<?xml version="1.0" encoding="UTF-8" ?>
<?xml-stylesheet type="text/xsl" href="https://community.arm.com/utility/feedstylesheets/rss.xsl" media="screen"?><rss version="2.0" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:slash="http://purl.org/rss/1.0/modules/slash/" xmlns:wfw="http://wellformedweb.org/CommentAPI/"><channel><title>what prevent parallel of non-fragment and fragment</title><link>https://community.arm.com/developer/tools-software/graphics/f/discussions/47977/what-prevent-parallel-of-non-fragment-and-fragment</link><description> 
 We see, the non-fragment and fragment workloads are not parellel. 
 &amp;quot; This expression defines the fragment queue utilization compared against the GPU active cycles. For GPU bound content it is expected that the GPU queues will process work in parallel</description><dc:language>en-US</dc:language><generator>Telligent Community 10</generator><item><title>RE: what prevent parallel of non-fragment and fragment</title><link>https://community.arm.com/thread/168460?ContentTypeID=1</link><pubDate>Thu, 05 Nov 2020 08:05:35 GMT</pubDate><guid isPermaLink="false">dd9e70c8-6d3c-4c71-b136-2456382a7b5c:08cc6225-ac90-40f5-83eb-a2db38511e9e</guid><dc:creator>Peter Harris</dc:creator><description>&lt;p&gt;Hi Shawn,&amp;nbsp;&lt;br /&gt;&lt;br /&gt;First of all, if you are hitting&amp;nbsp;your target framerate, e.g. 30 or 60&amp;nbsp;FPS, you may not be able to get perfect overlap. In that situation it really depends on the frame GPU load and the available GPU frequencies on that target device.&amp;nbsp;&lt;br /&gt;&lt;br /&gt;If you are not hitting 60 FPS, you&amp;#39;re generally looking to avoid dependencies between render passes. &lt;br /&gt;&lt;br /&gt;In OpenGL ES the main causes of this are:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Fences or queries which are waited on too early, so CPU must block and wait for the GPU to resolve the fence/query.&lt;/li&gt;
&lt;li&gt;Using glMapBuffer without GL_MAP_UNSYNCHRONIZED while the buffer is still referenced, so the CPU must block and wait for the GPU dependency to resovled (it might, it might also just ghost the resource).&lt;/li&gt;
&lt;li&gt;&lt;span&gt;Having a&amp;nbsp;vertex shader consume the output of the previous render pass as an input&amp;nbsp;texture or image.&amp;nbsp;&lt;/span&gt;&lt;/li&gt;
&lt;li&gt;Having a compute shader consume the output of the previous render pass as an input&amp;nbsp;texture or image.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;In Vulkan:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Same problems with fences and queries.&lt;/li&gt;
&lt;li&gt;Having conservative resource dependencies, for example if the srcStage for generating a resource is &amp;quot;BOTTOM_OF_PIPE&amp;quot; and dstStage for consuming it is &amp;quot;TOP_OF_PIPE&amp;quot; then the workloads will serialize, a &amp;quot;TOP&amp;quot; (vertex shading) must wait for &amp;quot;BOTTOM&amp;quot; (fragment shading) to complete before it can start.&amp;nbsp;&amp;nbsp;&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;span&gt;This might help provide some more background: &lt;a href="https://developer.arm.com/solutions/graphics-and-gaming/developer-guides/learn-the-basics/workload-pipelining"&gt;developer.arm.com/.../workload-pipelining&lt;/a&gt;&lt;br /&gt;&lt;/span&gt;&lt;br /&gt;Kind regards,&amp;nbsp;&lt;br /&gt;Pete&lt;/p&gt;&lt;div style="clear:both;"&gt;&lt;/div&gt;</description></item></channel></rss>