Module: Mesa Branch: main Commit: a94319d29b466c48ff0e00d6f986e36b99d2ed7b URL: http://cgit.freedesktop.org/mesa/mesa/commit/?id=a94319d29b466c48ff0e00d6f986e36b99d2ed7b
Author: Marek Olšák <marek.ol...@amd.com> Date: Sun Jan 7 15:11:26 2024 -0500 winsys/amdgpu: don't clear buffer list elements after IB submission amdgpu_winsys_bo_reference(dst, NULL) sets *dst to NULL, but we never read *dst again because we set num_buffers to 0. So don't touch the buffer list elements and only decrement the BO refcount. It makes a difference when you have 10k BOs. The CS thread overhead in VP2020/Catia1: 9.23% -> 8.74% Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-pra...@amd.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/27060> --- src/gallium/include/winsys/radeon_winsys.h | 16 ++++++++++++++++ src/gallium/winsys/amdgpu/drm/amdgpu_bo.h | 14 ++++++++++++++ src/gallium/winsys/amdgpu/drm/amdgpu_cs.c | 8 ++++---- 3 files changed, 34 insertions(+), 4 deletions(-) diff --git a/src/gallium/include/winsys/radeon_winsys.h b/src/gallium/include/winsys/radeon_winsys.h index 22d10cbe370..02f7ee48d6d 100644 --- a/src/gallium/include/winsys/radeon_winsys.h +++ b/src/gallium/include/winsys/radeon_winsys.h @@ -792,6 +792,22 @@ radeon_bo_reference(struct radeon_winsys *rws, struct pb_buffer_lean **dst, *dst = src; } +/* Same as radeon_bo_reference, but ignore the value in *dst. */ +static inline void +radeon_bo_set_reference(struct pb_buffer_lean **dst, struct pb_buffer_lean *src) +{ + *dst = src; + pipe_reference(NULL, &src->reference); /* only increment refcount */ +} + +/* Unreference dst, but don't assign anything. */ +static inline void +radeon_bo_drop_reference(struct radeon_winsys *rws, struct pb_buffer_lean *dst) +{ + if (pipe_reference(&dst->reference, NULL)) /* only decrement refcount */ + rws->buffer_destroy(rws, dst); +} + /* The following bits describe the heaps managed by slab allocators (pb_slab) and * the allocation cache (pb_cache). */ diff --git a/src/gallium/winsys/amdgpu/drm/amdgpu_bo.h b/src/gallium/winsys/amdgpu/drm/amdgpu_bo.h index 9d2bb04f66e..1fcb233d921 100644 --- a/src/gallium/winsys/amdgpu/drm/amdgpu_bo.h +++ b/src/gallium/winsys/amdgpu/drm/amdgpu_bo.h @@ -265,4 +265,18 @@ amdgpu_winsys_bo_reference(struct amdgpu_winsys *ws, struct amdgpu_winsys_bo **d (struct pb_buffer_lean**)dst, (struct pb_buffer_lean*)src); } +/* Same as amdgpu_winsys_bo_reference, but ignore the value in *dst. */ +static inline void +amdgpu_winsys_bo_set_reference(struct amdgpu_winsys_bo **dst, struct amdgpu_winsys_bo *src) +{ + radeon_bo_set_reference((struct pb_buffer_lean**)dst, (struct pb_buffer_lean*)src); +} + +/* Unreference dst, but don't assign anything. */ +static inline void +amdgpu_winsys_bo_drop_reference(struct amdgpu_winsys *ws, struct amdgpu_winsys_bo *dst) +{ + radeon_bo_drop_reference(&ws->dummy_ws.base, &dst->base); +} + #endif diff --git a/src/gallium/winsys/amdgpu/drm/amdgpu_cs.c b/src/gallium/winsys/amdgpu/drm/amdgpu_cs.c index 3bef8a1ed28..4b52c99894a 100644 --- a/src/gallium/winsys/amdgpu/drm/amdgpu_cs.c +++ b/src/gallium/winsys/amdgpu/drm/amdgpu_cs.c @@ -617,7 +617,7 @@ amdgpu_do_add_buffer(struct amdgpu_cs_context *cs, struct amdgpu_winsys_bo *bo, struct amdgpu_cs_buffer *buffer = &list->buffers[idx]; memset(buffer, 0, sizeof(*buffer)); - amdgpu_winsys_bo_reference(cs->ws, &buffer->bo, bo); + amdgpu_winsys_bo_set_reference(&buffer->bo, bo); list->num_buffers++; unsigned hash = bo->unique_id & (BUFFER_HASHLIST_SIZE-1); @@ -882,7 +882,7 @@ static void amdgpu_cs_context_cleanup_buffers(struct amdgpu_winsys *ws, struct a unsigned num_buffers = cs->buffer_lists[i].num_buffers; for (unsigned j = 0; j < num_buffers; j++) - amdgpu_winsys_bo_reference(ws, &buffers[j].bo, NULL); + amdgpu_winsys_bo_drop_reference(ws, buffers[j].bo); cs->buffer_lists[i].num_buffers = 0; } @@ -1638,10 +1638,10 @@ cleanup: for (i = 0; i < num_dec_buffers; i++) { p_atomic_dec(&buffers[i].bo->num_active_ioctls); - amdgpu_winsys_bo_reference(ws, &buffers[i].bo, NULL); + amdgpu_winsys_bo_drop_reference(ws, buffers[i].bo); } for (; i < num_buffers; i++) - amdgpu_winsys_bo_reference(ws, &buffers[i].bo, NULL); + amdgpu_winsys_bo_drop_reference(ws, buffers[i].bo); cs->buffer_lists[list].num_buffers = 0; }
