Code Meeting 10/07/2014 - Eddie

HLLC Solver

  • The solver passed all 5 tests from Ch. 6 in Toro. The test parameters are given on my wiki page: u/ehansen/buildcode
  • I tried adding transverse velocity (vy) to see if it would affect the solution to the Riemann problem. I used test 4 which is a shock/contact/shock test.
  • I tried running my clump simulation with the clump and ambient/wind moving in the x-direction instead of the y-direction as before. This would test if the sweep method is causing asymmetry. If we find that the sweep method is the culprit, it would still be strange for the HLLC solver to produce more asymmetry than the Exact solver. Perhaps it is the combination of the sweep method with HLLC that is bad.

movie​


Pulsed Jet Restarts

  • The hydro run finished, but it took almost 35 hours on 256 procs. I know that my original runs used more procs, so they ran in less time, but this still seems a bit high to me.
    • I found 18 restarts due to High CFL, which seems higher than normal.
    • How much slower should the code run because of the SII, SIII, and SIV tracers?
  • For the MHD runs, there are a lot of restarts that will make it impossible to finish. I tried the beta=5 model, and I only got 20 out of 100 frames in 48 hours on 256 procs. The estimated wall time remaining got up to about 1 month.
    • These are 'restarts due to nan in flux', which do not occur during the hydro run.
    • I attached a list of restarts that show position, time, CFL, maxspeed,etc. It seems like most of them occur at the head of the jet, but it's a bit hard to tell without overplotting things, and I didn't have enough time to figure out how to to that.

movie​

Attachments (5)

Comments

No comments.