1) Implement the broadcast on a mesh in SR. Each node in the mesh should be able to run on a separate processor. Do not use global memory to pass informantion between nodes.
2) Implement the allsums algorithm on a shuffle network in SR. See page 343-245 of the Akl text. Each node in the mesh should be able to run on a separate processor. Do not use global memory to pass informantion between nodes.