Jeff Hammond
Jeff Hammond
Currently, only these are missing: ``` MPI_BSEND_OVERHEAD is missing MPI_KEYVAL_INVALID is missing MPI_ROOT is missing MPI_MAX_STRINGTAG_LEN is missing MPI_MAX_PSET_NAME_LEN is missing MPI_PACKED is missing ``` Some constants may be present...
Untested, but present: ``` mpi_group_f.F90: module procedure MPI_Group_rank_f08 mpi_group_f.F90: module procedure MPI_Group_size_f08 mpi_group_f.F90: module procedure MPI_Group_translate_ranks_f08 mpi_group_f.F90: module procedure MPI_Group_compare_f08 mpi_group_f.F90: module procedure MPI_Group_union_f08 mpi_group_f.F90: module procedure MPI_Group_intersection_f08 mpi_group_f.F90: module...
also, i'm not even detecting `MPI_BOTTOM` anywhere right now 🤣
it is impossible to determine if a buffer is `MPI_BOTTOM` in Fortran, without resorting to `LOC()`, which is non-standard.
design tests: - https://github.com/jeffhammond/HPCInfo/blob/master/mpi/abi/sentinel.F90 - https://github.com/jeffhammond/HPCInfo/blob/master/mpi/abi/sentinel2.F90 - https://github.com/jeffhammond/HPCInfo/blob/master/mpi/abi/sentinel3.F90
This is what is there now: - MPI_File_open - MPI_File_close - MPI_File_delete - MPI_File_set_size - MPI_File_preallocate - MPI_File_get_size - MPI_File_set_view - MPI_File_read_at - MPI_File_read_at_all - MPI_File_read - MPI_File_read_all - MPI_File_write_at...
Set to true and assume compilers do this properly.
This is what is there now: - MPI_Type_commit - MPI_Type_free - MPI_Type_contiguous - MPI_Type_vector - MPI_Type_create_subarray
Do this with MPICH scripts instead. @dalcinl knows how to do it.
I'll use whatever works. The stuff you used to generate the ill-fated stubs would be really useful.