Early Results From the Porting of the Computational Fluid Dynamics Code, F3D, tothe Silicon Graphics Power Challenge

摘要

This report briefly discusses some of the issues involved in porting acomputationally intensive code from a Cray vector supercomputer to the Silicon Graphics (SGI) Power Challenge (75-MHz Mips R8000-based shared memory Symmetric Multiprocessor). Additionally, some evidence is presented to indicate that similar results should be expected when taking this code to comparable boxes from other vendors (e.g., Digital Equipment, Convex, HP, or SUN). Performance results are also discussed. This discussion deals with both the highly successful nature of these results in terms of absolute levels of performance and the even more impressive results in terms of cost effective performance. This report concludes with some thoughts on the future of this project.