GRASS for Cray

[Duane F Marble]

Jan Hartmann asks about compiling GRASS on a Cray. While this could
technically be done, I feel that it would not be very effective if
approached from the standpoint that a Cray is just another computer.

First, the Cray is a batch-oriented machine and doing graphics normally
involves post-processing (e.g., we have a YMP 8/64 and can pass output
from a model to a Stellar 2000 for visualization). Second, as a multi-
processor machine the algorithms used should be changed to match the
architecture. Third, the Cray compilers do not provide any significant
degree of optimization over the multiple processors -- this must be
done largely by hand. If it is not done, you are not obtaining the
maximum benefit from the special architecture.

Some relevant information is contained in:

Jay Sandhu and Duane Marble, 1988. "An Investigation into the Utility
  of the Cray X-MP Supercomputer for Handling Spatial Data," PROCEEDINGS,
  Third International Symposium on Spatial Data Handling, pp. 253-268.


-- 
Duane F. Marble                  E-mail: dmarble at magnus.acs.ohio-state.edu
Department of Geography
The Ohio State University        Telephone: (614) 292-2250
Columbus, Ohio 43210             Fax: (614) 292-6213