On Wed, Nov 23, 2011 at 18:48, Dmitry Karpeev
<karpeev@gmail.com> wrote:
> I wonder if we can express all useful communication with a more restrictive interface similar to a local-to-global map (but mapping between any vectors) where one side (typically called "local", but doesn't need to actually be a local Vec) has at most one edge.
I don't know what you mean. Say we have a scatter from X (sometimes "local") to Y (sometimes "global"). One way to describe my communication graph is as a single-valued function on X (containing either the index into Y or NULL). I'm asking whether the single-valued constraint is too restrictive for some important use case.
The primitives I can readily provide are
broadcast from Y to X
reduce values on X into Y (min, max, sum, replace, etc)
gather values on X into arrays for each point in Y (in some ordering, rank ordering if you like)
scatter values in arrays for each point in Y to the point in X
These latter two involve a data structure that is as big as the number of sharing processes (because that is explicitly being requested), but are still implemented where the owner of that point in Y has no knowledge of which processes it is communicating with. With the capability of hardware RMA, this can actually be more efficient than if the two-sided graph was available (and it's less code and fewer data structures).