J. Bruck, D. Dolev, C.-T. Ho, M.-C. Rosu, and R. Strong, "Efficient message passing interface (mpi) for parallel computing on clusters of workstations," Tech. Rep. TR95-1474, Cornell University, February 1995.  Home/Search   Document Details and Download   Summary   Related Articles

....the other. Second, the effect of group size may depend on the underlying networking technology. Third, the plot can be used to compare an existing implementation with a theoretical lower bound. Our performance results for various broadcast operations include both results from the literature [17, 23, 24, 25] as well as our own measurements. Using these data, we produced the plotshown in Figure 4; several parallel platforms are represented. The data for the IBM SP 2 are from Xu and Huang s study [17] in which two different communication libraries, MPI and MPL, were compared on a 512 node SP 2. The ....

....are from Xu and Huang s study [17] in which two different communication libraries, MPI and MPL, were compared on a 512 node SP 2. The data for the Intel Paragon is from measurements of the InterComm collective communication library, also on a 512 node system [23] The data for the MPI CCL library [24] is from an implementation atop unreliable hardware supported multicast on Ethernet; group sizes of 4 and 8 are represented. The SS10 ASX100 data represent software implementations of broadcast operations through FORE systems ATM application interface forming spanning binomial tree configuration, ....

J. Bruck, D. Dolev, C.-T. Ho, M.-C. Rosu, and R. Strong, "Efficient message passing interface (mpi) for parallel computing on clusters of workstations," Tech. Rep. TR95-1474, Cornell University, February 1995.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC