[Mpi3-rma] GetAccumulate restriction needed?
Underwood, Keith D
keith.d.underwood at intel.com
Sun Apr 3 21:45:55 CDT 2011
> > Would you think the user would expect that having the same source and
> > target buffer would cause the implementation to:
> >
> > -allocate a buffer
> > -perform a get into that buffer
> > -wait for get completion
> > -perform an atomic
> > -wait for atomic completion
> > -copy the data from the internal buffer to the user buffer
>
> Why are you trying to it atomically? Get_accumulate is only element-
> wise
> atomic.
Er, atomic only in the sense that Get_accumulate implies... So, replace "atomic" with "put" above and you still have something ugly. Point is: you have to wait for the get of a segment to complete before you can do the put. You have to wait for the put of a segment to complete before you can copy the result of the get (otherwise your retransmits are going to cause you great fun).
Keith
More information about the mpiwg-rma
mailing list