Sorry for the delay. I have now runned your big system, and I am not able to reproduce the problem. I would suggest you try with other linear algebra packages. The standard shipped BLAS+LAPACK packages are also *extremely* slow.
Den ons. 4. mar. 2020 kl. 22.12 skrev Giacomo Giorgi <giacsp...@gmail.com>: > Dear all, > following a previous unanswered message I would like to ask you your > opinion about this weird behavior of my calculations with siesta. > > I am working with TiO2 anatase. > Bulk worked perfectly. Also the (101) surface unit cell worked perfectly. > > Now I need a (5x3) supercell. (attached the fdf). > Running it as a replica of the initial unit cell I get the following error: > > .... > .... > cdiag: Error in Cholesky factorisation > Stopping Program from Node: 4 > 160 > cdiag: Error in Cholesky factorisation > Stopping Program from Node: 5 > 160 > cdiag: Error in Cholesky factorisation > Stopping Program from Node: 0 > -------------------------------------------------------------------------- > MPI_ABORT was invoked on rank 6 in communicator MPI COMMUNICATOR 3 CREATE > FROM 0 > with errorcode 1. > > NOTE: invoking MPI_ABORT causes Open MPI to kill all MPI processes. > You may or may not see output from other processes, depending on > exactly when Open MPI kills them. > -------------------------------------------------------------------------- > [node03:300332] PMIX ERROR: UNREACHABLE in file server/pmix_server.c at > line 2079 > [node03:300332] PMIX ERROR: UNREACHABLE in file server/pmix_server.c at > line 2079 > [node03:300332] PMIX ERROR: UNREACHABLE in file server/pmix_server.c at > line 2079 > [node03:300332] PMIX ERROR: UNREACHABLE in file server/pmix_server.c at > line 2079 > [node03:300332] PMIX ERROR: UNREACHABLE in file server/pmix_server.c at > line 2079 > [node03:300332] 23 more processes have sent help message help-mpi-api.txt > / mpi-abort > [node03:300332] Set MCA parameter "orte_base_help_aggregate" to 0 to see > all help / error messages > > > Maybe the structure is too large (strange...), thus I decided to run a > calculation on a smaller (1x3) supercell (still fdf atteched along with psf > used). > Then this time I get this new error: > ... > ... > Stopping Program from Node: 19 > Bad DM normalization: Qtot, Tr[D*S] = 576.00000000 > 638080922.01711190 > Stopping Program from Node: 21 > Bad DM normalization: Qtot, Tr[D*S] = 576.00000000 > 638080922.01711190 > Stopping Program from Node: 22 > -------------------------------------------------------------------------- > MPI_ABORT was invoked on rank 15 in communicator MPI COMMUNICATOR 3 CREATE > FROM 0 > with errorcode 1. > > NOTE: invoking MPI_ABORT causes Open MPI to kill all MPI processes. > You may or may not see output from other processes, depending on > exactly when Open MPI kills them. > -------------------------------------------------------------------------- > [node03:306841] 23 more processes have sent help message help-mpi-api.txt > / mpi-abort > [node03:306841] Set MCA parameter "orte_base_help_aggregate" to 0 to see > all help / error messages > > Since I do not find anyone able to fix this issue, which actually seems to > be system-independent ( I have had similar behaviour with other slabs), I > wonder if there is any of you able to help me. > > Thanks a lot- > Regards, > Giacomo > > > Siesta Version : v4.1-b4 > Architecture : unknown > Compiler version: GNU Fortran (GCC) 4.8.5 20150623 (Red Hat 4.8.5-36) > Compiler flags : mpif90 -O2 -fPIC -ftree-vectorize > PP flags : -DFC_HAVE_ABORT -DMPI -DSIESTA__DIAG_2STAGE > Libraries : libsiestaLAPACK.a libsiestaBLAS.a > /opt/share/scalapack_gnu/libscalapack.a > PARALLEL version > > * Running on 24 nodes in parallel > >> Start of run: 3-MAR-2020 10:41:40 > > > > > -- > G > > > "Oltre le illusioni di Timbuctù e le gambe lunghe di Babalù c'era questa > strada...Questa strada zitta che vola via come una farfalla, una nostalgia, > nostalgia al gusto di curaçao...Forse un giorno meglio mi spiegherò" > > (Paolo Conte, "Hemingway") > > -- > SIESTA is supported by the Spanish Research Agency (AEI) and by the > European H2020 MaX Centre of Excellence (http://www.max-centre.eu/) > -- Kind regards Nick
-- SIESTA is supported by the Spanish Research Agency (AEI) and by the European H2020 MaX Centre of Excellence (http://www.max-centre.eu/)