diff options
| author | Marek Olšák <marek.olsak@amd.com> | 2021-01-24 01:14:54 -0500 |
|---|---|---|
| committer | Marge Bot <eric+marge@anholt.net> | 2021-02-03 21:53:34 +0000 |
| commit | 586f1397133f2d918fc440c6bfce858d73f5cb80 (patch) | |
| tree | bd88fbdf3d369f2ef9edb0a358db6e7ef36a9733 /src/gallium/winsys/amdgpu/drm | |
| parent | e97af11ba95f2563579414c4c68f2cbd5af4928b (diff) | |
winsys/amdgpu: expand the slab allocation range to [256 B, 1 MB]
This increases the wasted memory to 140 MB for DeusExMD, still below
the original number. The advantage is that we now get 2 MB pages for more
buffers and fewer total buffers allocated by the kernel, enabling faster
GPU page translation and slightly lower kernel overhead.
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8683>
Diffstat (limited to 'src/gallium/winsys/amdgpu/drm')
| -rw-r--r-- | src/gallium/winsys/amdgpu/drm/amdgpu_winsys.c | 4 |
1 files changed, 2 insertions, 2 deletions
diff --git a/src/gallium/winsys/amdgpu/drm/amdgpu_winsys.c b/src/gallium/winsys/amdgpu/drm/amdgpu_winsys.c index 693414abd40..15cbafe69ff 100644 --- a/src/gallium/winsys/amdgpu/drm/amdgpu_winsys.c +++ b/src/gallium/winsys/amdgpu/drm/amdgpu_winsys.c @@ -445,8 +445,8 @@ amdgpu_winsys_create(int fd, const struct pipe_screen_config *config, (aws->info.vram_size + aws->info.gart_size) / 8, amdgpu_bo_destroy, amdgpu_bo_can_reclaim); - unsigned min_slab_order = 9; /* 512 bytes */ - unsigned max_slab_order = 18; /* 256 KB - higher numbers increase memory usage */ + unsigned min_slab_order = 8; /* 256 bytes */ + unsigned max_slab_order = 20; /* 1 MB (slab size = 2 MB) */ unsigned num_slab_orders_per_allocator = (max_slab_order - min_slab_order) / NUM_SLAB_ALLOCATORS; |