<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
</head>
<body text="#000000" bgcolor="#FFFFFF">
<div class="moz-cite-prefix">Ok, found something that works. Xonotic
in lowest resolution, lowest effects quality (e.g. totally CPU
bound):<br>
<br>
Without per process BOs:<br>
<br>
Xonotic 0.8:<br>
pts/xonotic-1.4.0 [Resolution: 800 x 600 - Effects Quality:
Low]<br>
Test 1 of 1<br>
Estimated Trial Run Count: 3<br>
Estimated Time To Completion: 3 Minutes<br>
Started Run 1 @ 21:13:50<br>
Started Run 2 @ 21:14:57<br>
Started Run 3 @ 21:16:03 [Std. Dev: 0.94%]<br>
<br>
Test Results:<br>
187.436577<br>
189.514724<br>
190.9605812<br>
<br>
Average: 189.30 Frames Per Second<br>
Minimum: 131<br>
Maximum: 355<br>
<br>
With per process BOs:<br>
<br>
Xonotic 0.8:<br>
pts/xonotic-1.4.0 [Resolution: 800 x 600 - Effects Quality:
Low]<br>
Test 1 of 1<br>
Estimated Trial Run Count: 3<br>
Estimated Time To Completion: 3 Minutes<br>
Started Run 1 @ 21:20:05<br>
Started Run 2 @ 21:21:07<br>
Started Run 3 @ 21:22:10 [Std. Dev: 1.49%]<br>
<br>
Test Results:<br>
203.0471676<br>
199.6622532<br>
197.0954183<br>
<br>
Average: 199.93 Frames Per Second<br>
Minimum: 132<br>
Maximum: 349<br>
<br>
Well that looks like some improvement.<br>
<br>
Regards,<br>
Christian.<br>
<br>
Am 28.08.2017 um 14:59 schrieb Zhou, David(ChunMing):<br>
</div>
<blockquote type="cite"
cite="mid:MWHPR1201MB0206D4E64E86D7AADA159A47B49E0@MWHPR1201MB0206.namprd12.prod.outlook.com">
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
<meta name="Generator" content="Microsoft Exchange Server">
<!-- converted from text -->
<style><!-- .EmailQuote { margin-left: 1pt; padding-left: 4pt; border-left: #800000 2px solid; } --></style>
<div>I will push our vulkan guys to test it, their bo list is very long.<br>
<br>
<p dir="ltr" style="display:inline"><span style="color:#888888"><span
style="font-size:0.81em">发自坚果 Pro</span></span></p>
<style type="text/css">
<!--
* body
{padding:0 16px 30px!important;
margin:0!important;
background-color:#ffffff;
line-height:1.4;
word-wrap:break-word;
word-break:normal}
div
{word-wrap:break-word;
word-break:normal}
p
{word-wrap:break-word;
word-break:normal;
text-indent:0pt!important}
span
{word-wrap:break-word;
word-break:normal}
a
{word-wrap:break-word;
word-break:normal}
td
{word-wrap:break-word;
word-break:break-all}
-->
</style>
<div class="x_quote">
<div style="margin:0 0px; font-size:105%"><font
style="line-height:1.4" color="#629140"><span>Christian
K鰊ig <a class="moz-txt-link-rfc2396E" href="mailto:deathsimple@vodafone.de"><deathsimple@vodafone.de></a> 于 2017年8月28日
下午7:55写道：</span></font></div>
<br type="attribution">
</div>
</div>
<font size="2"><span style="font-size:10pt;">
<div class="PlainText">Am 28.08.2017 um 06:21 schrieb zhoucm1:<br>
><br>
><br>
> On 2017年08月27日 18:03, Christian König wrote:<br>
>> Am 25.08.2017 um 21:19 schrieb Christian König:<br>
>>> Am 25.08.2017 um 18:22 schrieb Marek Olšák:<br>
>>>> On Fri, Aug 25, 2017 at 3:00 PM, Christian
König<br>
>>>> <a class="moz-txt-link-rfc2396E" href="mailto:deathsimple@vodafone.de"><deathsimple@vodafone.de></a> wrote:<br>
>>>>> Am 25.08.2017 um 12:32 schrieb zhoucm1:<br>
>>>>>><br>
>>>>>><br>
>>>>>> On 2017年08月25日 17:38, Christian
König wrote:<br>
>>>>>>> From: Christian König
<a class="moz-txt-link-rfc2396E" href="mailto:christian.koenig@amd.com"><christian.koenig@amd.com></a><br>
>>>>>>><br>
>>>>>>> Add the IOCTL interface so that
applications can allocate per VM <br>
>>>>>>> BOs.<br>
>>>>>>><br>
>>>>>>> Still WIP since not all corner
cases are tested yet, but this <br>
>>>>>>> reduces<br>
>>>>>>> average<br>
>>>>>>> CS overhead for 10K BOs from
21ms down to 48us.<br>
>>>>>> Wow, cheers, eventually you get per
vm bo to same reservation <br>
>>>>>> with PD/pts,<br>
>>>>>> indeed save a lot of bo list.<br>
>>>>><br>
>>>>> Don't cheer to loud yet, that is a
completely constructed test case.<br>
>>>>><br>
>>>>> So far I wasn't able to archive any
improvements with any real <br>
>>>>> game on this<br>
>>>>> with Mesa.<br>
> With thinking more, too many BOs share one reservation,
which could <br>
> result in reservation lock often is busy, if eviction
or destroy also <br>
> happens often in the meaning time, then which could
effect VM update <br>
> and CS submission as well.<br>
<br>
That's exactly the reason why I've added code to the BO
destroy path to <br>
avoid at least some of the problems. But yeah, that's only
the tip of <br>
the iceberg of problems with that approach.<br>
<br>
> Anyway, this is very good start and try that we reduce
CS overhead, <br>
> especially we've seen "reduces average CS overhead for
10K BOs from <br>
> 21ms down to 48us. ".<br>
<br>
Actually, it's not that good. See this is a completely build
up test <br>
case on a kernel with lockdep and KASAN enabled.<br>
<br>
In reality we usually don't have so many BOs and so far I
wasn't able to <br>
find much of an improvement in any real world testing.<br>
<br>
Regards,<br>
Christian.<br>
</div>
</span></font>
<br>
<fieldset class="mimeAttachmentHeader"></fieldset>
<br>
<pre wrap="">_______________________________________________
amd-gfx mailing list
<a class="moz-txt-link-abbreviated" href="mailto:amd-gfx@lists.freedesktop.org">amd-gfx@lists.freedesktop.org</a>
<a class="moz-txt-link-freetext" href="https://lists.freedesktop.org/mailman/listinfo/amd-gfx">https://lists.freedesktop.org/mailman/listinfo/amd-gfx</a>
</pre>
</blockquote>
<p><br>
</p>
</body>
</html>