drm/amdgpu: stop mapping BOs to GTT
authorChristian König <christian.koenig@amd.com>
Fri, 30 Jun 2017 10:19:42 +0000 (12:19 +0200)
committerAlex Deucher <alexander.deucher@amd.com>
Fri, 14 Jul 2017 15:06:21 +0000 (11:06 -0400)
No need to map BOs to GTT on eviction and intermediate transfers any more.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c

index 4a34b61..fb9c698 100644 (file)
@@ -199,7 +199,6 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo,
                .lpfn = 0,
                .flags = TTM_PL_MASK_CACHING | TTM_PL_FLAG_SYSTEM
        };
-       unsigned i;
 
        if (!amdgpu_ttm_bo_is_amdgpu_bo(bo)) {
                placement->placement = &placements;
@@ -217,20 +216,6 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo,
                        amdgpu_ttm_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_CPU);
                } else {
                        amdgpu_ttm_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_GTT);
-                       for (i = 0; i < abo->placement.num_placement; ++i) {
-                               if (!(abo->placements[i].flags &
-                                     TTM_PL_FLAG_TT))
-                                       continue;
-
-                               if (abo->placements[i].lpfn)
-                                       continue;
-
-                               /* set an upper limit to force directly
-                                * allocating address space for the BO.
-                                */
-                               abo->placements[i].lpfn =
-                                       adev->mc.gtt_size >> PAGE_SHIFT;
-                       }
                }
                break;
        case TTM_PL_TT:
@@ -391,7 +376,7 @@ static int amdgpu_move_vram_ram(struct ttm_buffer_object *bo,
        placement.num_busy_placement = 1;
        placement.busy_placement = &placements;
        placements.fpfn = 0;
-       placements.lpfn = adev->mc.gtt_size >> PAGE_SHIFT;
+       placements.lpfn = 0;
        placements.flags = TTM_PL_MASK_CACHING | TTM_PL_FLAG_TT;
        r = ttm_bo_mem_space(bo, &placement, &tmp_mem,
                             interruptible, no_wait_gpu);
@@ -438,7 +423,7 @@ static int amdgpu_move_ram_vram(struct ttm_buffer_object *bo,
        placement.num_busy_placement = 1;
        placement.busy_placement = &placements;
        placements.fpfn = 0;
-       placements.lpfn = adev->mc.gtt_size >> PAGE_SHIFT;
+       placements.lpfn = 0;
        placements.flags = TTM_PL_MASK_CACHING | TTM_PL_FLAG_TT;
        r = ttm_bo_mem_space(bo, &placement, &tmp_mem,
                             interruptible, no_wait_gpu);