<div dir="ltr">Dear Aman, I am assuming you are talking about Xeon Phi x100 family (KNC). To prepare an ARCH file for Xeon Phi "native" is about cross-compiling the code using the "-mmic" flag. I think there was some practical experience at EPCC / University of Edinburgh. On the other hand there was also an attempt to fit into the offload infrastructure, which is in place for the GPGPU / ACC backend. This code did not make it into production, and it is hence stripped from the main branch. It is still present in the intel branch, but I have not put any more work into it recently. If you are talking about Xeon Phi x200 family (KNL), this is a whole different story and it works nicely. Let me summarize the steps for KNL: <ul><li>It works since CP2K 3.0 (released), and onward (CP2K 4.0-development). If you want to use the release or the master/trunk:</li><ul><li>git clone <a href="https://github.com/cp2k/cp2k.git" target="_blank">https://github.com/cp2k/cp2k.git</a> cp2k.git (or from SourceForge via Subversion or download of the official release)</li><li>ln -s cp2k.git/cp2k cp2k</li></ul><li>It works since pre-CP2K 3.0 and onward using the (current) CP2K/intel branch.</li><ul><li>git clone --branch intel <a href="https://github.com/cp2k/cp2k.git" target="_blank">https://github.com/cp2k/cp2k.git</a> cp2k.git</li><li>ln -s cp2k.git/cp2k cp2k</li></ul><li>I strongly suggest to use LIBXSMM (although you can run LIBSMM, which currently only supports up to AVX2).</li><ul><li>git clone <a href="https://github.com/hfp/libxsmm.git" target="_blank">https://github.com/hfp/libxsmm.git</a></li><li>git checkout tags/1.4.3 (if you don't want the master)</li></ul><li>I suggest to use my outgrown ARCH file(s), but you can also build your own (minimal) ARCH file.</li><ul><li>The ARCH files work for CP2K 3.0, and development (regardless of whether it is the master or the intel branch).</li><li>For 3.0/master/trunk branch, you have to download the following ARCH files into the cp2k/arch folder: <a href="goog_1622229635">https://github.com/cp2k/cp2k/</a><a href="goog_1622229635">raw/intel/cp2k/arch/Linux-x86-</a><a href="https://github.com/cp2k/cp2k/raw/intel/cp2k/arch/Linux-x86-64-intel.x">64-intel.x</a> (needed for PSMP, POPT, SOPT, and SSMP) <a href="https://github.com/cp2k/cp2k/raw/intel/cp2k/arch/Linux-x86-64-intel.psmp">https://github.com/cp2k/cp2k/raw/intel/cp2k/arch/Linux-x86-64-intel.psmp</a> POPT, SOPT, and SSMP configurations are available as well.</li></ul><li>Building CP2K</li><ul><li>LIBXSMM is auto-detected if the "cp2k" folder (perhaps a symlink; see above) and the "libxsmm" folder are located in parallel, if not: use LIBXSMMROOT=/path/to/libxsmm along with the make command.</li><li>source /opt/intel/compilers_and_libraries_2017.0.064/linux/bin/compilervars.sh intel64 </li><li>make ARCH=Linux-x86-64-intel VERSION=psmp AVX=3 MIC=1</li><li>There are quite some additional key-value pairs supported:</li><ul><li>LIBXSMM=2: intercepts all GEMM calls to go through LIBXSMM (gives some minor/additional performance)</li><li>SYM=1: add symbols "-g" (it also enables VTune support in LIBXSMM if Intel VTune is source'd, see <a href="https://github.com/hfp/libxsmm/#profiling">here</a>) </li><li>DBG=1: enables unoptimized code along with symbols (not for production runs, but for debug)</li><li>MKL=2: enables parallel MKL (default is sequential, but sequential should be fine)</li></ul><li>I recommend to use LIBINT, LIBXC, and ELPA (unless you know that your workload cannot benefit). If you want/need (optimal) recipes for these libraries, let me know.</li><ul><li>LIBINTROOT=/path/to/libint</li><li>LIBXCROOT=/path/to/libxc</li><li>ELPAROOT=/path/to/elpa</li></ul></ul></ul> Hans Am Freitag, 3. Juni 2016 16:21:29 UTC+2 schrieb Aman Jindal:<blockquote class="gmail_quote" style="margin:0;margin-left:0.8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr">Dear Users, Has anyone of you tried installing cp2k on Intel xeonphi. if yes, which arch file should be used (with what modifications)? thanks in advance, Aman </div></blockquote></div>