[CP2K-user] too many communicators

[Ronald Cohen]

I am getting the following crashes reproducibly after several thousand time 
steps:

 *******************************************************************************
 *             MPI error 672742415 in mpi_cart_sub @ mp_cart_sub : Other 
MPI   *
 *   ___         error, error stack:
PMPI_Cart_sub(228)..................:     *
 *  /   \        MPI_Cart_sub(comm=0xc4040010, 
remain_dims=0x7f200ce27e80,     *
 * [ABORT]                       
comm_new=0x2b5549318b64)                      *
 *  \___/              failed
PMPI_Cart_sub(167)..................:            *
 *    |                   
MPIR_Comm_split_impl(253)...........:               *
 *  O/|        
MPIR_Get_contextid_sparse_group(604): Too many communicators   *
 * /| |                 (0/32768 free on this process; 
ignore_id=0)            *
 * / \                                                    
dbcsr_mpiwrap.F:1098 *
 *******************************************************************************


 ===== Routine Calling Stack ===== 

           14 mp_cart_sub
           13 dbcsr_complete_redistribute
           12 copy_dbcsr_to_fm
           11 cp_dbcsr_sm_fm_multiply
           10 make_basis_sm
            9 reorthogonalize_vectors
            8 wfi_extrapolate
            7 scf_env_initial_rho_setup
            6 init_scf_run
            5 qs_energies
            4 qs_forces
            3 velocity_verlet
            2 qs_mol_dyn_low
            1 CP2K

It seems that there is a communicator leak--more communicators are created 
than destroyed. This is using the latest code for github . I am using 2019 
intel compilers and mkl ,
but it seems to be a problem in cp2k dbscr. I am running psmp hybrid with 
384 cores and OMP_NUM_THREADS=4 . I am stuck until I can get pas this.

Thanks!

Ron


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20190909/b3d7eefa/attachment.htm>