[mpiwg-rma] {MAX,MIN}LOC in RMA

[Jed Brown]

Jeff Hammond <jeff.science at gmail.com> writes:

>>> If NIC does FP for complex, this is fine.
>> 
>> Show me a NIC that does.
>
> Prove that one doesn't exist already. 
>
> See also: shared memory. 

Uh, what shared memory system provides atomic complex multiply?

>>> M**Loc isn't a commons HW atomic.

Prove that this doesn't exist.

You'll recall that I've nearly given up on RMA because Ops are not
extensible.  I can't implement anything important using RMA without also
implementing it over non-RMA because my code has to work with types like
float128 and longer pair types.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 835 bytes
Desc: not available
URL: <http://lists.mpi-forum.org/pipermail/mpiwg-rma/attachments/20140313/68314f8b/attachment-0001.pgp>