<?xml version="1.0" encoding="UTF-8"?>
<rss xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#" xmlns:taxo="http://purl.org/rss/1.0/modules/taxonomy/" version="2.0">
  <channel>
    <title>topic Hello Yan, in Intel® oneAPI Math Kernel Library</title>
    <link>https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/MKL-LLC-usage-policy/m-p/1134745#M25936</link>
    <description>&lt;P&gt;Hello Yan,&lt;/P&gt;

&lt;P&gt;I may recommend you to use the Intel Vtune&amp;nbsp; Amplifier XE, it&amp;nbsp;can explore the LLC&amp;nbsp;missing&amp;nbsp;, so you can compare the saturation become worse or not .&lt;/P&gt;

&lt;P&gt;Best Regards,&lt;/P&gt;

&lt;P&gt;Ying &amp;nbsp;&amp;nbsp;&lt;/P&gt;</description>
    <pubDate>Mon, 12 Feb 2018 04:11:10 GMT</pubDate>
    <dc:creator>Ying_H_Intel</dc:creator>
    <dc:date>2018-02-12T04:11:10Z</dc:date>
    <item>
      <title>MKL LLC usage policy</title>
      <link>https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/MKL-LLC-usage-policy/m-p/1134742#M25933</link>
      <description>&lt;P&gt;&lt;SPAN style="font-size: 1em;"&gt;Hello, experts,&lt;/SPAN&gt;&lt;/P&gt;

&lt;P&gt;&lt;SPAN style="font-size: 1em;"&gt;Assume that I have 4 cores machine, each core has 2MB of LLC slice and LLC includes L2.&amp;nbsp;&lt;/SPAN&gt;&lt;/P&gt;

&lt;P&gt;&lt;SPAN style="font-size: 1em;"&gt;1) If I use single-threaded MKL, the MKL instance will use 2MB of LLC or use 8MB LLC?&lt;/SPAN&gt;&lt;/P&gt;

&lt;P&gt;&lt;SPAN style="font-size: 1em;"&gt;2) If I use openmp threads to control the&amp;nbsp;&lt;/SPAN&gt;parallelism, will MKL instance determine available LLC based on thread num?&lt;/P&gt;

&lt;P&gt;Any help is appreciated. Thanks.&lt;/P&gt;

&lt;P&gt;Best Regards&lt;/P&gt;

&lt;P&gt;&amp;nbsp;&lt;/P&gt;

&lt;P&gt;&amp;nbsp;&lt;/P&gt;

&lt;P&gt;&amp;nbsp;&lt;/P&gt;</description>
      <pubDate>Wed, 07 Feb 2018 03:25:34 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/MKL-LLC-usage-policy/m-p/1134742#M25933</guid>
      <dc:creator>Dai__Yan</dc:creator>
      <dc:date>2018-02-07T03:25:34Z</dc:date>
    </item>
    <item>
      <title>Hi Yan,</title>
      <link>https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/MKL-LLC-usage-policy/m-p/1134743#M25934</link>
      <description>Hi Yan, 

Could you please tell what is the MKL function you are calling and what is problem size?  why do you want to check the LLC from application level? 

MKL don't directly control to use which LLC.  As i understand, the feature  is supposed determined by CPU feature  or cpu schedule/Bios to change. MKL is just like any of an application, doesn't control the Cache behavior.  In more case, the LLC is designed to shared , any process or multi-threads will share the 8M memory. 

And for MKL, if you want to to make sure the high performance and keep the better memory locality,  you may control OpenMP affinity, please refer to OpenMP thread affinity; 
If your machine is NUMA, then maybe the numa control may help.  

Please see the article


Intel® MKL NUMA Notes | Intel® Software


&lt;A href="https://en.wikipedia.org/wiki/CPU_cache#LLC" target="_blank"&gt;https://en.wikipedia.org/wiki/CPU_cache#LLC&lt;/A&gt;

Best Regards,
Ying</description>
      <pubDate>Fri, 09 Feb 2018 05:30:59 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/MKL-LLC-usage-policy/m-p/1134743#M25934</guid>
      <dc:creator>Ying_H_Intel</dc:creator>
      <dc:date>2018-02-09T05:30:59Z</dc:date>
    </item>
    <item>
      <title>Hello,  Ying</title>
      <link>https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/MKL-LLC-usage-policy/m-p/1134744#M25935</link>
      <description>&lt;P&gt;&lt;SPAN style="font-size: 1em;"&gt;Hello, &amp;nbsp;Ying&lt;/SPAN&gt;&lt;/P&gt;

&lt;P&gt;&lt;SPAN style="font-size: 1em;"&gt;​Thanks for your&amp;nbsp;&lt;/SPAN&gt;kind help.&lt;SPAN style="font-size: 1em;"&gt;&amp;nbsp;&lt;/SPAN&gt;&lt;/P&gt;

&lt;P&gt;Assume that in i7-4770K, I have &lt;STRONG&gt;4 threads application&lt;/STRONG&gt; and each thread will call &lt;STRONG&gt;single-threaded&lt;/STRONG&gt; sgemm routine.&lt;/P&gt;

&lt;P&gt;&lt;SPAN style="font-size: 1em;"&gt;And my question is that assuming the LLC is &lt;STRONG&gt;inclusive&lt;/STRONG&gt;(before Skylake Server) and each sgemm will generate its own memory&amp;nbsp;&lt;/SPAN&gt;traffic and may &lt;STRONG&gt;overwrite&lt;/STRONG&gt; data from other threads in LLC. And if single-threaded sgemm will use whole LLC, such situation will become much worse. So May I know whether this situation may happen?&lt;/P&gt;

&lt;P&gt;Best Regards&lt;/P&gt;

&lt;P&gt;&amp;nbsp;&lt;/P&gt;

&lt;P&gt;&amp;nbsp;&lt;/P&gt;

&lt;P&gt;&amp;nbsp;&lt;/P&gt;</description>
      <pubDate>Sun, 11 Feb 2018 05:17:00 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/MKL-LLC-usage-policy/m-p/1134744#M25935</guid>
      <dc:creator>Dai__Yan</dc:creator>
      <dc:date>2018-02-11T05:17:00Z</dc:date>
    </item>
    <item>
      <title>Hello Yan,</title>
      <link>https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/MKL-LLC-usage-policy/m-p/1134745#M25936</link>
      <description>&lt;P&gt;Hello Yan,&lt;/P&gt;

&lt;P&gt;I may recommend you to use the Intel Vtune&amp;nbsp; Amplifier XE, it&amp;nbsp;can explore the LLC&amp;nbsp;missing&amp;nbsp;, so you can compare the saturation become worse or not .&lt;/P&gt;

&lt;P&gt;Best Regards,&lt;/P&gt;

&lt;P&gt;Ying &amp;nbsp;&amp;nbsp;&lt;/P&gt;</description>
      <pubDate>Mon, 12 Feb 2018 04:11:10 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/MKL-LLC-usage-policy/m-p/1134745#M25936</guid>
      <dc:creator>Ying_H_Intel</dc:creator>
      <dc:date>2018-02-12T04:11:10Z</dc:date>
    </item>
  </channel>
</rss>

