fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-01-04 15:40:11 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

radv/amdgpu: skip adding per VM BOs for sparse during CS BO list build

Browse source 
This should be similar but it will help for re-introducing
OP_MAP/ON_UNMAP instead of OP_REPLACE.

Cc: mesa-stable
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23741>
This commit is contained in:
Samuel Pitoiset 2023-06-20 12:52:26 +02:00 • committed by Marge Bot
parent 035e55f5ac
commit 51caece74c
2 changed files with 14 additions and 15 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

15
src/amd/vulkan/winsys/amdgpu/radv_amdgpu_bo.c
View file

@ -136,21 +136,6 @@ radv_amdgpu_winsys_bo_virtual_bind(struct radeon_winsys *_ws, struct radeon_wins
return VK_ERROR_OUT_OF_DEVICE_MEMORY;
}
/* Do not add the BO to the virtual BO list if it's already in the global list to avoid dangling
* BO references because it might have been destroyed without being previously unbound. Resetting
* it to NULL clears the old BO ranges if present.
*
* This is going to be clarified in the Vulkan spec:
* https://gitlab.khronos.org/vulkan/vulkan/-/issues/3125
*
* The issue still exists for non-global BO but it will be addressed later, once we are 100% it's
* RADV fault (mostly because the solution looks more complicated).
*/
if (bo && bo->base.use_global_list) {
bo = NULL;
bo_offset = 0;
}
/* We have at most 2 new ranges (1 by the bind, and another one by splitting a range that
* contains the newly bound range). */
if (parent->range_capacity - parent->range_count < 2) {

14
src/amd/vulkan/winsys/amdgpu/radv_amdgpu_cs.c
View file

@ -787,6 +787,20 @@ radv_amdgpu_add_cs_to_bo_list(struct radv_amdgpu_cs *cs, struct drm_amdgpu_bo_li
for (unsigned k = 0; k < virtual_bo->bo_count; ++k) {
struct radv_amdgpu_winsys_bo *bo = virtual_bo->bos[k];
bool found = false;
/* Do not add the BO to the virtual BO list if it's already in the global list to avoid
* dangling BO references because it might have been destroyed without being previously
* unbound. Resetting it to NULL clears the old BO ranges if present.
*
* This is going to be clarified in the Vulkan spec:
* https://gitlab.khronos.org/vulkan/vulkan/-/issues/3125
*
* The issue still exists for non-global BO but it will be addressed later, once we are
* 100% it's RADV fault (mostly because the solution looks more complicated).
*/
if (bo->base.use_global_list)
continue;
for (unsigned m = 0; m < num_handles; ++m) {
if (handles[m].bo_handle == bo->bo_handle) {
found = true;