DOE ACTS COLLECTION Workshop Robust and High Performance Tools for Scientific Computing September 4-7, 2002

(a) Incorporating parallel solver(s) to run MODFLOW and SUTRA on Linux clusters

(b) Developing code to model subsurface transport of colloids (bacteria, protozoans).

The USGS California District is planning to install one or more Linux clusters in the near future. To realize the full potential of these clusters, we will need to incororate message-passing parallel solvers into several of our modeling codes. I am currently evaluating the feasibility of incorporating Aztec, PETSc and/or ScaLAPAK into several groundwater flow and transport codes (MODFLOW, MOC3D, MT3D/RT3D) written in FORTRAN-77.

I am also working with two other programs in the California District office which are also investigating incorporating message-passing parallel solvers into their modeling codes. One of these programs is developing density-dependend surface water flow and transport codes for the Sacramento-San Joaquin delta (currently run on a shared-memory parallel computer). The other is developing models linking precipitation to unsaturated zone flow in the western US, including use of TOUGH2 (which has a parallel solver from Aztrec).

My main goals from this workshop would be to learn (a) how to incorporate parallel sovers into legacy code, and (b) how to compile the resulting programs for operation on linux clusters (i.e. what is different than compiling for single-processor operation). It would be helpful if the workshop included some treatment of the various compiler options for optimizing parallel compilation in linux, but I realize this may be compiler- or computer-specific.Thanks for offering this workshop!