<html>
    <head>
      <base href="https://bugs.freedesktop.org/">
    </head>
    <body><table border="1" cellspacing="0" cellpadding="8">
        <tr>
          <th>Bug ID</th>
          <td><a class="bz_bug_link 
          bz_status_NEW "
   title="NEW - [PERF] Updating ubo offset via vkCmdBindDescriptorSets is causing flush that is taking 50% of rendering time"
   href="https://bugs.freedesktop.org/show_bug.cgi?id=105695">105695</a>
          </td>
        </tr>

        <tr>
          <th>Summary</th>
          <td>[PERF] Updating ubo offset via vkCmdBindDescriptorSets is causing flush that is taking 50% of rendering time
          </td>
        </tr>

        <tr>
          <th>Product</th>
          <td>Mesa
          </td>
        </tr>

        <tr>
          <th>Version</th>
          <td>git
          </td>
        </tr>

        <tr>
          <th>Hardware</th>
          <td>x86-64 (AMD64)
          </td>
        </tr>

        <tr>
          <th>OS</th>
          <td>Linux (All)
          </td>
        </tr>

        <tr>
          <th>Status</th>
          <td>NEW
          </td>
        </tr>

        <tr>
          <th>Severity</th>
          <td>normal
          </td>
        </tr>

        <tr>
          <th>Priority</th>
          <td>medium
          </td>
        </tr>

        <tr>
          <th>Component</th>
          <td>Drivers/Vulkan/intel
          </td>
        </tr>

        <tr>
          <th>Assignee</th>
          <td>intel-3d-bugs@lists.freedesktop.org
          </td>
        </tr>

        <tr>
          <th>Reporter</th>
          <td>ytriffy@gmail.com
          </td>
        </tr>

        <tr>
          <th>QA Contact</th>
          <td>intel-3d-bugs@lists.freedesktop.org
          </td>
        </tr>

        <tr>
          <th>CC</th>
          <td>jason@jlekstrand.net
          </td>
        </tr></table>
      <p>
        <div>
        <pre>I do vkCmdBindDescriptorSets per draw call to change transformation matrix of
an object. Basically I change only one number - offset into dynamic ubo. Then I
call vkCmdDrawIndexed and it calls gen9_cmd_buffer_flush_state that amounts to
50% of rendering time (in terms of instr fetch metric). I dig deeper and find
two culprits: flush_descriptor_sets, cmd_buffer_flush_push_constants. And I
don't even use push constants, I prefer dynamic ubos. Emitting binding tables
is huge amount of work (23% of total rendering time). I don't understand why so
much work has to be done just to change offset in memory. And this is the most
popular usecase - everyone wants to change matrix per object. My opengl
implementation is 2 times faster than this. Are there any plans on improving
performance in that area?

I got this data from valgrind profiling:
<a href="https://www.dropbox.com/s/wsri01x69kwciyo/callgrind.cullingvk_ubo.out?dl=0">https://www.dropbox.com/s/wsri01x69kwciyo/callgrind.cullingvk_ubo.out?dl=0</a></pre>
        </div>
      </p>


      <hr>
      <span>You are receiving this mail because:</span>

      <ul>
          <li>You are the QA Contact for the bug.</li>
          <li>You are the assignee for the bug.</li>
      </ul>
    </body>
</html>