A set of routines which embeds the parallelization methodology used for CAYENNE and greatly facilitates parallel program development is given in the file CAYEN.FOR. This file along with the files PLIBFOR.FOR and PLIBMAC.MAR form the library of routines developed for the CAYENNE methodology.
Two SPICE input files are also in the directory: BJTADDER.SPI and MOSADDER.SPI. Outputs for these input files are: BJTADDERBST.SPO and MOSADDERBST.SPO. These files will verify the CAYENNE application.
Benchmarking this application on a VAX 8300 MP has yielded performance results from 1.5 to over 1.8 times the single stream version of SPICE. Results will vary due to the size of the data sets. Larger SPICE 2G.6 data sets will tend to yield greater performance; hence greater throughout.
Several command procedures have been included for ease of use. CAYENNE may be run in single stream or as a parallel application by specifying the number of subprocesses desired. The specification of zero subprocesses, at start up, would yield a single stream execution of CAYENNE, while a specification of two subprocesses would be ideal to run CAYENNE in parallel across two processors.
Click on DOC to display additional information about the product. Click on FTP to download from the FTP Achives.