Most of the computationally expensive parts of QE have been progressively extracted from the code in the form of "libraries", which are parallelized both with MPI and openmp and mostly support GPU. Please have a look at this paper: https://arxiv.org/abs/2104.10502 kind regards -- Lorenzo Paulatto - Paris On Jan 29 2022, at 10:16 pm, Husak Michal <michal.hu...@vscht.cz> wrote: > Hi > > > I was comparing the speed of QE build with > --enable-openmp > and > --disable-openmp > MPI was disabled ... > > From what I see OpenMP is activated but the speed gain is zero ... > I was searching for OpenMP optimization flags in QE fortran code like > !$OMP PARALLEL etc ... > But I did not find any in important source codes ... > Did I not catch something or is there really no OpenMP optimization in QE and > eventual MP speed up can be obtained only by the use of external (linear > algebra, fft) libraries ? > Michal Husak UCT Prague > _______________________________________________ > Quantum ESPRESSO is supported by MaX (www.max-centre.eu) > users mailing list users@lists.quantum-espresso.org > https://lists.quantum-espresso.org/mailman/listinfo/users >
_______________________________________________ Quantum ESPRESSO is supported by MaX (www.max-centre.eu) users mailing list users@lists.quantum-espresso.org https://lists.quantum-espresso.org/mailman/listinfo/users