Parallel Software Products     Please select sample OpenMP parallelisations of Fortran code performed by ParaWise from the menu below:

Runtime on up to 30 processors for 3 Mesh Sizes

Overflow runtimes

Speedup on up to 30 processors for 3 Mesh Sizes

Overflow speedups

NASA Overflow

The following results show the performance of the ParaWise generated parallel OVERFLOW Overset Grid Computational Fluid Dynamics simulation code on an SGI Origin 2000 (at NASA Ames). The parallelisation was carried out at NASA Ames by Dr H.Jin.

Code information: 100,000 lines of source and 870 subroutines

Total Parallelization Time using ParaWise/CAPO : Approximately 3 days.

User Time: Approximately 1 day.

Summary of OpenMP directives inserted by CAPO

PARALLEL regions : 243

PARALLEL + DO regions : 375

Parallel DO loops : 377

REDUCTION loops : 21

ATOMIC/CRITICAL sections : 68

Regions with FIRSTPRIVATE: 8

Duplicated routines : 26

As can be seen in the performance graphs on the left, for three different sizes of test case, the ParaWise generated OpenMP code shows scalability, with the larger test case size running over 15 times faster than the serial version when using 30 processors.