Dear all, I managed to compile QE4.1 with MPI, MKL and (perhaps) with MKL_FFTW3, with better performances, but I'm still getting strange timing results. For examples, the "tests" are finished in twice the time of the reference (1500 s instead of about 700), but the example21 with 32 water molecules is done in 1'25", whereas the 64 water molecules is done in 6'50"
Since mpiexec apparently run normally, I suppose QE is correctly compiled with MPI. Now, my question is: how can I be sure that MKL_FFTW3 is used? I cannot understand if the clib/fftw.h should be (or not?) correctly replaced by /opt/intel-10.0/Compiler/11.0/081/mkl/include/fftw/fftw.h I included the dir in make.sys, but the LIBOBJS (containing flib.a and clib.a) are (I suppose) linked before the LIBS, therefore I suppose that the fftw.h used is the clib/fftw.h Thank you, Carlo -- ------------------------------------------------------ Carlo Nervi carlo.nervi at unito.it Tel:+39 011 6707507/8 Fax: +39 011 6707855 - Dipartimento di Chimica IFM via P. Giuria 7, 10125 Torino, Italy http://lem.ch.unito.it/
