Parallel Studio XE Requirement

[philip_o_]

We are deciding between the professional and cluster editions for use with a pair of Phis.

Is it mandatory to get the Cluster edition, or will the Professional edition suffice? The only differences seem to be Intel MPI related.

Would we not be able to use, for example GCC OpenMP? Could we instead use something like TBB?

 https://software.intel.com/en-us/articles/intelr-cluster-studio-xe-works-on-xeonr-phi-coprocessor-openmp-tbb-mpi deals with Cluster edition exclusively, and from https://software.intel.com/en-us/forums/topic/360754, an output appears to rely on Intel OpenMP.

You can check the shared library dependencies for this binary using the following command:

~#/usr/linux-k1om-4.7/bin/x86_64-k1om-linux-readelf -d ./a.out | grep NEEDED

0x0000000000000001 (NEEDED) Shared library: [libm.so.6]
0x0000000000000001 (NEEDED) Shared library: [libiomp5.so]

However, the Developer Quick Start explicitly appears to state that we can use Professional edition

You can purchase software development tools at http://software.intel.com/en-us/linux-tool-suites. Select the tool(s) that fit(s) your need (e.g., “Intel® Parallel Studio XE 2015 Cluster Edition” or the Intel® Parallel Studio XE Professional Edition)

I am hoping for some clarity on this concern so that we can make an informed choice for the purchase.

[TimP]

You are correct that cluster edition is needed only if you want mpi .  "Symmetric" mpi, where mpi runs on coprocessors as well as host, is available only for linux host. Vtune and openmp for mic are present in all parallel studio versions. If you are interested in mpi, it may be the best way to use multiple mic coprocessors. Tbb is in all c++ packages.

gcc openmp isn't available for knc coprocessor, nor are the vpu instructions implemented for knc, so it is important to have an intel compiler. I don't know whether tbb has been built with native gcc on mic, but it would likely be disappointing without vpu support.

there has been work on open source mpi for mic, but i haven't seen discussion of current status.

[jimdempseyatthecove]

You can also use asynchronous offloads with the XE version. This can be performed in the serial section of host code or from threads within a parallel region. You could also split the work amongst the host and your two Xeon Phi's.

Jim Dempsey