<?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 One big reason to implement support for MacOS X! in OpenCL* for CPU</title>
    <link>https://community.intel.com/t5/OpenCL-for-CPU/One-big-reason-to-implement-support-for-MacOS-X/m-p/774139#M220</link>
    <description>Hi,&lt;DIV&gt;I was thinking about MacOS X support and I believe you must support because basically Apple OpenCL CPU implementation has a big limitation it only admits workgroups of 1 element so really can't execute programs optimized to be run on GPUs where workgroups are minimally some multiple of wavefront size (AMD speaking) or warp size (Nvidia speaking) so tipically &amp;gt;=128 at least..&lt;/DIV&gt;&lt;DIV&gt;Also lots of CUDA code avaiable when ported to OpenCL will not run on MacOSX CPU backend for this very reason..&lt;/DIV&gt;&lt;DIV&gt;this renders the backend useless so I hope you port the backend..&lt;/DIV&gt;&lt;DIV&gt;Also you have advanced autovectorization,etc..&lt;/DIV&gt;&lt;DIV&gt;As in Macos X there isn't any ICD model I wouldn't have any objection linking to a Intel OCL library if that allows to use your efficient implementation..&lt;/DIV&gt;&lt;DIV&gt;Thanks&lt;/DIV&gt;</description>
    <pubDate>Wed, 01 Dec 2010 08:29:07 GMT</pubDate>
    <dc:creator>rtfss1gmail_com</dc:creator>
    <dc:date>2010-12-01T08:29:07Z</dc:date>
    <item>
      <title>One big reason to implement support for MacOS X!</title>
      <link>https://community.intel.com/t5/OpenCL-for-CPU/One-big-reason-to-implement-support-for-MacOS-X/m-p/774139#M220</link>
      <description>Hi,&lt;DIV&gt;I was thinking about MacOS X support and I believe you must support because basically Apple OpenCL CPU implementation has a big limitation it only admits workgroups of 1 element so really can't execute programs optimized to be run on GPUs where workgroups are minimally some multiple of wavefront size (AMD speaking) or warp size (Nvidia speaking) so tipically &amp;gt;=128 at least..&lt;/DIV&gt;&lt;DIV&gt;Also lots of CUDA code avaiable when ported to OpenCL will not run on MacOSX CPU backend for this very reason..&lt;/DIV&gt;&lt;DIV&gt;this renders the backend useless so I hope you port the backend..&lt;/DIV&gt;&lt;DIV&gt;Also you have advanced autovectorization,etc..&lt;/DIV&gt;&lt;DIV&gt;As in Macos X there isn't any ICD model I wouldn't have any objection linking to a Intel OCL library if that allows to use your efficient implementation..&lt;/DIV&gt;&lt;DIV&gt;Thanks&lt;/DIV&gt;</description>
      <pubDate>Wed, 01 Dec 2010 08:29:07 GMT</pubDate>
      <guid>https://community.intel.com/t5/OpenCL-for-CPU/One-big-reason-to-implement-support-for-MacOS-X/m-p/774139#M220</guid>
      <dc:creator>rtfss1gmail_com</dc:creator>
      <dc:date>2010-12-01T08:29:07Z</dc:date>
    </item>
    <item>
      <title>One big reason to implement support for MacOS X!</title>
      <link>https://community.intel.com/t5/OpenCL-for-CPU/One-big-reason-to-implement-support-for-MacOS-X/m-p/774140#M221</link>
      <description>&lt;P&gt;Thanks for sharing this great feedback with us.&lt;/P&gt;&lt;P&gt;&lt;BR /&gt;
Intel does not comment on features and availability of third party products. We
are aware of Apple plans and evaluating support on our CPU-based platforms for
different Operating Systems. We suggest you speak directly with any such third
party for direct comment.&lt;/P&gt;&lt;P&gt;&lt;/P&gt;</description>
      <pubDate>Wed, 01 Dec 2010 09:49:49 GMT</pubDate>
      <guid>https://community.intel.com/t5/OpenCL-for-CPU/One-big-reason-to-implement-support-for-MacOS-X/m-p/774140#M221</guid>
      <dc:creator>ARNON_P_Intel</dc:creator>
      <dc:date>2010-12-01T09:49:49Z</dc:date>
    </item>
  </channel>
</rss>

