[Mesa-dev] [PATCH] radv: ignore the loadOp if the first use of an attachment is a resolve

[Bas Nieuwenhuizen]

On Wed, May 22, 2019 at 11:20 AM Samuel Pitoiset
<samuel.pitoiset at gmail.com> wrote:
>
> Based on ANV.
>
> Signed-off-by: Samuel Pitoiset <samuel.pitoiset at gmail.com>
> ---
>  src/amd/vulkan/radv_meta_resolve.c | 21 ++++++++++++---------
>  1 file changed, 12 insertions(+), 9 deletions(-)
>
> diff --git a/src/amd/vulkan/radv_meta_resolve.c b/src/amd/vulkan/radv_meta_resolve.c
> index ade5d438438..6a8abce1ddb 100644
> --- a/src/amd/vulkan/radv_meta_resolve.c
> +++ b/src/amd/vulkan/radv_meta_resolve.c
> @@ -618,15 +618,6 @@ radv_cmd_buffer_resolve_subpass(struct radv_cmd_buffer *cmd_buffer)
>         struct radv_meta_saved_state saved_state;
>         enum radv_resolve_method resolve_method = RESOLVE_HW;
>
> -       /* FINISHME(perf): Skip clears for resolve attachments.
> -        *
> -        * From the Vulkan 1.0 spec:
> -        *
> -        *    If the first use of an attachment in a render pass is as a resolve
> -        *    attachment, then the loadOp is effectively ignored as the resolve is
> -        *    guaranteed to overwrite all pixels in the render area.
> -        */
> -
>         if (!subpass->has_resolve)
>                 return;
>
> @@ -637,6 +628,18 @@ radv_cmd_buffer_resolve_subpass(struct radv_cmd_buffer *cmd_buffer)
>                 if (dest_att.attachment == VK_ATTACHMENT_UNUSED)
>                         continue;
>
> +               if (cmd_buffer->state.attachments[dest_att.attachment].pending_clear_aspects) {
> +                       /* From the Vulkan 1.0 spec:
> +                        *
> +                        *    If the first use of an attachment in a render
> +                        *    pass is as a resolve attachment, then the loadOp
> +                        *    is effectively ignored as the resolve is
> +                        *    guaranteed to overwrite all pixels in the render
> +                        *    area.
> +                       */
> +                       cmd_buffer->state.attachments[dest_att.attachment].pending_clear_aspects = 0;
> +               }
> +

You can drop the if statement. Also this seems like a correctness
issue, as we only cleared the first time it was used as a color
attachment, which might happen after a resolve?

>                 struct radv_image *dst_img = cmd_buffer->state.framebuffer->attachments[dest_att.attachment].attachment->image;
>                 struct radv_image *src_img = cmd_buffer->state.framebuffer->attachments[src_att.attachment].attachment->image;
>
> --
> 2.21.0
>
> _______________________________________________
> mesa-dev mailing list
> mesa-dev at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/mesa-dev