Attachment 127548 Details for Bug 97872 – attachment-31724-0.html

attachment-31724-0.html

attachment-31724-0.html (text/html), 3.62 KB, created by Sean V Kelley on 2016-10-26 06:44:04 UTC

(hide)

Description:

Filename:

MIME Type:

Creator: Sean V Kelley

Created: 2016-10-26 06:44:04 UTC

Size: 3.62 KB

patch

obsolete

><html><head><meta http-equiv="Content-Type" content="text/html charset=utf-8"><base></head><body style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space;" class=""><meta http-equiv="Content-Type" content="text/html charset=utf-8" class=""><base class=""><div style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space;" class=""><br class=""><div class=""><blockquote type="cite" class=""><div class="">On 25 DFÃ³mh 2016, at 22:49, <a href="mailto:bugzilla-daemon@freedesktop.org" class="">bugzilla-daemon@freedesktop.org</a> wrote:</div><br class="Apple-interchange-newline"><div class="">
>    
>      <base href="https://bugs.freedesktop.org/" class="">
>    
>    <div class=""><div class="">
>        <br class="webkit-block-placeholder"></div><div class="">
>            <b class=""><a class=" bz_status_NEW bz_bug_link 
>" title="NEW - GPU hang with libva (gstreamer)" href="https://bugs.freedesktop.org/show_bug.cgi?id=97872#c16">Comment # 16</a>
>              on <a class=" bz_status_NEW bz_bug_link 
>" title="NEW - GPU hang with libva (gstreamer)" href="https://bugs.freedesktop.org/show_bug.cgi?id=97872">bug 97872</a>
>              from <span class="vcard"><a class="email" href="mailto:haihao.xiang@intel.com" title="haihao &lt;haihao.xiang@intel.com&gt;"> <span class="fn">haihao</span></a>
></span></b>
>        <pre class="">(In reply to Sean V Kelley from <a href="x-msg://4/show_bug.cgi?id=97872#c13" class="">comment #13</a>)
><span class="quote">&gt; Haihao,
>&gt; 
>&gt; The 2nd VDBOX on SKL is not a complete VDBOX, it only contains MFX. I would
>&gt; recommend shunting all MFX workloads to the 2nd VDBOX and using the 1st
>&gt; VDBOX for HCP, VDENC, HuC.</span>
>
>Usually user doesn't use different codecs at the same time. I don't think using
>BSD1 only for MFX is better choice for VP8/H264/MPEG2 etc.
></pre></div></div></div></blockquote><div class=""><br class=""></div>You canât assume that. &nbsp;In fact, VDENC will likely dominate.&nbsp;<br class=""><blockquote type="cite" class=""><div class=""><div class=""><div class=""><pre class=""><span class="quote">&gt; What we need for a permanent fix is an i915
>&gt; kernel patch that manages the loads between the engines based on input from
>&gt; UMD.</span>
>
>Currently i915 kernel can manage the loads between the engines.  But i915
>kernel doesn't know HCP/HuC commands must be ran from the 2nd ring unless UMD
>driver can tell the kernel. which is why I915_EXEC_BSD_RING1 and
>I915_EXEC_BSD_RING2 are added to the execution ioctl.
></pre></div></div></div></blockquote><div class=""><br class=""></div>Yes, that is the whole point of the patch submitted to the i915, but it is still a hack. &nbsp;Iâm well aware of how this works. &nbsp;And that means every time we use a new codec that is not balanced between the VDBoxen we add the hack flag. &nbsp;Again, we need to do better.</div><div class=""><br class=""></div><div class="">Sean<br class=""><blockquote type="cite" class=""><div class=""><div class=""><div class=""><pre class=""><span class="quote">&gt; 
>&gt; So, while I find you patch servicable for the immediate need, I want to see
>&gt; a long term fix along the lines above that I suggest.  I will look into the
>&gt; kernel patch.
>&gt; 
>&gt; Thanks,
>&gt; 
>&gt; Sean</span></pre>
>        </div><div class=""><br class="webkit-block-placeholder"></div>
>
>
>      <hr class="">
>      <span class="">You are receiving this mail because:</span>
>
>      <ul class="">
>          <li class="">You are the QA Contact for the bug.</li>
>      </ul>
>    </div>
></div></blockquote></div><br class=""></div></body></html>=

<html><head><meta http-equiv="Content-Type" content="text/html charset=utf-8"><base></head><body style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space;" class=""><meta http-equiv="Content-Type" content="text/html charset=utf-8" class=""><base class=""><div style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space;" class=""><br class=""><div class=""><blockquote type="cite" class=""><div class="">On 25 DFÃ³mh 2016, at 22:49, <a href="mailto:bugzilla-daemon@freedesktop.org" class="">bugzilla-daemon@freedesktop.org</a> wrote:</div><br class="Apple-interchange-newline"><div class="">
    
      <base href="https://bugs.freedesktop.org/" class="">
    
    <div class=""><div class="">
        <br class="webkit-block-placeholder"></div><div class="">
            <b class=""><a class=" bz_status_NEW bz_bug_link 
" title="NEW - GPU hang with libva (gstreamer)" href="https://bugs.freedesktop.org/show_bug.cgi?id=97872#c16">Comment # 16</a>
              on <a class=" bz_status_NEW bz_bug_link 
" title="NEW - GPU hang with libva (gstreamer)" href="https://bugs.freedesktop.org/show_bug.cgi?id=97872">bug 97872</a>
              from <span class="vcard"><a class="email" href="mailto:haihao.xiang@intel.com" title="haihao &lt;haihao.xiang@intel.com&gt;"> <span class="fn">haihao</span></a>
</span></b>
        <pre class="">(In reply to Sean V Kelley from <a href="x-msg://4/show_bug.cgi?id=97872#c13" class="">comment #13</a>)
<span class="quote">&gt; Haihao,
&gt; 
&gt; The 2nd VDBOX on SKL is not a complete VDBOX, it only contains MFX. I would
&gt; recommend shunting all MFX workloads to the 2nd VDBOX and using the 1st
&gt; VDBOX for HCP, VDENC, HuC.</span>

Usually user doesn't use different codecs at the same time. I don't think using
BSD1 only for MFX is better choice for VP8/H264/MPEG2 etc.
</pre></div></div></div></blockquote><div class=""><br class=""></div>You canât assume that. &nbsp;In fact, VDENC will likely dominate.&nbsp;<br class=""><blockquote type="cite" class=""><div class=""><div class=""><div class=""><pre class=""><span class="quote">&gt; What we need for a permanent fix is an i915
&gt; kernel patch that manages the loads between the engines based on input from
&gt; UMD.</span>

Currently i915 kernel can manage the loads between the engines.  But i915
kernel doesn't know HCP/HuC commands must be ran from the 2nd ring unless UMD
driver can tell the kernel. which is why I915_EXEC_BSD_RING1 and
I915_EXEC_BSD_RING2 are added to the execution ioctl.
</pre></div></div></div></blockquote><div class=""><br class=""></div>Yes, that is the whole point of the patch submitted to the i915, but it is still a hack. &nbsp;Iâm well aware of how this works. &nbsp;And that means every time we use a new codec that is not balanced between the VDBoxen we add the hack flag. &nbsp;Again, we need to do better.</div><div class=""><br class=""></div><div class="">Sean<br class=""><blockquote type="cite" class=""><div class=""><div class=""><div class=""><pre class=""><span class="quote">&gt; 
&gt; So, while I find you patch servicable for the immediate need, I want to see
&gt; a long term fix along the lines above that I suggest.  I will look into the
&gt; kernel patch.
&gt; 
&gt; Thanks,
&gt; 
&gt; Sean</span></pre>
        </div><div class=""><br class="webkit-block-placeholder"></div>


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

      <ul class="">
          <li class="">You are the QA Contact for the bug.</li>
      </ul>
    </div>
</div></blockquote></div><br class=""></div></body></html>=

Actions: View

Attachments on bug 97872: 126659 | 127525 | 127534 | 127548

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.