On Wed, Jun 21, 2017 at 06:19:57PM -0700, Jason Ekstrand wrote:
> On Wed, Jun 21, 2017 at 5:15 PM, Nanley Chery <nanleych...@gmail.com> wrote:
>
> > In the future, we plan on using this method to resolve images whose
> > surface state fast-clear value is dynamically updated during command
> > buffer execution. Start using it now for testing and to reduce churn
> > later on.
> >
> > Signed-off-by: Nanley Chery <nanley.g.ch...@intel.com>
> > Reviewed-by: Iago Toral Quiroga <ito...@igalia.com>
> > ---
> > src/intel/vulkan/anv_blorp.c | 29 ++++++++++-------------------
> > 1 file changed, 10 insertions(+), 19 deletions(-)
> >
> > diff --git a/src/intel/vulkan/anv_blorp.c b/src/intel/vulkan/anv_blorp.c
> > index c4c744868c..7f6ed0efe4 100644
> > --- a/src/intel/vulkan/anv_blorp.c
> > +++ b/src/intel/vulkan/anv_blorp.c
> > @@ -1493,7 +1493,6 @@ anv_image_ccs_clear(struct anv_cmd_buffer
> > *cmd_buffer,
> >
> > static void
> > ccs_resolve_attachment(struct anv_cmd_buffer *cmd_buffer,
> > - struct blorp_batch *batch,
> > uint32_t att)
> > {
> > struct anv_framebuffer *fb = cmd_buffer->state.framebuffer;
> > @@ -1589,12 +1588,6 @@ ccs_resolve_attachment(struct anv_cmd_buffer
> > *cmd_buffer,
> > if (resolve_op == BLORP_FAST_CLEAR_OP_NONE)
> > return;
> >
> > - struct blorp_surf surf;
> > - get_blorp_surf_for_anv_image(image, VK_IMAGE_ASPECT_COLOR_BIT,
> > - att_state->aux_usage, &surf);
> > - if (att_state->fast_clear)
> > - surf.clear_color = vk_to_isl_color(att_state->clear_value.color);
> > -
> > /* From the Sky Lake PRM Vol. 7, "Render Target Resolve":
> > *
> > * "When performing a render target resolve, PIPE_CONTROL with end
> > of
> > @@ -1610,12 +1603,8 @@ ccs_resolve_attachment(struct anv_cmd_buffer
> > *cmd_buffer,
> > cmd_buffer->state.pending_pipe_bits |=
> > ANV_PIPE_RENDER_TARGET_CACHE_FLUSH_BIT | ANV_PIPE_CS_STALL_BIT;
> >
> > - for (uint32_t layer = 0; layer < fb->layers; layer++) {
> > - blorp_ccs_resolve(batch, &surf,
> > - iview->isl.base_level,
> > - iview->isl.base_array_layer + layer,
> > - iview->isl.format, resolve_op);
> > - }
> > + anv_ccs_resolve(cmd_buffer, att_state->color_rt_state, image,
> > + iview->isl.base_level, fb->layers, resolve_op);
> >
> > cmd_buffer->state.pending_pipe_bits |=
> > ANV_PIPE_RENDER_TARGET_CACHE_FLUSH_BIT | ANV_PIPE_CS_STALL_BIT;
> > @@ -1633,16 +1622,13 @@ anv_cmd_buffer_resolve_subpass(struct
> > anv_cmd_buffer *cmd_buffer)
> > struct anv_subpass *subpass = cmd_buffer->state.subpass;
> >
> >
> > - struct blorp_batch batch;
> > - blorp_batch_init(&cmd_buffer->device->blorp, &batch, cmd_buffer, 0);
> > -
> > for (uint32_t i = 0; i < subpass->color_count; ++i) {
> > const uint32_t att = subpass->color_attachments[i].attachment;
> > if (att == VK_ATTACHMENT_UNUSED)
> > continue;
> >
> > assert(att < cmd_buffer->state.pass->attachment_count);
> > - ccs_resolve_attachment(cmd_buffer, &batch, att);
> > + ccs_resolve_attachment(cmd_buffer, att);
> > }
> >
> > if (subpass->has_resolve) {
> > @@ -1681,6 +1667,10 @@ anv_cmd_buffer_resolve_subpass(struct
> > anv_cmd_buffer *cmd_buffer)
> > const VkRect2D render_area = cmd_buffer->state.render_area;
> >
> > assert(src_iview->aspect_mask == dst_iview->aspect_mask);
> > +
> > + struct blorp_batch batch;
> > + blorp_batch_init(&cmd_buffer->device->blorp, &batch,
> > cmd_buffer, 0);
> > +
> > resolve_image(&batch, src_iview->image,
> > src_iview->isl.base_level,
> > src_iview->isl.base_array_layer,
> > @@ -1692,11 +1682,12 @@ anv_cmd_buffer_resolve_subpass(struct
> > anv_cmd_buffer *cmd_buffer)
> > render_area.offset.x, render_area.offset.y,
> > render_area.extent.width,
> > render_area.extent.height);
> >
> > - ccs_resolve_attachment(cmd_buffer, &batch, dst_att);
> > + blorp_batch_finish(&batch);
> >
>
> At some point, I'd like to make blorp smarter so that it can avoid
> re-emitting all of the state on every blorp op. When that happens, we'll
> want to figure out how to move blorp_batch_init/finish back out so that
> they happen outside of the loop. Not a big deal today though.
>
>
Thankfully, we can remove the loop when we eventually make the resolves
happen only on layout transitions.
> > +
> > + ccs_resolve_attachment(cmd_buffer, dst_att);
> > }
> > }
> >
> > - blorp_batch_finish(&batch);
> > }
> >
> > void
> > --
> > 2.13.1
> >
> > _______________________________________________
> > mesa-dev mailing list
> > mesa-dev@lists.freedesktop.org
> > https://lists.freedesktop.org/mailman/listinfo/mesa-dev
> >
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/mesa-dev
