Brandon D. Shroyer 2007-06-20

I successfully ran two tests of mpibear today, one with MaxLevels = 1 and one with MaxLevels = 2 (both of them had a CFL of 0.3). I was going to run a CFL = 0.4, MaxLevels = 0.2 test, but I decided to wait until the results of the single-processor tests (see below) completed.

Kris came up with several ideas to try to find the CFL problem in the single-processor run:

1. Try restarting from frame 12 to see if the problem persists (unsuccessful).

2. Set the viscLOF parameter in the physics.data file to 0.0 (down from 0.02) (unsuccessful).

3. Set the b4step function in contrib/astro/problem.f90 to RETURN without executing any code. This should apparently bypass a correction algorithm that hasn't been properly tested yet. I'm concerned about this one, simply because the program is moving a lot faster than it should (which suggests to me that an awful lot of code is not being executed), but we'll see how it goes.

UPDATE: Test 3 was successful in that it did not get a bad CFL. It did, however, stop after 22 frames claiming that bear.data options didn't want it to write any further.