<div dir="auto"><div><div class="gmail_extra"><div class="gmail_quote">On Feb 23, 2017 8:03 AM, "Lionel Landwerlin" <<a href="mailto:lionel.g.landwerlin@intel.com">lionel.g.landwerlin@intel.com</a>> wrote:<br type="attribution"><blockquote class="quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">I must be missing something, but I'm struggling to see why this patch change anything.<br>
Why does it matter if we write a clear color in SURFACE_STATE?<br>
Shouldn't the usage of that color value only depend on whether we enable 3DSTATE_PS.RenderTargetFastCle<wbr>arEnable for instance?<br></blockquote></div></div></div><div dir="auto"><br></div><div dir="auto">Because ISL will assert on us if it's a color the surface state doesn't allow. In this particular case it was a non-0/1 color which isn't even representable on Broadwell.</div><div dir="auto"><br></div><div dir="auto"><div class="gmail_extra"><div class="gmail_quote"><blockquote class="quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
-<br>
Lionel<div class="elided-text"><br>
<br>
On 22/02/17 03:31, Jason Ekstrand wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
Not all clear colors are valid. In particular, on Broadwell and<br>
earlier, only 0/1 colors are allowed in surface state. No CTS tests are<br>
affected outright by this because, apparently, the CTS coverage for<br>
different clear colors is pretty terrible. However, when multisample<br>
compression is enabled, we do hit it with CTS tests and this commit<br>
prevents regressions when enabling MCS on Broadwell and earlier.<br>
<br>
Cc: "13.0 17.0" <<a href="mailto:mesa-stable@lists.freedesktop.org" target="_blank">mesa-stable@lists.freedesktop<wbr>.org</a>><br>
---<br>
src/intel/vulkan/anv_blorp.c | 5 +++--<br>
1 file changed, 3 insertions(+), 2 deletions(-)<br>
<br>
diff --git a/src/intel/vulkan/anv_blorp.c b/src/intel/vulkan/anv_blorp.c<br>
index 4e7078b..8db03e4 100644<br>
--- a/src/intel/vulkan/anv_blorp.c<br>
+++ b/src/intel/vulkan/anv_blorp.c<br>
@@ -1198,9 +1198,10 @@ anv_cmd_buffer_clear_subpass(s<wbr>truct anv_cmd_buffer *cmd_buffer)<br>
struct blorp_surf surf;<br>
get_blorp_surf_for_anv_image(i<wbr>mage, VK_IMAGE_ASPECT_COLOR_BIT,<br>
att_state->aux_usage, &surf);<br>
- surf.clear_color = vk_to_isl_color(att_state->cle<wbr>ar_value.color);<br>
if (att_state->fast_clear) {<br>
+ surf.clear_color = vk_to_isl_color(att_state->cle<wbr>ar_value.color);<br>
+<br>
blorp_fast_clear(&batch, &surf, iview->isl.format,<br>
iview->isl.base_level,<br>
iview->isl.base_array_layer, fb->layers,<br>
@@ -1224,7 +1225,7 @@ anv_cmd_buffer_clear_subpass(s<wbr>truct anv_cmd_buffer *cmd_buffer)<br>
render_area.offset.x, render_area.offset.y,<br>
render_area.offset.x + render_area.extent.width,<br>
render_area.offset.y + render_area.extent.height,<br>
- surf.clear_color, NULL);<br>
+ vk_to_isl_color(att_state->cl<wbr>ear_value.color), NULL);<br>
}<br>
att_state->pending_clear_aspec<wbr>ts = 0;<br>
</blockquote>
<br>
<br>
</div></blockquote></div><br></div></div></div>