drm/amdgpu: use the maximum possible fragment size on Vega/Raven

The fragment size controls only the L1 on Vega/Raven and we now don't
have any extra overhead any more because of larger fragments.

Signed-off-by: Christian König <[email protected]>
Reviewed-by: Huang Rui <[email protected]>
Reviewed-by: Felix Kuehling <[email protected]>
Acked-by: Junwei Zhang <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
index 45343501c1f3079f012e6c41b04f543487624362..2d2d6197c3dcc3128d22939b8c7b246f6f72f754 100644 (file)
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
@@ -1540,8 +1540,16 @@ static void amdgpu_vm_fragment(struct amdgpu_pte_update_params *params,
         * larger. Thus, we try to use large fragments wherever possible.
         * Userspace can support this by aligning virtual base address and
         * allocation size to the fragment size.
+        *
+        * Starting with Vega10 the fragment size only controls the L1. The L2
+        * is now directly feed with small/huge/giant pages from the walker.
         */
-       unsigned max_frag = params->adev->vm_manager.fragment_size;
+       unsigned max_frag;
+
+       if (params->adev->asic_type < CHIP_VEGA10)
+               max_frag = params->adev->vm_manager.fragment_size;
+       else
+               max_frag = 31;
 
        /* system pages are non continuously */
        if (params->src || !(flags & AMDGPU_PTE_VALID)) {