Dear Nick, Thank you very much. Bests, Mahmoud
From: Nick Papior <[email protected]> To: [email protected] Date: Wed, 2 Jan 2019 09:39:58 +0100 Subject: Re: [SIESTA-L] problem with "--map-by" Intel MPI is based on MPICH, the manual states that the used command is only known to work with OpenMPI. Please refer to Intel MPI command mpiexec for details on bindings. Den tir. 1. jan. 2019 kl. 22.01 skrev Mahmoud Payami Shabestari <[email protected] [mailto:[email protected]]>: Dear Siesta Users and Developers, First, HAPPY NEW YEAR to you! Second, I have compiled siesta-4.1-b4 with support both openmp and mpi. I am interested to check the performance gain on a two-sockets four-cores per socket box the following command: <<mpirun --map-by ppr:1:socket:pe=4 -x OMP_NUM_THREADS=4 -x OMP_PROC_BIND=true siesta RUN.fdf>> The compilation is done using intel compiler 2017. However, it complains about the "--map-by" switch. I would be grateful if you could please help to manage it. Best regards, Mahmoud Payami -- Kind regards Nick
