… However, intranode communication performance is much greater than internode communication
performance. In this paper, we show a simple approach that takes into account only …

… Low-power low-complexity wireless loop technology in small base units can be integrated …
personal communications services (PCS) to small, lightweight, low-power personal voice and/…

… of this can be seen in Figure 1, which shows how the common assignment of MPI processes
consecutively on a node can lead to significantly greater internode communication than a …

… William D Gropp is the director and chief scientist of the National Center for Supercomputing
Applications and holds the Thomas M. Siebel Chair in the Department of Computer Science …

… describing the data to be moved, even if noncontiguous, in the MPI communication
routines. This can eliminate a extra copy performed by the user into a separate buffer (unfortunately, …

… can accurately estimate the cost of communication, but at a significantly increased cost [12,
16, 17]. Network contention has been previously modeled for collective communication, …

The "postal" model of communication [3, 8] T = α + βn, for sending n bytes of data between
two processes with latency α and bandwidth 1/β, is perhaps the most commonly used …

… can send six messages and receive six messages at the same time. Communication algorithms
that take advantage of this can … can be up to six-fold quicker than its generic counterpart. …

Now that the performance of individual cores has plateaued, future supercomputers will depend
upon increasing parallelism for performance. Processor counts are now in the hundreds …

… William D Gropp is a director and chief scientist of the National Center for Supercomputing
Applications and holds the Thomas M. Siebel Chair in the Department of Computer Science …