T. Brewer, "A Highly Scalable System Utilizing up to 128 PA-RISC Processors," in Digest of Papers, COMPCON'95, pp. 133--140, Mar. 1995.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....scope of this paper. Table 1 shows the set of problem sizes analyzed in this study. For conducting comparisons, smaller problem sizes were also analyzed, as reported in [11] The scientific benchmarks were analyzed on 1 to 32 processors using the default options. We analyzed the Convex Exemplar [12] implementation of NPB. However, to get a general characterization of these benchmarks, we undid some of the Exemplar specific optimizations. The six scientific benchmarks were instrumented, compiled, and analyzed on a 4 node Exemplar SPP1600 running SPP UX 4.2. The six scientific benchmarks are ....

T. Brewer, "A Highly Scalable System Utilizing up to 128 PA-RISC Processors," in Digest of Papers, COMPCON'95, pp. 133--140, Mar. 1995.

....size I. The SPLASH 2 benchmarks were developed in Stanford University to facilitate shared memory multiprocessor research and are written in C. The NPB are specified algorithmically so that computer vendors can implement them on a wide range of parallel machines. We analyzed the Convex Exemplar [13] implementation of NPB which is written in Fortran. The performance of an earlier version of this implementation is reported in [14] However, to get a general characterization of these benchmarks, we undid some of the Exemplar specific optimizations. The six scientific benchmarks were ....

T. Brewer, "A highly scalable system utilizing up to 128 PA-RISC processors," in Digest of papers, COMPCON'95, pp. 133--140, Mar. 1995.

....quantitative comparisons among different multiprocessors and selecting the best available application implementation techniques. In this paper, we present an experimental methodology and use it to characterize the memory, scheduling, and synchronization performance of the Convex SPP1000 [3]. We extend the microbenchmark techniques used in previous studies [4, 5, 6, 7, 8] by adapting them to deal with the important attributes of a DSM system. Like previous studies, this paper analyzes the cache, memory, and interconnection network performance. Moreover, special attention is given to ....

T. Brewer, "A Highly Scalable System Utilizing up to 128 PA-RISC Processors," in Digest of papers, COMPCON'95, pp. 133--140, Mar. 1995.

....large scale computational fluid dynamic applications. These benchmarks are specified algorithmically so that computer vendors can implement them on a wide range of parallel machines. In this report we present our analysis of the HP Convex implementation of NPB on the Convex Exemplar multiprocessor [3]. This implementation was mainly developed by Herb Rothmund of the HP Convex Technology Center. The performance of an earlier version of this implementation is reported in a NPB results report [4] We have analyzed two of the NPB kernels (CG and MG) and the three pseudo applications (SP, LU, and ....

T. Brewer, "A highly scalable system utilizing up to 128 PA-RISC processors," in Digest of papers, COMPCON'95, pp. 133--140, Mar. 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