M. D. Dikaiakos and J. Stadel, A performance study of cosmological simulations on messagepassing and shared-memory multiprocessors, in "Proc. 10th ACM International Conference on Supercomputing, ACM SIGARCH, Philadelphia, PA, 1996," pp. 94--101.  Home/Search   Document Details and Download   Summary   Related Articles   Check

.... [16] and a number of graphics algorithms [15] Dikaiakos and Stadel conducted performance comparisons of cosmological simulations on the Intel Paragon and KSR 2 machines, and found that the shared memory version running on the KSR 2 outperformed the message passing code running on the Paragon [2]. More recently, Oliker and Biswas examined the performance of a dynamic unstructured mesh adaptation algorithm using three different programming models and concluded that a multithreaded implementation on the Tera MTA was the simplest and showed the most promise [8] However, each programming ....

M.D. Dikaiakos and J. Stadel, "A performance study of cosmological simulations on message-passing and shared-memory multiprocessors," 10th International Conference on Supercomputing, 1996, 94-- 101.

....we consider is based upon splitting the longest side of each cell. A k d tree has the property that its height is logarithmic in the number of points. K d trees have been used in a number of N body implementations, including the early implementation of Appel[App85] and in the UW KD grav code[DS95] We show that in the worst case, an N body algorithm using k d trees is inferior to an algorithm with a spatially balanced decomposition. There is a bad case point set where the points all lie on a line. This one dimensional bad case can then be repeated in two and three dimensions to get the ....

M. Dikaiakos and J. Stadel. A performance study of cosmological simulations on messagepassing and shared-memory multiprocessors. This paper is available from http://wwwhpcc. astro.washington.edu/, 1995.

....we consider is based upon splitting the longest side of each cell. A k d tree has the property that its height is logarithmic in the number of points. K d trees have been used in a number of N body implementations, including the early implementation of Appel[App85] and in the UW KD grav code[DS95] We show that in the worst case, an N body algorithm using k d trees is inferior to an algorithm with a spatially balanced decomposition. There is a bad case point set where the points all lie on a line. This one dimensional bad case can then be repeated in two and three dimensions to get the ....

M. Dikaiakos and J. Stadel. A performance study of cosmological simulations on message-passing and shared-memory multiprocessors. This paper is available from http://www-hpcc.astro.washington.edu/, 1995.

....and systems, the optimization of application designs, and the extrapolation of properties of future systems. So far, systematic performance studies and experiments have provided important insights for the design of parallel and distributed systems and the tuning of applications (see, for instance, [7, 14, 16, 10, 11, 21]) The issue of performance is also very important in emerging Internet systems: numerous studies show that performance of systems and applications determines to a large extent the popularity of Internet services and user perceived Quality of Service [2, 18] Nevertheless, the more complex a ....

M. Dikaiakos and J. Stadel. A Performance Study of Cosmological Simulation on Message-Passing and Shared-Memory Multiprocessors. In Proceedings of the 10th ACM International Conference on Supercomputing. ACM, May 1996.

....the great volume of data needed to describe computations and communications performed during parallel execution, and the low availability of high end multiprocessor systems. Numerous research projects have collected and studied communication and computation patterns from challenging applications [3,4,17]. Most proceed by conducting instrumentation and performance monitoring on top of parallel software and hardware platforms. This approach has its obvious merits as it addresses the performance assessment of programs running on existing systems. It is accompanied, however, by inevitable drawbacks: ....

M. Dikaiakos and J. Stadel, A Performance Study of Cosmological Simulations on Message-Passing and Shared-Memory Multiprocessors. (Submitted for Publication, March 1995). 12

....The purpose of our research is to understand performance bottlenecks that arise when running TIPSY on networked environments. To this end, we employ input files generated from real simulations conducted by researchers of the HPCC group in the Department of Astronomy at the University of Washington [11, 6], and typical TIPSY sessions conducted in the same Department. The main question we seek to answer is where is time spent during TIPSY runs Furthermore, we focus on I O requirements of TIPSY, characterize its I O behavior, and compare our observations with insights derived from studies of ....

M. Dikaiakos and J. Stadel. A Performance Study of Cosmological Simulation on Message-Passing and Shared-Memory Multiprocessors. In Proceedings of the 10th ACM International Conference on Supercomputing. ACM, May 1996.

No context found.

M. D. Dikaiakos and J. Stadel, A performance study of cosmological simulations on messagepassing and shared-memory multiprocessors, in "Proc. 10th ACM International Conference on Supercomputing, ACM SIGARCH, Philadelphia, PA, 1996," pp. 94--101.

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