HP-MPI V2.3 for Linux Release Note

Table 1-4 MPICH Wrappers
MPICH2MPICH1
mpirun.mpich2mpirun.mpich
mpicc.mpich2mpicc.mpich
mpif77.mpich2mpif77.mpich
mpif90.mpich2mpif90.mpich
Object files built with HP-MPI MPICH compiler wrappers can be used by an application
that uses the MPICH implementation. Applications built using MPICH compliant
libraries should be relinked to use HP-MPI in MPICH compatibility mode.
NOTE: Do not use MPICH compatibility mode to produce a single executable to run
under both MPICH and HP-MPI.
For more information on MPICH compatibility, see the HP-MPI User's Guide.
1.2.7.4 Support for Large Messages
The current MPI standard allows the amount of data transferred using standard API
calls to be greater than 2GB. For example, if you make a call to MPI_Send() that
contains a count of 1024 elements that each have a size of 2049KB, the resulting message
size in bytes would be greater than what could be stored in a signed 32-bit integer. In
some cases, previous versions of HP-MPI handled these incorrectly. HP-MPI V2.3 has
been updated to fix these deficiencies.
Additionally, some users working with extremely large data sets on 64-bit architectures
need to explicitly pass a count that is greater than the size of a 32-bit integer. This is
not allowed by the current MPI-2.1 standard. Until the standards committee releases
a new API that will handle this, HP-MPI V2.3 provides new APIs to handle large
message counts. These new APIs are extensions to the MPI-2.1 standard and will not
be portable across other MPI implementations. These new APIs will contain a trailing
L. For example, to pass a 10GB count to an MPI send operation, MPI_SendL() must
be called, not MPI_Send().
IMPORTANT: These interfaces will be deprecated when official APIs are included in
the MPI standard.
The other API through which large integer counts can be passed into HP-MPI calls is
the Fortran autodouble -i8 interface (which is also nonstandard). This interface has
been supported in previous HP-MPI releases, but historically had the limitation that
the values passed in must still fit in 32-bit integers since the large integer input
arguments were cast down to 32-bit values. For HP-MPI V2.3, that restriction is removed.
1.2 What’s in This Version 13