Parallel graphics algorithms on a 1024-processor hypercube
We have developed four parallel graphics algorithms for visualization of complex problems in PDE simulations, radar simulation, and other large applications on a 1024-node ensemble with a 16-node graphics device. We discuss the impact of system parameters on algorithm development and performance. Algorithmic issues include multistage routing of graphics data through the ensemble, non-hypercube mappings from the ensemble to the graphics system, synchronization between ensemble and graphics nodes, and synchronization between graphic nodes. These issues apply to both the present and anticipated future systems which combine highly parallel ensembles and parallel I/O devices. ''Best'' solutions are described for routing, mapping and synchronization on the current hardware. Implications are discussed for future hardware and software for massively parallel computers. 6 refs., 2 figs., 2 tabs.
- Research Organization:
- Sandia National Labs., Albuquerque, NM (USA)
- DOE Contract Number:
- AC04-76DP00789
- OSTI ID:
- 5956894
- Report Number(s):
- SAND-89-0550C; CONF-890372-12; ON: DE89015492
- Country of Publication:
- United States
- Language:
- English
Similar Records
A radar simulation program for a 1024-processor hypercube
Molecular dynamics simulations of short-range force systems on 1024-node hypercubes