[Mesa-dev] [PATCH] r600: increase performance for DRI PRIME offloading if 2nd GPU is Evergreen+

[Marek Olšák]

Applied. Thanks!

Marek

On Fri, Aug 26, 2016 at 6:59 PM, Mario Kleiner
<mario.kleiner.de at gmail.com> wrote:
> This is a direct port of Marek Olšáks patch
> "radeonsi: increase performance for DRI PRIME
> offloading if 2nd GPU is CIK or VI" to r600.
>
> It uses SDMA for the detiling blit from renderoffload VRAM
> to GTT, as SDMA is much faster for tiled->linear blits from
> VRAM to GTT.
>
> Testing on a dual Radeon HD-5770 setup reduced the time
> for the render offload gpu to get its rendering into
> system RAM from approximately 16 msecs for simple rendering
> at 1920x1080 pixel 32 bpp to 5 msecs, a > 3x speedup!
>
> This was measured using ftrace to trace the time the radeon kms
> driver waited on the dmabuf fence of the renderoffload gpu to
> complete.
>
> All in all this brought the time for a flip down from 20 msecs
> to 9 msecs, so the prime setup can display at full 60 fps instead
> of barely 30 fps vsync'ed.
>
> The current r600 implementation supports SDMA on Evergreen and
> later, but not R600/R700 due to some bugs apparently present
> in their SDMA implementation.
>
> Signed-off-by: Mario Kleiner <mario.kleiner.de at gmail.com>
> Cc: Marek Olšák <marek.olsak at amd.com>
> ---
>  src/gallium/drivers/r600/r600_blit.c | 19 +++++++++++++++++++
>  1 file changed, 19 insertions(+)
>
> diff --git a/src/gallium/drivers/r600/r600_blit.c b/src/gallium/drivers/r600/r600_blit.c
> index adf616e..8fdc51c 100644
> --- a/src/gallium/drivers/r600/r600_blit.c
> +++ b/src/gallium/drivers/r600/r600_blit.c
> @@ -850,11 +850,30 @@ static void r600_blit(struct pipe_context *ctx,
>                        const struct pipe_blit_info *info)
>  {
>         struct r600_context *rctx = (struct r600_context*)ctx;
> +       struct r600_texture *rdst = (struct r600_texture *)info->dst.resource;
>
>         if (do_hardware_msaa_resolve(ctx, info)) {
>                 return;
>         }
>
> +       /* Using SDMA for copying to a linear texture in GTT is much faster.
> +        * This improves DRI PRIME performance.
> +        *
> +        * resource_copy_region can't do this yet, because dma_copy calls it
> +        * on failure (recursion).
> +        */
> +       if (rdst->surface.level[info->dst.level].mode ==
> +           RADEON_SURF_MODE_LINEAR_ALIGNED &&
> +           rctx->b.dma_copy &&
> +           util_can_blit_via_copy_region(info, false)) {
> +               rctx->b.dma_copy(ctx, info->dst.resource, info->dst.level,
> +                                info->dst.box.x, info->dst.box.y,
> +                                info->dst.box.z,
> +                                info->src.resource, info->src.level,
> +                                &info->src.box);
> +               return;
> +       }
> +
>         assert(util_blitter_is_blit_supported(rctx->blitter, info));
>
>         /* The driver doesn't decompress resources automatically while
> --
> 2.7.4
>
> _______________________________________________
> mesa-dev mailing list
> mesa-dev at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/mesa-dev