Re: Proposed Example for 11.6.2
Bill Nitzberg (nitzberg@nas.nasa.gov)
Wed, 25 Sep 1996 18:55:08 -0700
In message <199609260117.UAA06811@abacus> Rajeev Thakur writes
>
> Since we are now supporting MPI_Write_shared_ordered and
> MPI_Read_shared_ordered, I'm wondering if scatter-gather and
> scatterv-gatherv would be needed at all. One could do the same using
> the shared_ordered routines.
>
> Rajeev
The *_shared_ordered routines are specifically targeted to supporting
I/O when the I/O access size keeps changing. They will certainly mimic
a scatter/gather, but the synchronization required to implement the
*_shared_ordered routines is likely to make them slower than an
equivalent *_all routine with a scatter/gather filetype.
- bill