[mpich-discuss] Why do predefined MPI_Ops function elementwise in MPI_Accumulate, but not in MPI-1 routines?

[Jed Brown]

On Wed, Apr 25, 2012 at 09:55, Jim Dinan <dinan at mcs.anl.gov> wrote:

> Yes, one of the design goals behind MPI is that it should be implementable
> in hardware.
>

The expectation is that every combination _will_ be implemented in
hardware? Or that the common combinations will be implemented in hardware
and others will fall back to software?

Could you tell me which hardware implements MPI_MINLOC on
MPI_LONG_DOUBLE_INT and MPI_PROD on MPI_COMPLEX32?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/mpich-discuss/attachments/20120425/05a59174/attachment.htm>