[GRASS5] GRASS on parallel CPUs

[Rich Shepard]

On Wed, 26 Apr 2000, Markus Neteler wrote:

> As far as I understand it is not a question of having a good compiler but
> a question of a re-write of GRASS libraries/modules. But I am not familiar
> with that.

Markus,

  I believe that there's more to it than that. The compiler has to know what
can be run in parallel and what cannot. The compiler also makes decisions on
which loops to unroll and how to optimize the code by inlining functions,
among other things. There are differences in the debuggers and profilers,
too. Now I've reached the end of my knowledge about modifying code to run on
multiple processors and/or clusters. :-)

  The other issue is the need for a high performance Fortran compliler along
with the C compiler. SWAT, for example, is written in Fortran and that's
probably where a lot of benefits from parallelization will be realized. For
example, We build a model of the Willamette River basin (from the Columbia
River at Portland on the north to below Eugene in the Siskiyou Mountains
about 200 miles the south) and define 300 subbasins. With a single processor
system, each subbasin is computed sequentially. With a 6-processor cluster,
a computational "run" would be 50 times faster because each processor would
run the same thread on a different subbasin.

  The Portland Group's cluster development kit (CDK) costs US$2,500, plus
$500 per year for minor upgrades and support. Cough! Cough! Compared with
the cost of GRASS and linux, this is equivalent to putting a US$100,000 mass
spectrophotometer as a detector on the back end of a $2,000 gas
chromatograph.

  That's my $0.25's worth (inflation, you know).

Rich

Dr. Richard B. Shepard, President

                       Applied Ecosystem Services, Inc. (TM)
              Making environmentally-responsible mining happen. (SM)         
                       --------------------------------
            2404 SW 22nd Street | Troutdale, OR 97060-1247 | U.S.A.
 + 1 503-667-4517 (voice) | + 1 503-667-8863 (fax) | rshepard at appl-ecosys.com


---------------------------------------- 
If you want to unsubscribe from GRASS Development
Team internal mailing list write to:
minordomo at geog.uni-hannover.de with
subject 'unsubscribe grass5'
length: 2999
max: 0