Run-time partitioning of scientific continuum calculations running on multiprocessors
A wide range of scientific continuum calculations typically concentrate computational effort non-uniformly over localized regions of physical space. A run-time partitioning strategy is presented intended for such methods, that distributes work evenly across a team of processors and that can exploit the spatial localization present in the original computation in order to avoid high overhead costs. The strategy was tested on Anderson's Method of Local Corrections, a type of vortex method for computational fluid dynamics. Because computational effort follows particles that congregate and disperse irregularly about the domain, this problem is hard to partition in a way that distributes the work evenly among the processors. Experiments were run on 32 processors of an Intel Personal Scientific Computer-a message-passing hypercube multiprocessor-and on 4 processors of a Cray X-MP-a shared-memory vector architecture- and achieved good parallel speedups of 22 and 3.6, respectively. The partitioner may be implemented as a virtual machine (VM) and made available to the programmer as a library of run-time utilities. The semantics of the VM are insensitive to the application and to the computer architecture on which the VM is implemented.
- Research Organization:
- California Univ., Berkeley (USA)
- OSTI ID:
- 7135306
- Country of Publication:
- United States
- Language:
- English
Similar Records
Run-time partitioning of scientific continuum calculations running on multiprocessors
Programming abstractions for run-time partitioning of scientific continuum calculations running on multiprocessors