A code for carrying out stochastic simulations of the MPS econometric model, a simultaneous, nonlinear, dynamic system of equations involving approximately 600 variables, was adapted to run on the CYBER 205 vector processor. The effi ciency gained relative to performance on an IBM 3081-GX was significant: the CPU time re quired for a given simulation was reduced by a factor of 10 to 12. However, this is considerably smaller than the theoretical gain expected from the manufacturer's ratings. The basic causes of this smaller gain are the fairly small real memory of the CYBER 205 and the inefficiency of the CYBER 205 vector processor in performing ele mentary function look-up. The primitive nature of the operating system and some aspects of the FORTRAN compiler on the CYBER 205 required more reprogramming of our large code than such a conversion should demand. However, potential gains from such a conversion will be great once the deficiencies described above are eliminated.
|Original language||English (US)|
|Number of pages||28|
|Journal||International Journal of High Performance Computing Applications|
|State||Published - Dec 1987|
All Science Journal Classification (ASJC) codes
- Theoretical Computer Science
- Hardware and Architecture