Home |
PC to parallel.
The phrase "PC to Parallel" has a double significance. First, it is a reminder that PHOENICS has for many years operated with full functionality on the whole range of hardware platforms, from personal computers (first DOS, now Windows) to super-computers.
An across-the-board parallelization of PHOENICS was effected during the 1990's. The adoption of the domain-decomposition technique enabled a single version to be employed for all the commonly-employed parallel-computer configurations.
These may be of shared-memory or distributed-memory configuration.
Recently, the second significance of the phrase has been realised: parallel PHOENICS now operates successfully, and with high efficiency, on clusters of personal computers linked by fast Ethernet, under the LINUX operating system.
The Benefits of Parallel CFD
The performance for the simulation of the flow of water around a ship's hull for different platforms is shown below; figure 1 shows the reduction of elapsed time brought about by increasing the number of processors, and figure 2 shows the speed increase versus number of processors.
The speed increase is seen to be almost linear with the number of processors.

Figure 1 Elapsed time versus number of processors

Figure 2 Speed increase versus number of processors.
Platforms
PHOENICS has now been ported and benchmarked on a wide range of parallel machines, indeed all MPP and SMP computers that support message passing (MPI or PVM) can now run parallel PHOENICS.
Systems Benchmarked:
CRAY J90 Intel Paragon
Digital AlphaServers Parsytec
IBM RS6000/SP, SiGr Power Challenge, Origin, Octane
Hewlett Packard HP-K240 , PC-Clusters (Intel Pentium 200 PCs)
Parallel PHOENICS is also available on the following systems:
Convex SPP1200, Hitachi SR2001
CRAY T3Em T3D and T90 NEC S-4
Fujitsu V-4, VPP500, VPP700Sun Servers