support new MPI "large count" routines

[roblatham00]

From slack:

On using the latest IOR version with MPICH(4.3.2), the actual MPIIO calls don’t occur in the DXT log.
MPIIO calls appear in the darshan log but things like MPIIO_BYTES_READ are zero
We think the issue might be the series of _C MPI(introduced with MPI 4.0) (eg. MPI_File_write_at_c) functions used by IOR; darshan may not be wrapping them.
(We didn’t have this issue with OpenMPI; it doesn’t have the _c functions, IOR just wraps the original set of MPI functions with _c suffix) (edited) 

recently MPI added "large count" support, but they didn't change the old routines that take 32 bit int. They added a bunch of new _c routines that take MPI_Count, a 64 bit (or maybe 128 bits on some platforms... something big) value.

These routines behave in every way like the traditional routines but are new symbols and Darshan will have to wrap them.

[carns]

Just to clarify, is IOR explicitly calling the _C variants?

[wkliao]

This may take some time.
Here is a list of MPI-IO large-count APIs, i.e. name ended with "_c".

int MPI_File_read_c(MPI_File fh, void *buf, MPI_Count count, MPI_Datatype datatype, MPI_Status *status)
int MPI_File_read_all_c(MPI_File fh, void *buf, MPI_Count count, MPI_Datatype datatype, MPI_Status *status)
int MPI_File_read_all_begin_c(MPI_File fh, void *buf, MPI_Count count, MPI_Datatype datatype)
int MPI_File_read_at_c(MPI_File fh, MPI_Offset offset, void *buf, MPI_Count count, MPI_Datatype datatype,
int MPI_File_read_at_all_c(MPI_File fh, MPI_Offset offset, void * buf, MPI_Count count,
int MPI_File_read_at_all_begin_c(MPI_File fh, MPI_Offset offset, void *buf, MPI_Count count,
int MPI_File_read_ordered_c(MPI_File fh, void *buf, MPI_Count count, MPI_Datatype datatype,
int MPI_File_read_ordered_begin_c(MPI_File fh, void *buf, MPI_Count count, MPI_Datatype datatype)
int MPI_File_read_shared_c(MPI_File fh, void *buf, MPI_Count count, MPI_Datatype datatype,
int MPI_File_write_c(MPI_File fh, const void *buf, MPI_Count count, MPI_Datatype datatype,
int MPI_File_write_all_c(MPI_File fh, const void *buf, MPI_Count count, MPI_Datatype datatype,
int MPI_File_write_all_begin_c(MPI_File fh, const void *buf, MPI_Count count, MPI_Datatype datatype)
int MPI_File_write_at_c(MPI_File fh, MPI_Offset offset, const void * buf, MPI_Count count,
int MPI_File_write_at_all_c(MPI_File fh, MPI_Offset offset, const void *buf, MPI_Count count,
int MPI_File_write_at_all_begin_c(MPI_File fh, MPI_Offset offset, const void *buf, MPI_Count count,
int MPI_File_write_ordered_c(MPI_File fh, const void *buf, MPI_Count count, MPI_Datatype datatype,
int MPI_File_write_ordered_begin_c(MPI_File fh, const void *buf, MPI_Count count, MPI_Datatype datatype)
int MPI_File_write_shared_c(MPI_File fh, const void *buf, MPI_Count count, MPI_Datatype datatype,
int MPI_File_iread_c(MPI_File fh, void *buf, MPI_Count count, MPI_Datatype datatype, MPIO_Request *request)
int MPI_File_iread_all_c(MPI_File fh, void *buf, MPI_Count count, MPI_Datatype datatype,
int MPI_File_iread_at_c(MPI_File fh, MPI_Offset offset, void *buf, MPI_Count count, MPI_Datatype datatype,
int MPI_File_iread_at_all_c(MPI_File fh, MPI_Offset offset, void *buf, MPI_Count count,
int MPI_File_iread_shared_c(MPI_File fh, void *buf, MPI_Count count, MPI_Datatype datatype,
int MPI_File_iwrite_c(MPI_File fh, const void *buf, MPI_Count count, MPI_Datatype datatype,
int MPI_File_iwrite_all_c(MPI_File fh, const void *buf, MPI_Count count, MPI_Datatype datatype,
int MPI_File_iwrite_at_c(MPI_File fh, MPI_Offset offset, const void *buf, MPI_Count count,
int MPI_File_iwrite_at_all_c(MPI_File fh, MPI_Offset offset, const void *buf, MPI_Count count,
int MPI_File_iwrite_shared_c(MPI_File fh, const void *buf, MPI_Count count, MPI_Datatype datatype,
int MPI_File_get_type_extent_c(MPI_File fh, MPI_Datatype datatype, MPI_Count *extent) ROMIO_API_PUBLIC;

[wkliao]

FYI. mpi4py version 4.0 calls MPI-IO large-count APIs.
When building mpi4py using MPICH, MPICH version 4.2.2 and later is required.

[roblatham00]

fixed in #1060