[Mpi-forum] MPI Count proposal from today's meeting
nmm1 at cam.ac.uk
Wed Jun 16 16:16:13 CDT 2010
On Jun 16 2010, Fab Tillier wrote:
> We reviewed the various plans that were presented since ticket #14 was
> originally introduced in September 2008. We then put up for a straw vote
> the following plan:
> Define MPI_Count in all existing APIs that use an int count.
> Implementations will decide whether to define MPI_Count as a C int, or
> something larger. Defining MPI_Count as a C int provides for backward
> compatibility with MPI 2.2 applications. Defining it as a 64-bit integer
> lets an implementation support large counts. Demand can drive which modes
> implementers will support.
That's sensible, except for one thing.
I hope nobody is insane enough to define it as a 64-bit integer in ANY
implementation. The alternative to 'int' should be the integer suitable
for array indexing, which might be of any size from 16 bits (due to MPI
tag and other constraints) upwards. This is not just a nit-pick.
Unfortunately, hope springs infernal, and he who hopes is sure to be
More information about the mpi-forum