Experiments and improvements of the ordinary differential equation solvers DEPAC on the CRAY computer
We have looked at timing results when solving large systems of differential equations using DEPAC on the CRAY computer. By reprogramming some critical (previously nonvectorized) loops, using loop unrolling where appropriate, and using the CRAY $SCILIB routines, significant speedup factors have been achieved. For example, when overhead is dominant (i.e., the derivative evaluation cost per component is minimal), we have observed cost reductions by factors of 2 to 7 when compared to the unmodified codes. We have also incorporated three additional user options which enable faster execution in appropriate circumstances. The modified DEPAC will replace the earlier package as part of the SLATEC library for use on the CRAY. The computational results produced are identical to those obtained by the earlier codes.
- Research Organization:
- Sandia National Labs., Albuquerque, NM (USA)
- DOE Contract Number:
- AC04-76DP00789
- OSTI ID:
- 5640536
- Report Number(s):
- SAND-85-0856; ON: DE85013260
- Country of Publication:
- United States
- Language:
- English
Similar Records
Data Locality Enhancement of Dynamic Simulations for Exascale Computing (Final Report)
Fast and Accurate Simulation of the Cray XMT Multithreaded Supercomputer