Four-ary tree-based barrier synchronization for 2D meshes without nonmember involvement

This paper proposes a Barrier Tree for Meshes (BTM) to minimize the barrier synchronization latency for two-dimensional (2D) meshes. The proposed BTM scheme has two distinguishing features. First, the synchronization tree is 4-ary. The synchronization latency of the BTM scheme is asymptotically Thet a (log(4) n), while that of the fastest scheme reported in the literature i s bounded between Omega (log(3) n) and O(n(1/2)), where n is the number of member nodes. Second, nonmember nodes are neither involved in the construct ion of a BTM nor actively participate in the synchronization operations, wh ich avoids interference among different process groups during synchronizati on. This not only results in low setup overhead, but also reduces the synch ronization latency. The low setup overhead is particularly effective for th e dynamic process model provided in MPI-2. Extensive simulation study shows that, for up to 64 x 64 meshes, the BTM scheme results in about 40 similar to 70 percent shorter synchronization latency and is more scalable than co nventional schemes.