Building the Intel Optimized MP LINPACK Benchmark

The Intel Optimized MP LINPACK Benchmark contains a few sample architecture makefiles, which you can edit to fit your configuration. Specifically:

• Set TOPdir to the directory where the Intel Optimized MP LINPACK Benchmark is being built.
• (Optional) Set MPI variables: MPdir, MPinc, and MPlib.
• Specify the location of Intel MKL and of files to be used (LAdir, LAinc, LAlib).
• Adjust compiler and compiler/linker options.

• Specify the architecture by setting the architecture parameter arch for the make command:

  • ia32 - for MPI processes to run on a 32-bit host processor.

  • intel64 - for MPI processes to run on a 64-bit host processor and use the Intel MKL hybrid offload methodology if Intel Xeon Phi coprocessors are available on the cluster.

  • mic - for MPI processes to run only on the Intel Xeon Phi coprocessors.

    In this case, the makefile creates pure native mode binaries. Note that this option is limited by the size of memory on the coprocessors. For more details, see Usage Modes of Intel Optimized MP LINPACK Benchmark for Intel® Xeon Phi™ Coprocessors.

• By default, the non-hybrid version of MP LINPACK is built. To build a different version, set the version parameter for the make command to hybrid or offload, as appropriate.

  For example:

  make arch=intel64 version=offload install

The makefile creates the binary in the bin/<arch> subdirectory of your mp_linpack directory (for example, bin/intel64).

The makefiles contain common values for some sample cases, such as Linux systems based on the Intel® 64 architecture. To change these values, you must be familiar with building an HPL and picking appropriate values for the variables in the makefiles.