radv: move is_local up to the winsys level.

We can avoid adding the buffer in the non-local case, this will
avoid all the overhead of the indirect call.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

src/amd/vulkan/radv_radeon_winsys.h

@@ -147,6 +147,7 @@ struct radeon_winsys_fence;
 
 struct radeon_winsys_bo {
 	uint64_t va;
+	bool is_local;
 };
 struct radv_winsys_sem_counts {
 	uint32_t syncobj_count;
@@ -284,6 +285,9 @@ static inline void radv_cs_add_buffer(struct radeon_winsys *ws,
 				      struct radeon_winsys_bo *bo,
 				      uint8_t priority)
 {
+	if (bo->is_local)
+		return;
+
 	ws->cs_add_buffer(cs, bo, priority);
 }
 

src/amd/vulkan/winsys/amdgpu/radv_amdgpu_bo.c

@@ -333,7 +333,7 @@ radv_amdgpu_winsys_bo_create(struct radeon_winsys *_ws,
 	if (!(flags & RADEON_FLAG_IMPLICIT_SYNC) && ws->info.drm_minor >= 22)
 		request.flags |= AMDGPU_GEM_CREATE_EXPLICIT_SYNC;
 	if (flags & RADEON_FLAG_NO_INTERPROCESS_SHARING && ws->info.drm_minor >= 20) {
-		bo->is_local = true;
+		bo->base.is_local = true;
 		request.flags |= AMDGPU_GEM_CREATE_VM_ALWAYS_VALID;
 	}
 

src/amd/vulkan/winsys/amdgpu/radv_amdgpu_bo.h

@@ -45,7 +45,6 @@ struct radv_amdgpu_winsys_bo {
 	uint64_t size;
 	struct radv_amdgpu_winsys *ws;
 	bool is_virtual;
-	bool is_local;
 	int ref_count;
 
 	union {

src/amd/vulkan/winsys/amdgpu/radv_amdgpu_cs.c

@@ -473,7 +473,7 @@ static void radv_amdgpu_cs_add_buffer(struct radeon_winsys_cs *_cs,
 		return;
 	}
 
-	if (bo->is_local)
+	if (bo->base.is_local)
 		return;
 
 	radv_amdgpu_cs_add_buffer_internal(cs, bo->bo, priority);