From version 2017a on of the Intel toolchains, the setup on the UAntwerp is different from the one on some other VSC clusters:

  • A full install of all Intel tools for which we have a license at UAntwerp has been performed in a single directory tree as intended by Intel. There is a single module for the C/C++/Fortran compilers, Intel MPI, the libraries MKL (Math Kernel Library), IPP (Integrated Performance Primitives), TBB (Threading Building Blocks) and DAAL (Data Analytics Acceleration Library) and Intel-provided GDB-based debuggers. The Intel tools for code and performance analysis VTune Amplifier XE, Intel Trace Analyzer and Collector (ITAC), Intel Advisor XE and Intel Inspector XE are also installed, but these still have separate module files as they rely on overloading libraries in some cases.
  • There should be no need to run any of the configuration scripts provided by Intel, all variables should be set correctly by the module file. Contact user support if this is not the case. The configuration scripts should work as intended by Intel though should you want to use the compilers without loading the module.
  • Several variables specific for the way software is set up at the VSC are defined the way they would be defined if the Intel toolchain was defined in the standard VSC way through the module tree. As such, we expect that you should be able to use any Makefile developed for the standard VSC-setup.
  • All compiler components needed to develop applications with offload to a Xeon Phi expansion board are also provided in anticipation of the installation of such a cluster node for testing purposes in Leibniz.

Compilers

  • The compilers work exactly in the way described on the regular Intel toolchain web page, including the MPI compiler wrappers. All links to the documentation on that page are also relevant.
  • Man pages for all commands have also been installed.

Debuggers

  • Intel-adapted GDB debuggers have been installed
    • Debugging regular Intel64 applications: gdb-ia
    • Debugging applications with offload to Xeon Phi: gdb-mic
  • Manual pages and GNU info pages are available for both commands

Libraries

Math Kernel Library (MKL)

MKL works exactly as in the regular VSC Intel toolchain. See the MKL section of web page on the VSC Intel toolchain for more information.

Integrated Performance Primitives (IPP)

Threading Building Blocks (TBB)

Data Analytics Acceleration Library (DAAL)

Code and performance analysis

VTune Amplifier XE

ITAC - Intel Trace Analyzer and Collector

Advisor

  • What? Advisor is a code analysis tool that works with the compilers to give advise on vectorization and threading for both the Xeon and Xeon Phi processors.
  • How? Advisor uses output generated by the compiler when building a full optimized release build and as such expects that some additional options are specified when compiling the application. The resulting compiler output can then be analized using the advixe-gui command.
  • Module: Advisor/<toolchain version>, e.g., Advisor/2017a.
  • Documentation

Inspector