<div dir="ltr"><div class="gmail_extra"><div class="gmail_quote">On Wed, Aug 31, 2016 at 8:29 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">This series enables Hierarchical depth buffer rendering and fast depth clears<br>
for render passes with a single subpass running on platforms BDW+. Platforms<br>
pre-BDW can test this feature with an environment variable. The FPS of some<br>
demos are roughly estimated to increase by as much as ~50% on a SKL GT2.<br>
<br>
This feature was partially implemented by Chad and Jason. Where applicable,<br>
I've tried to accurately note the modifications I've made to their patches<br>
without being too verbose. I've also tried to maintain the authorship of their<br>
patches when the core of their work remained.<br>
<br>
The only patch which wasn't retained due to the core of the work being lost<br>
was a patch to create a HiZ surface. This was replaced with my patch to update<br>
an existing function which does so. This diverged enough for me to feel at<br>
risk of misrepresenting the original author's work.<br>
<br>
Any suggestions with respect to my annotating method, notices of incorrectly<br>
attributed credit, or general comments are welcome.<br></blockquote><div><br></div><div>Feel free to take more credit. :) Chad and I wrote sketchy, untested, skeleton patches. You were the one who got it working!<br><br></div><div>Patches 1, 3-7, and 10-12 have a few comments here and there. Assuming those comments are addressed, those patches are<br><br></div><div>Reviewed-by: Jason Ekstrand <<a href="mailto:jason@jlekstrand.net">jason@jlekstrand.net</a>><br><br></div><div>We talked about 2 offline and I sent my little 6-patch series that makes the original plan work.<br><br></div><div>On patch 9, I gave a bunch of comments but one thing was clear: We need tests. In the interest of merging patches, I think I'd recommend that we disable HiZ for mipmapped surfaces (we can just not allocate the surface) and don't do fast-clears for anything other than full-RT clears. That seems like the shortest path to getting the patches merged quickly with some guarantee of correctness.<br><br>For partial clears and mipmapped HiZ, I think we need more tests. There may be CTS tests for partial depth clears (In particular, the subpass tests) but I'm not sure. I'll leave it up to you as to whether you'd rather write CTS tests or crucible tests. Crucible may be easier, but the CTS needs those tests too, so maybe we should be good citizens and put them there?<br></div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<br>
Chad Versace (4):<br>
anv: Add anv_image::hiz_surface<br>
anv: Add func anv_image_has_hiz()<br>
anv: Allocate hiz surface<br>
genX/cmd_buffer: Enable rendering to HiZ<br>
<br>
Jason Ekstrand (3):<br>
anv: Move BindImageMemory to anv_image.c<br>
anv/image: Memset hiz surfaces to 0 when binding memory<br>
anv/cmd_buffer: Add code for performing HZ operations<br>
<br>
Nanley Chery (5):<br>
isl: Correct a comment in the isl_format enum<br>
isl: Update isl_surf_get_hiz_surf()<br>
isl: Make MSAA pixel scaling function public<br>
genX/cmd_buffer: Enable fast depth clears<br>
anv/TODO: Update the HiZ task<br>
<br>
src/intel/isl/isl.c | 41 ++++++++++--<br>
src/intel/isl/isl.h | 6 +-<br>
src/intel/vulkan/TODO | 2 +-<br>
src/intel/vulkan/anv_device.c<wbr> | 20 ------<br>
src/intel/vulkan/anv_genX.h | 3 +<br>
src/intel/vulkan/anv_image.c | 67 ++++++++++++++++++-<br>
src/intel/vulkan/anv_pass.c | 11 +++<br>
src/intel/vulkan/anv_private.<wbr>h | 18 +++++<br>
src/intel/vulkan/gen7_cmd_buf<wbr>fer.c | 5 ++<br>
src/intel/vulkan/gen8_cmd_buf<wbr>fer.c | 134 ++++++++++++++++++++++++++++++<wbr>+++++++<br>
src/intel/vulkan/genX_cmd_buf<wbr>fer.c | 45 +++++++++++--<br>
11 files changed, 313 insertions(+), 39 deletions(-)<br>
<span><font color="#888888"><br>
--<br>
2.9.3<br>
<br>
</font></span></blockquote></div><br></div></div>