Hello GRASS hoppers:
Has someone succesfully ported GRASS to a Silicon Graphics
ORIGIN 2000 and if so, what are the switches for the compiler
to include in the head file.
Our ORIGIN 2000 has 10 RS10,000 processors, 1.25GBytes of RAM and
runs IRIX 6.4 OS
I suppose the first step is to port it for one processor and then try
with multi-threads. I want to port the modules that take a lot of time to run
on workstations like r.watershed, r.surf.contour, s.surf.idw,
s.surf.tps.
Any help or hint is welcomed.
Best regards
Alejandro Hinojosa
alhinc@cicese.mx http://www.cicese.mx/~alhinc
Research Associate
Earth Sciences Division
CICESE
Ensenada, Baja California, Mexico.
Dear Alejandro,
Has someone succesfully ported GRASS to a Silicon Graphics
ORIGIN 2000 and if so, what are the switches for the compiler
to include in the head file.
Our ORIGIN 2000 has 10 RS10,000 processors, 1.25GBytes of RAM and
runs IRIX 6.4 OS
I suppose the first step is to port it for one processor and then try
with multi-threads. I want to port the modules that take a lot of time to run
on workstations like r.watershed, r.surf.contour, s.surf.idw,
s.surf.tps.
Being the original designer/author of r.watershed and r.surf.contour
algorithms (with experience helping someone multi-thread my code), I can
safely say that r.surf.contour can be made faster using multiple
processors. Unfortunately, r.watershed's A* algorithm doesn't
multi-thread well: The time consuming parts must be done one cell at a
time, sequentially.
Based on what I know about s.surf.tps and s.surf.idw, they should be
multi-thread-able.
Good luck, chuck
--
Charles R. Ehlschlaeger, Assistant Professor
Department of Geography, Hunter College
695 Park Avenue New York, NY 10021
w: 212-772-5321 fax: 212-772-5268
e: chuck@everest.hunter.cuny.edu
http://everest.hunter.cuny.edu/~chuck/
"Come to New York, its not as filthy as you think"
- from Crazy People