<?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 No performance boost with mkl on Pentium 4 ! Why ? in Intel® oneAPI Math Kernel Library</title>
    <link>https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/No-performance-boost-with-mkl-on-Pentium-4-Why/m-p/796034#M2678</link>
    <description>Hello,&lt;BR /&gt;&lt;BR /&gt;I am compiling very simple program for my school like this:&lt;BR /&gt;&lt;B&gt;gfortran ddot.f ddot_long.g -o ddot_long.exe&lt;/B&gt;&lt;BR /&gt;&lt;BR /&gt;and this is working fine.&lt;BR /&gt;&lt;BR /&gt;ddot.f can be downloaded from:&lt;BR /&gt;&lt;A href="http://lavica.fesb.hr/VPR/vektorske_operacije/ddot.f" target="_blank"&gt;http://lavica.fesb.hr/VPR/vektorske_operacije/ddot.f&lt;/A&gt;&lt;BR /&gt;ddot_long.g can be downloaded from:&lt;BR /&gt;&lt;A href="http://lavica.fesb.hr/VPR/vektorske_operacije/ddot_1.f" target="_blank"&gt;http://lavica.fesb.hr/VPR/vektorske_operacije/ddot_1.f&lt;/A&gt;&lt;BR /&gt;&lt;BR /&gt;with this compiling I get&lt;BR /&gt;Mflops = 113.62863&lt;BR /&gt;&lt;BR /&gt;If I use:&lt;BR /&gt;&lt;B&gt;gfortran -O3 ddot.f ddot_long.g -o ddot_long_03.exe&lt;/B&gt;&lt;BR /&gt;-O3 is maximum at my machine&lt;BR /&gt;then I get Mflops = 124.99220&lt;BR /&gt;&lt;BR /&gt;Which is normal because I use optimizations (-O3).&lt;BR /&gt;&lt;BR /&gt;But when I use mkl for optimization i do not get any performance.&lt;BR /&gt;Why is that so ?&lt;BR /&gt;I compile it like this:&lt;BR /&gt;&lt;B&gt;gfortran -L/opt/intel/mkl/10.2.6.038/lib/32/ -lmkl_core ddot.f ddot_long.f -o ddot_long_mkl.exe&lt;/B&gt;&lt;BR /&gt;&lt;BR /&gt;When I compile it like this I get Mflops = 113.62990, just like with no optimization.&lt;BR /&gt;Why is that so ?&lt;BR /&gt;Is it possible because I have old computer and I downloaded the newest mkl that mkl are not optimized anymore for my old processor (Pentimum 4 at 2.6Ghz).&lt;BR /&gt;Or am I just compiling it wrong ?&lt;BR /&gt;&lt;BR /&gt;I am working on debian-linux on P4 at 2.6 GHz&lt;BR /&gt;&lt;BR /&gt;thanks&lt;BR /&gt;</description>
    <pubDate>Mon, 01 Nov 2010 21:45:22 GMT</pubDate>
    <dc:creator>misko_st</dc:creator>
    <dc:date>2010-11-01T21:45:22Z</dc:date>
    <item>
      <title>No performance boost with mkl on Pentium 4 ! Why ?</title>
      <link>https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/No-performance-boost-with-mkl-on-Pentium-4-Why/m-p/796034#M2678</link>
      <description>Hello,&lt;BR /&gt;&lt;BR /&gt;I am compiling very simple program for my school like this:&lt;BR /&gt;&lt;B&gt;gfortran ddot.f ddot_long.g -o ddot_long.exe&lt;/B&gt;&lt;BR /&gt;&lt;BR /&gt;and this is working fine.&lt;BR /&gt;&lt;BR /&gt;ddot.f can be downloaded from:&lt;BR /&gt;&lt;A href="http://lavica.fesb.hr/VPR/vektorske_operacije/ddot.f" target="_blank"&gt;http://lavica.fesb.hr/VPR/vektorske_operacije/ddot.f&lt;/A&gt;&lt;BR /&gt;ddot_long.g can be downloaded from:&lt;BR /&gt;&lt;A href="http://lavica.fesb.hr/VPR/vektorske_operacije/ddot_1.f" target="_blank"&gt;http://lavica.fesb.hr/VPR/vektorske_operacije/ddot_1.f&lt;/A&gt;&lt;BR /&gt;&lt;BR /&gt;with this compiling I get&lt;BR /&gt;Mflops = 113.62863&lt;BR /&gt;&lt;BR /&gt;If I use:&lt;BR /&gt;&lt;B&gt;gfortran -O3 ddot.f ddot_long.g -o ddot_long_03.exe&lt;/B&gt;&lt;BR /&gt;-O3 is maximum at my machine&lt;BR /&gt;then I get Mflops = 124.99220&lt;BR /&gt;&lt;BR /&gt;Which is normal because I use optimizations (-O3).&lt;BR /&gt;&lt;BR /&gt;But when I use mkl for optimization i do not get any performance.&lt;BR /&gt;Why is that so ?&lt;BR /&gt;I compile it like this:&lt;BR /&gt;&lt;B&gt;gfortran -L/opt/intel/mkl/10.2.6.038/lib/32/ -lmkl_core ddot.f ddot_long.f -o ddot_long_mkl.exe&lt;/B&gt;&lt;BR /&gt;&lt;BR /&gt;When I compile it like this I get Mflops = 113.62990, just like with no optimization.&lt;BR /&gt;Why is that so ?&lt;BR /&gt;Is it possible because I have old computer and I downloaded the newest mkl that mkl are not optimized anymore for my old processor (Pentimum 4 at 2.6Ghz).&lt;BR /&gt;Or am I just compiling it wrong ?&lt;BR /&gt;&lt;BR /&gt;I am working on debian-linux on P4 at 2.6 GHz&lt;BR /&gt;&lt;BR /&gt;thanks&lt;BR /&gt;</description>
      <pubDate>Mon, 01 Nov 2010 21:45:22 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/No-performance-boost-with-mkl-on-Pentium-4-Why/m-p/796034#M2678</guid>
      <dc:creator>misko_st</dc:creator>
      <dc:date>2010-11-01T21:45:22Z</dc:date>
    </item>
    <item>
      <title>No performance boost with mkl on Pentium 4 ! Why ?</title>
      <link>https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/No-performance-boost-with-mkl-on-Pentium-4-Why/m-p/796035#M2679</link>
      <description>You are mistaken in assuming that the last compilation actually uses MKL. Since you included ddot.f in the command line, the linker would have no unsatisfied external routines that it would need to search for in the MKL libraries. If you asked a map file to be produced by the linker, you would see in it that no MKL routines were included in your executable. Alternatively, you can use nm on the executable and try to see if any MKL symbols are included.&lt;BR /&gt;&lt;BR /&gt;Leave out ddot.f from the command line and try again.&lt;BR /&gt;</description>
      <pubDate>Mon, 01 Nov 2010 22:25:34 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/No-performance-boost-with-mkl-on-Pentium-4-Why/m-p/796035#M2679</guid>
      <dc:creator>mecej4</dc:creator>
      <dc:date>2010-11-01T22:25:34Z</dc:date>
    </item>
    <item>
      <title>No performance boost with mkl on Pentium 4 ! Why ?</title>
      <link>https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/No-performance-boost-with-mkl-on-Pentium-4-Why/m-p/796036#M2680</link>
      <description>As mecej4 said, the presence of a ddot.o object in your build will prevent loading of ddot.o from MKL library, regardless of whether your ddot.f has the same or different function than the usual BLAS function. I don't think we can guess what your .g file is, or what use gfortran or ifort could make of it.&lt;BR /&gt;If you have vectorized ddot by your own compilation, the gain you may get from MKL ddot may be limited.</description>
      <pubDate>Mon, 01 Nov 2010 23:49:17 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/No-performance-boost-with-mkl-on-Pentium-4-Why/m-p/796036#M2680</guid>
      <dc:creator>TimP</dc:creator>
      <dc:date>2010-11-01T23:49:17Z</dc:date>
    </item>
    <item>
      <title>No performance boost with mkl on Pentium 4 ! Why ?</title>
      <link>https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/No-performance-boost-with-mkl-on-Pentium-4-Why/m-p/796037#M2681</link>
      <description>I understand what you have said.&lt;BR /&gt;&lt;BR /&gt;but when I try:&lt;BR /&gt;&lt;B&gt;gfortran ddot_long.f -L/opt/intel/mkl/10.2.6.038/lib/32/ -lmkl_core -o ddot_long_mkl.exe&lt;/B&gt;&lt;BR /&gt;&lt;BR /&gt;I get:&lt;BR /&gt;&lt;B&gt;/tmp/ccTxEZ73.o: In function `MAIN__':&lt;BR /&gt;ddot_long.f:(.text+0xb6): undefined reference to `ddot_'&lt;BR /&gt;collect2: ld returned 1 exit status&lt;/B&gt;&lt;BR /&gt;&lt;BR /&gt;It look like that ddot can not be found in MKL.&lt;BR /&gt;Is my conclusion correct ?&lt;BR /&gt;&lt;BR /&gt;&lt;B&gt;ddot_long.f can be downloded from:&lt;BR /&gt;&lt;/B&gt;&lt;A href="http://lavica.fesb.hr/VPR/vektorske_operacije/ddot_long.f" target="_blank"&gt;http://lavica.fesb.hr/VPR/vektorske_operacije/ddot_long.f&lt;/A&gt;&lt;BR /&gt;&lt;BR /&gt;So can you please explain me how to compile &lt;B&gt;ddot_long.f &lt;/B&gt;with mkl.&lt;BR /&gt;&lt;BR /&gt;Thanks</description>
      <pubDate>Tue, 02 Nov 2010 16:51:11 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/No-performance-boost-with-mkl-on-Pentium-4-Why/m-p/796037#M2681</guid>
      <dc:creator>misko_st</dc:creator>
      <dc:date>2010-11-02T16:51:11Z</dc:date>
    </item>
    <item>
      <title>No performance boost with mkl on Pentium 4 ! Why ?</title>
      <link>https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/No-performance-boost-with-mkl-on-Pentium-4-Why/m-p/796038#M2682</link>
      <description>The MKL library comprises many pieces, and linking it properly from Intel Fortran, and more so from another Fortran compiler, is not trivial. &lt;BR /&gt;&lt;BR /&gt;However, if you use &lt;A href="http://software.intel.com/en-us/articles/intel-mkl-link-line-advisor/"&gt;the link line advisor &lt;/A&gt;, things work fine. For example, using the 32-bit compiler and libraries, link using&lt;BR /&gt;&lt;BR /&gt; &lt;PRE&gt;[bash]gfortran -m32 ddot_long.o -L$MKLPATH $MKLPATH/libmkl_solver_sequential.a -Wl,--start-group -lmkl_gf -lmkl_sequential -lmkl_core -Wl,--end-group -lpthread[/bash]&lt;/PRE&gt; &lt;BR /&gt;</description>
      <pubDate>Tue, 02 Nov 2010 19:38:53 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/No-performance-boost-with-mkl-on-Pentium-4-Why/m-p/796038#M2682</guid>
      <dc:creator>mecej4</dc:creator>
      <dc:date>2010-11-02T19:38:53Z</dc:date>
    </item>
    <item>
      <title>No performance boost with mkl on Pentium 4 ! Why ?</title>
      <link>https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/No-performance-boost-with-mkl-on-Pentium-4-Why/m-p/796039#M2683</link>
      <description>Thnak you mecej4, it is working for me.</description>
      <pubDate>Sat, 06 Nov 2010 16:28:25 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/No-performance-boost-with-mkl-on-Pentium-4-Why/m-p/796039#M2683</guid>
      <dc:creator>misko_st</dc:creator>
      <dc:date>2010-11-06T16:28:25Z</dc:date>
    </item>
  </channel>
</rss>

