<div dir="ltr"><div class="gmail_extra"><div class="gmail_quote">On Wed, Jun 21, 2017 at 5:15 PM, Nanley Chery <span dir="ltr"><<a href="mailto:nanleychery@gmail.com" target="_blank">nanleychery@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><span class="">In the future, we plan on using this method to resolve images whose<br>
surface state fast-clear value is dynamically updated during command<br>
buffer execution. Start using it now for testing and to reduce churn<br>
later on.<br>
<br>
Signed-off-by: Nanley Chery <<a href="mailto:nanley.g.chery@intel.com">nanley.g.chery@intel.com</a>><br>
</span>Reviewed-by: Iago Toral Quiroga <<a href="mailto:itoral@igalia.com">itoral@igalia.com</a>><br>
<span class="">---<br>
 src/intel/vulkan/anv_blorp.c | 29 ++++++++++-------------------<br>
 1 file changed, 10 insertions(+), 19 deletions(-)<br>
<br>
diff --git a/src/intel/vulkan/anv_blorp.c b/src/intel/vulkan/anv_blorp.c<br>
</span>index c4c744868c..7f6ed0efe4 100644<br>
--- a/src/intel/vulkan/anv_blorp.c<br>
+++ b/src/intel/vulkan/anv_blorp.c<br>
@@ -1493,7 +1493,6 @@ anv_image_ccs_clear(struct anv_cmd_buffer *cmd_buffer,<br>
<span class=""><br>
 static void<br>
 ccs_resolve_attachment(struct anv_cmd_buffer *cmd_buffer,<br>
-Â Â Â Â Â Â Â Â Â Â Â Â struct blorp_batch *batch,<br>
            uint32_t att)<br>
 {<br>
  struct anv_framebuffer *fb = cmd_buffer->state.framebuffer;<br>
</span>@@ -1589,12 +1588,6 @@ ccs_resolve_attachment(struct anv_cmd_buffer *cmd_buffer,<br>
<span class="">Â Â if (resolve_op == BLORP_FAST_CLEAR_OP_NONE)<br>
    return;<br>
<br>
-Â Â struct blorp_surf surf;<br>
-Â Â get_blorp_surf_for_anv_image(<wbr>image, VK_IMAGE_ASPECT_COLOR_BIT,<br>
-Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â att_state->aux_usage, &surf);<br>
-Â Â if (att_state->fast_clear)<br>
-Â Â Â surf.clear_color = vk_to_isl_color(att_state-><wbr>clear_value.color);<br>
-<br>
  /* From the Sky Lake PRM Vol. 7, "Render Target Resolve":<br>
   *<br>
   *  "When performing a render target resolve, PIPE_CONTROL with end of<br>
</span>@@ -1610,12 +1603,8 @@ ccs_resolve_attachment(struct anv_cmd_buffer *cmd_buffer,<br>
<span class="">Â Â cmd_buffer->state.pending_<wbr>pipe_bits |=<br>
    ANV_PIPE_RENDER_TARGET_CACHE_<wbr>FLUSH_BIT | ANV_PIPE_CS_STALL_BIT;<br>
<br>
-Â Â for (uint32_t layer = 0; layer < fb->layers; layer++) {<br>
-Â Â Â blorp_ccs_resolve(batch, &surf,<br>
-Â Â Â Â Â Â Â Â Â Â Â Â iview->isl.base_level,<br>
-Â Â Â Â Â Â Â Â Â Â Â Â iview->isl.base_array_layer + layer,<br>
-Â Â Â Â Â Â Â Â Â Â Â Â iview->isl.format, resolve_op);<br>
-Â Â }<br>
+Â Â anv_ccs_resolve(cmd_buffer, att_state->color_rt_state, image,<br>
+Â Â Â Â Â Â Â Â Â Â iview->isl.base_level, fb->layers, resolve_op);<br>
<br>
  cmd_buffer->state.pending_<wbr>pipe_bits |=<br>
    ANV_PIPE_RENDER_TARGET_CACHE_<wbr>FLUSH_BIT | ANV_PIPE_CS_STALL_BIT;<br>
</span>@@ -1633,16 +1622,13 @@ anv_cmd_buffer_resolve_<wbr>subpass(struct anv_cmd_buffer *cmd_buffer)<br>
<span class="">Â Â struct anv_subpass *subpass = cmd_buffer->state.subpass;<br>
<br>
<br>
-Â Â struct blorp_batch batch;<br>
-Â Â blorp_batch_init(&cmd_buffer-><wbr>device->blorp, &batch, cmd_buffer, 0);<br>
-<br>
  for (uint32_t i = 0; i < subpass->color_count; ++i) {<br>
    const uint32_t att = subpass->color_attachments[i].<wbr>attachment;<br>
    if (att == VK_ATTACHMENT_UNUSED)<br>
     continue;<br>
<br>
    assert(att < cmd_buffer->state.pass-><wbr>attachment_count);<br>
-Â Â Â ccs_resolve_attachment(cmd_<wbr>buffer, &batch, att);<br>
+Â Â Â ccs_resolve_attachment(cmd_<wbr>buffer, att);<br>
  }<br>
<br>
  if (subpass->has_resolve) {<br>
</span>@@ -1681,6 +1667,10 @@ anv_cmd_buffer_resolve_<wbr>subpass(struct anv_cmd_buffer *cmd_buffer)<br>
<span class="">Â Â Â Â Â const VkRect2D render_area = cmd_buffer->state.render_area;<br>
<br>
     assert(src_iview->aspect_mask == dst_iview->aspect_mask);<br>
+<br>
+Â Â Â Â Â struct blorp_batch batch;<br>
+Â Â Â Â Â blorp_batch_init(&cmd_buffer-><wbr>device->blorp, &batch, cmd_buffer, 0);<br>
+<br>
     resolve_image(&batch, src_iview->image,<br>
            src_iview->isl.base_level,<br>
            src_iview->isl.base_array_<wbr>layer,<br>
</span>@@ -1692,11 +1682,12 @@ anv_cmd_buffer_resolve_<wbr>subpass(struct anv_cmd_buffer *cmd_buffer)<br>
<div class="HOEnZb"><div class="h5">Â Â Â Â Â Â Â Â Â Â Â Â render_area.offset.x, render_area.offset.y,<br>
            render_area.extent.width, render_area.extent.height);<br>
<br>
-Â Â Â Â Â ccs_resolve_attachment(cmd_<wbr>buffer, &batch, dst_att);<br>
+     blorp_batch_finish(&batch);<br></div></div></blockquote><div><br></div><div>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.<br></div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div class="HOEnZb"><div class="h5">
+<br>
+Â Â Â Â Â ccs_resolve_attachment(cmd_<wbr>buffer, dst_att);<br>
    }<br>
  }<br>
<br>
-Â Â blorp_batch_finish(&batch);<br>
 }<br>
<br>
 void<br>
--<br>
2.13.1<br>
<br>
______________________________<wbr>_________________<br>
mesa-dev mailing list<br>
<a href="mailto:mesa-dev@lists.freedesktop.org">mesa-dev@lists.freedesktop.org</a><br>
<a href="https://lists.freedesktop.org/mailman/listinfo/mesa-dev" rel="noreferrer" target="_blank">https://lists.freedesktop.org/<wbr>mailman/listinfo/mesa-dev</a><br>
</div></div></blockquote></div><br></div></div>