Equation Solution High Performance by Design |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
JCL v. GPS
JCL and GPS are two heuristic algorithms. GPS source code (i.e., 582.for) is a public domain software. Both JCL and GPS are in Fortran. The following is a comparison of JCL and GPS on the examples in "Algorithms for reducing the bandwidth and profile of a sparse matrix", published in computers & structures, vol. 44, no. 3, 1992. The OPTPRO option of GPS is set to False which means bandwidth reduction has a higher priority.
From the above comparison, it can be seen that JCL completely beats GPS in bandwidth reduction. In profile reduction, JCL wins 7, and GPS has 3. Even JCL is bandwidth-reduction based, JCL is still better than GPS in profile reduction. JCL is a reliable heuristic algorithm. Even JCL outperforms GPS, GPS has been wide-applied in scientific and engineering computing. There are some research papers comparing "average bandwidth" to GPS's bandwidth reduction. This site believes it is not a fair comparison. GPS's concept is to stretch a level structure to a maximal depth. A maximal-depth level structure is for a minimal bandwidth, not a minimal "average bandwidth". "Average bandwidth" belongs to the scope of profile reduction. It is inappropriate to compare a profile reduction algorithm to a bandwidth reduction algorithm. This site compares JCL with GPS, both of which are bandwidth reduction algorithm. |
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||