[MPI3-IO] [EXTERNAL] New MPI-IO routines
chaarawi at hdfgroup.org
Thu May 24 17:19:04 CDT 2012
Ok I understood your original question differently, but anyway, yes I
can create a library on top of MPI to do what you mentioned. I don't see
why, for this case, it is better.. A lot of the information that is
needed to determine whether an operation depends on the progress of
another is present in the MPI library. Doing the same thing again at a
higher layer (i.e. decoding datatypes, hand out generalized requests,
etc..) isn't very encouraging.
On 5/24/2012 4:29 PM, Dries Kimpe wrote:
> And on the other hand, if this can be done equally well on top of MPI, it
> should be done on top as that is the fastest way to provide this new
> functionality to the general public (without having to wait for MPI-3.1)
I agree with you there..
More information about the mpiwg-io