<?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 OpenMP: Missing Privatization Warnings in Intel® Fortran Compiler</title>
    <link>https://community.intel.com/t5/Intel-Fortran-Compiler/OpenMP-Missing-Privatization-Warnings/m-p/745631#M3720</link>
    <description>The official way is by setting default(none) so as to require everything to be designated private or shared. This helps catch more mistakes than simply writing to shared variable.&lt;BR /&gt;The long way is to run with Intel Thread Checker, which should flag races caused by multiple threads modifying shared variables. A new replacement for it is under development.&lt;BR /&gt;Sun Studio compilers also performed some analysis, and would simply decline to observe omp parallel for various reasons, including such mistakes. It was quite a hassle to determine when and why parallelization was dropped, although some optional diagnostics were available. ifort takes more of the attitude "programmer knows best."</description>
    <pubDate>Thu, 15 Apr 2010 16:45:22 GMT</pubDate>
    <dc:creator>TimP</dc:creator>
    <dc:date>2010-04-15T16:45:22Z</dc:date>
    <item>
      <title>OpenMP: Missing Privatization Warnings</title>
      <link>https://community.intel.com/t5/Intel-Fortran-Compiler/OpenMP-Missing-Privatization-Warnings/m-p/745630#M3719</link>
      <description>&lt;P&gt;Mymost frequent error with OpenMP is forgetting to privatize variables. It's because I got used to the good old HP-UX f90 which would warn me in such cases. &lt;BR /&gt;Is anybody aware of a way to haveifort issue a warning if a scalar is modified inside a parallel section and if it was not privatized ? Found no hints in official documentation. This would be very valuable and should be trivial toimplement (if it's not already there and unknown).&lt;/P&gt;</description>
      <pubDate>Thu, 15 Apr 2010 13:13:32 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-Fortran-Compiler/OpenMP-Missing-Privatization-Warnings/m-p/745630#M3719</guid>
      <dc:creator>mriedman</dc:creator>
      <dc:date>2010-04-15T13:13:32Z</dc:date>
    </item>
    <item>
      <title>OpenMP: Missing Privatization Warnings</title>
      <link>https://community.intel.com/t5/Intel-Fortran-Compiler/OpenMP-Missing-Privatization-Warnings/m-p/745631#M3720</link>
      <description>The official way is by setting default(none) so as to require everything to be designated private or shared. This helps catch more mistakes than simply writing to shared variable.&lt;BR /&gt;The long way is to run with Intel Thread Checker, which should flag races caused by multiple threads modifying shared variables. A new replacement for it is under development.&lt;BR /&gt;Sun Studio compilers also performed some analysis, and would simply decline to observe omp parallel for various reasons, including such mistakes. It was quite a hassle to determine when and why parallelization was dropped, although some optional diagnostics were available. ifort takes more of the attitude "programmer knows best."</description>
      <pubDate>Thu, 15 Apr 2010 16:45:22 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-Fortran-Compiler/OpenMP-Missing-Privatization-Warnings/m-p/745631#M3720</guid>
      <dc:creator>TimP</dc:creator>
      <dc:date>2010-04-15T16:45:22Z</dc:date>
    </item>
    <item>
      <title>OpenMP: Missing Privatization Warnings</title>
      <link>https://community.intel.com/t5/Intel-Fortran-Compiler/OpenMP-Missing-Privatization-Warnings/m-p/745632#M3721</link>
      <description>You might also try the Source Checker feature of the compiler - I know it will detect many kinds of OpenMP coding errors, but am not certain about this one.</description>
      <pubDate>Thu, 15 Apr 2010 19:04:12 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-Fortran-Compiler/OpenMP-Missing-Privatization-Warnings/m-p/745632#M3721</guid>
      <dc:creator>Steven_L_Intel1</dc:creator>
      <dc:date>2010-04-15T19:04:12Z</dc:date>
    </item>
    <item>
      <title>OpenMP: Missing Privatization Warnings</title>
      <link>https://community.intel.com/t5/Intel-Fortran-Compiler/OpenMP-Missing-Privatization-Warnings/m-p/745633#M3722</link>
      <description>What do you mean Source Checker? Is this a sort of lint-ing functionality? &lt;BR /&gt;&lt;BR /&gt;it looks like something like&lt;BR /&gt;&lt;BR /&gt;ifort -diag-enable warn ... &lt;BR /&gt;&lt;BR /&gt;might invoke this functionality. Is this correct? Can this run without building object files (i.e. as linting combined with -syntax-only)?</description>
      <pubDate>Thu, 15 Apr 2010 22:20:31 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-Fortran-Compiler/OpenMP-Missing-Privatization-Warnings/m-p/745633#M3722</guid>
      <dc:creator>Izaak_Beekman</dc:creator>
      <dc:date>2010-04-15T22:20:31Z</dc:date>
    </item>
    <item>
      <title>OpenMP: Missing Privatization Warnings</title>
      <link>https://community.intel.com/t5/Intel-Fortran-Compiler/OpenMP-Missing-Privatization-Warnings/m-p/745634#M3723</link>
      <description>I recommend that you read the documentation for use of Source checker. Yes, it's sort of a cross-program-unit lint and it does not generate objects when used, so you would typically have an alternate build configuration to use it.&lt;BR /&gt;&lt;BR /&gt;I find that it often puts out a lot of noise for Fortran programs, so don't pay attention to everything it complains about, but it can be very useful.</description>
      <pubDate>Fri, 16 Apr 2010 00:40:01 GMT</pubDate>
      <guid>https://community.intel.com/t5/Intel-Fortran-Compiler/OpenMP-Missing-Privatization-Warnings/m-p/745634#M3723</guid>
      <dc:creator>Steven_L_Intel1</dc:creator>
      <dc:date>2010-04-16T00:40:01Z</dc:date>
    </item>
  </channel>
</rss>

