<?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 MKL with gnu 4.2/4.3 compilers in Intel® oneAPI Math Kernel Library</title>
    <link>https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/MKL-with-gnu-4-2-4-3-compilers/m-p/856058#M7001</link>
    <description>I just completed some testing with gcc and gfortran, including gnu openmp, on linux x86-64. Apparently, it's inadvisable to activate gnu OpenMP when calling MKL OpenMP functions, due to incompatibilities between the OpenMP versions, although it "works" fairly often.&lt;BR /&gt;gfortran 4.3 implements an option -fexternal-blas, which implements MATMUL directly as a call to ?gemm. Performance calling MKL by this option may be better than the MATMUL incorporated in either gfortran or ifort libraries, but not as good as implementing the BLAS call directly.&lt;BR /&gt;I have also tested gfortran successfully on MPI applications, using the Intel MPI and MKL libraries.&lt;BR /&gt;</description>
    <pubDate>Fri, 08 Jun 2007 21:54:59 GMT</pubDate>
    <dc:creator>TimP</dc:creator>
    <dc:date>2007-06-08T21:54:59Z</dc:date>
    <item>
      <title>MKL with gnu 4.2/4.3 compilers</title>
      <link>https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/MKL-with-gnu-4-2-4-3-compilers/m-p/856058#M7001</link>
      <description>I just completed some testing with gcc and gfortran, including gnu openmp, on linux x86-64. Apparently, it's inadvisable to activate gnu OpenMP when calling MKL OpenMP functions, due to incompatibilities between the OpenMP versions, although it "works" fairly often.&lt;BR /&gt;gfortran 4.3 implements an option -fexternal-blas, which implements MATMUL directly as a call to ?gemm. Performance calling MKL by this option may be better than the MATMUL incorporated in either gfortran or ifort libraries, but not as good as implementing the BLAS call directly.&lt;BR /&gt;I have also tested gfortran successfully on MPI applications, using the Intel MPI and MKL libraries.&lt;BR /&gt;</description>
      <pubDate>Fri, 08 Jun 2007 21:54:59 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/MKL-with-gnu-4-2-4-3-compilers/m-p/856058#M7001</guid>
      <dc:creator>TimP</dc:creator>
      <dc:date>2007-06-08T21:54:59Z</dc:date>
    </item>
    <item>
      <title>Re: MKL with gnu 4.2/4.3 compilers</title>
      <link>https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/MKL-with-gnu-4-2-4-3-compilers/m-p/856059#M7002</link>
      <description>&lt;P&gt;This sounds like a status report as much as anything. &lt;/P&gt;
&lt;P&gt;You don't state anything about the impact on performance when using threaded gnu SW and MKL. In general, at the current time, this is an inadvisable thing to do, as noted in the user notes with MKL. In time, MKL will have a solution for this (successful threading in both the user program &lt;EM&gt;and&lt;/EM&gt; MKL using, and using a non-Intel compiler for the user program).&lt;/P&gt;</description>
      <pubDate>Mon, 11 Jun 2007 20:47:19 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/MKL-with-gnu-4-2-4-3-compilers/m-p/856059#M7002</guid>
      <dc:creator>Intel_C_Intel</dc:creator>
      <dc:date>2007-06-11T20:47:19Z</dc:date>
    </item>
    <item>
      <title>Re: MKL with gnu 4.2/4.3 compilers</title>
      <link>https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/MKL-with-gnu-4-2-4-3-compilers/m-p/856060#M7003</link>
      <description>Performance of MKL DGEMM was reduced when gfortran -fopenmp was used, even though DGEMM is called from a serial region.&lt;BR /&gt;</description>
      <pubDate>Mon, 11 Jun 2007 22:48:05 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/MKL-with-gnu-4-2-4-3-compilers/m-p/856060#M7003</guid>
      <dc:creator>TimP</dc:creator>
      <dc:date>2007-06-11T22:48:05Z</dc:date>
    </item>
  </channel>
</rss>

