NIMROD TEAM MEETING Sunday, November 16 8:30 am - 2:00 pm Somerset Room DoubleTree Hotel Pittsburgh, PA MINUTES 1. Status of Code - Advection Advection in the momentum equation has been coded and included. Time advance is simple predictor/corrector, and no upwinding With the mass matrix the CFL number must be strictly less than 1 (not .le.). The predictor/corrector is quite diffusive. We may eventually need to find a higher order approach. Advection has not yet been put into Ohm's law. So far this term has had no noticeable effect on the nonlinear evolution. - Two-fluid Code still has same Ohm's law as at the July review. - Pressure Same as in the July review. Includes advection, but this term is not done with predictor/corrector. A predictor/corrector should be added to handle this term. Perhaps this accounts for some recent problems in the nonlinear regime. - Performance Performance depends on the preconditioner. Code spends most (65%-90%) of its time in the linear solve. Performance has been improved by incorporating ILU pre-conditioning. BLAS 1 and 2 routines have been tried, but actually slowed down the code. There are indications that the time step can be accurately decoupled from the Alfv=E9n time scale, so arbitrarily large time steps with accuracy are likely. 2. Status of GUI Being updated, but functionality has not been improved. - Is anyone using it? Most developers don't use it because they are already familiar with the code 2E - Is it useful? Remains to be seen. 3. Status of Graphics Alan has made a movie using the new fast method of Poincar=E9 mapping. It looks really good. Alan and Dan Laney have are developing 3-D graphics for DX showing isosurfaces of q. Homegrown graphics based on OPENGL is being considered as an option to replac DX. Nina Popova has completed the contour plots with multiple rblocks and tblocks. Her next tasks will be vector plots, menus, and JPEG output. 4. Status of Validation Plan - Ideal modes D3D ideal MHD cases at various values of bN have compared well with GATO results. Same for Soloviev cases. Galkin cases are underway. - Resistive modes Good comparison with DEBS (linear and nonlinear). - 2-fluid Stable whistler waves have had a successful convergence study. Rick is starting a comparison of growing solutions between NIMROD and DEBS. - Nonlinear evolution Good results for D3D in a torus and tearing modes in a doubly periodic cylinder. 5. Discussion of GA Acceptance Plan The team will develop a counter-proposal. The major disagreement is on the goals of the beta-test. 6. The Vision Thing Alan will write a compromise statement and circulate it to the group. Just a few bullets. - Dalton's vision - Carl's vision - Concensus 7. Next steps/plans Packed gridding near singular surfaces. The next high priority code development item is the vacuum region/moving separatrix. Rblock grid in vacuum region (Alan). 8. Linear Solver - LLNL?? Awaiting response from Steve Ashby. Alfonso will be checking ISIS on T3E to see of it can speed up the code. This should be increased over the Galkin case, which awaits Alan's packed grid. 9. Documentation/Library Alfonso has started the library. This will be linked to the web page. 10. Time and Date for next teleconference Meeting January/February in San Diego. SAIC will host. Teleconference in mid-December. DALTON WILL MAIL JULY PRESENTATIONS TO EVERYBODY. SEND COPIES OF PRESENTATIONS TO CURT, HE CAN PDF THEM AND GIVE THEM TO SCOTT TO PUT IT ON THE WEB. ----------------------------------------------------------------------------- Dalton D. Schnack