[PATCH 6/6] drm/amdgpu: use more than 64KB fragment size if possible

[Alex Deucher]

On Tue, Aug 9, 2016 at 8:52 AM, Christian König <deathsimple at vodafone.de> wrote:
> From: Christian König <christian.koenig at amd.com>
>
> We align to 64KB, but when userspace aligns even more we can easily use more.
>
> Signed-off-by: Christian König <christian.koenig at amd.com>

For the series:
Reviewed-by: Alex Deucher <alexander.deucher at amd.com>

> ---
>  drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 12 ++++++++----
>  1 file changed, 8 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
> index e6c030b..88f4109 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
> @@ -817,13 +817,13 @@ static void amdgpu_vm_frag_ptes(struct amdgpu_pte_update_params   *params,
>          * allocation size to the fragment size.
>          */
>
> -       /* SI and newer are optimized for 64KB */
> -       uint64_t frag_flags = AMDGPU_PTE_FRAG(AMDGPU_LOG2_PAGES_PER_FRAG);
> -       uint64_t frag_align = 1 << AMDGPU_LOG2_PAGES_PER_FRAG;
> +       const uint64_t frag_align = 1 << AMDGPU_LOG2_PAGES_PER_FRAG;
>
>         uint64_t frag_start = ALIGN(start, frag_align);
>         uint64_t frag_end = end & ~(frag_align - 1);
>
> +       uint32_t frag;
> +
>         /* system pages are non continuously */
>         if (params->src || params->pages_addr || !(flags & AMDGPU_PTE_VALID) ||
>             (frag_start >= frag_end)) {
> @@ -832,6 +832,10 @@ static void amdgpu_vm_frag_ptes(struct amdgpu_pte_update_params    *params,
>                 return;
>         }
>
> +       /* use more than 64KB fragment size if possible */
> +       frag = lower_32_bits(frag_start | frag_end);
> +       frag = likely(frag) ? __ffs(frag) : 31;
> +
>         /* handle the 4K area at the beginning */
>         if (start != frag_start) {
>                 amdgpu_vm_update_ptes(params, vm, start, frag_start,
> @@ -841,7 +845,7 @@ static void amdgpu_vm_frag_ptes(struct amdgpu_pte_update_params     *params,
>
>         /* handle the area in the middle */
>         amdgpu_vm_update_ptes(params, vm, frag_start, frag_end, dst,
> -                             flags | frag_flags);
> +                             flags | AMDGPU_PTE_FRAG(frag));
>
>         /* handle the 4K area at the end */
>         if (frag_end != end) {
> --
> 2.5.0
>
> _______________________________________________
> amd-gfx mailing list
> amd-gfx at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/amd-gfx