue;
+
fence = &p->jobs[i]->base.s_fence->scheduled;
r = amdgpu_sync_fence(&leader->sync, fence);
if (r)
fence = &p->jobs[i]->base.s_fence->scheduled;
r = amdgpu_sync_fence(&leader->sync, fence);
if (r)
list_for_each_entry(e, &p->validated, tv.head) {
/* Everybody except for the gang leader uses READ */
list_for_each_entry(e, &p->validated, tv.head) {
/* Everybody except for the gang leader uses READ */
- for (i = 0; i < (p->gang_size - 1); ++i) {
+ for (i = 0; i < p->gang_size; ++i) {
+ if (p->jobs[i] == leader)
+ continue;
+
dma_resv_add_fence(e->tv.bo->base.resv,
&p->jobs[i]->base.s_fence->finished,
DMA_RESV_USAGE_READ);
dma_resv_add_fence(e->tv.bo->base.resv,
&p->jobs[i]->base.s_fence->finished,
DMA_RESV_USAGE_READ);
- seq = amdgpu_ctx_add_fence(p->ctx, p->entities[p->gang_size - 1],
+ seq = amdgpu_ctx_add_fence(p->ctx, p->entities[p->gang_leader_idx],
p->fence);
amdgpu_cs_post_dependencies(p);
p->fence);
amdgpu_cs_post_dependencies(p);
/* scheduler job objects */
unsigned int gang_size;
/* scheduler job objects */
unsigned int gang_size;
+ unsigned int gang_leader_idx;
struct drm_sched_entity *entities[AMDGPU_CS_GANG_SIZE];
struct amdgpu_job *jobs[AMDGPU_CS_GANG_SIZE];
struct amdgpu_job *gang_leader;
struct drm_sched_entity *entities[AMDGPU_CS_GANG_SIZE];
struct amdgpu_job *jobs[AMDGPU_CS_GANG_SIZE];
struct amdgpu_job *gang_leader;