234. Atomicity


Up: Semantics and Correctness Next: Progress Previous: Semantics and Correctness

The outcome of concurrent accumulates to the same location, with the same operation and predefined datatype, is as if the accumulates where done at that location in some serial order. On the other hand, if two locations are both updated by two accumulate calls, then the updates may occur in reverse order at the two locations. Thus, there is no guarantee that the entire call to MPI_ACCUMULATE is executed atomically. The effect of this lack of atomicity is limited: The previous correctness conditions imply that a location updated by a call to MPI_ACCUMULATE, cannot be accessed by load or an RMA call other than accumulate, until the MPI_ACCUMULATE call has completed (at the target). Different interleavings can lead to different results only to the extent that computer arithmetics are not truly associative or commutative.



Up: Semantics and Correctness Next: Progress Previous: Semantics and Correctness


Return to MPI-2.2 Standard Index
Return to MPI Forum Home Page

(Unofficial) MPI-2.2 of September 4, 2009
HTML Generated on September 10, 2009