3D Pulsed Jet on DAVinCI

I tested the new pulsed jet module on Rice's machine DAVinCI, and it seems to work well. This pulsed jet is 3D, MHD, with non-equilibrium cooling.

movie​

movie​

The code ran pretty well, producing chombos at a steady pace with no restarts. Here are some numbers to show how well this run did on DAVinCI:

Resolution 32 cells/Rjet
Cores 576
Run Time 2.5 hrs
Cost $0.49
% Run Completed 30%

The run does slow down as it moves along, so it's hard to say how long the full run would take at this resolution. The code estimates 11.2 more hrs after frame 30, but the actual time is probably closer to 24 hrs. If I assume that the time between chombos grows linearly, then I can estimate the total runtime to be about 24 hours.

Thus, the total run at 32 cells/Rjet would take 24 hours on 576 cores, costing 4.70. This is the maximum number of cores that I can use on DAVinCI, and I can only submit for a maximum of 8 hours at a time. So even at this low resolution a run is going to take more than a day if you include wait time and restarting. If we double the resolution to 64 cells/Rjet, this run could take 2 weeks or more instead of 1 day, and the cost would be more than 50.


UPDATE

Two new models: hydro (no MHD), and an MHD with beta = 1 (same as above) but with precession. Precession amplitude is 3 degrees and the precession frequency is 2.5 times the pulsation frequency.

movie​

movie​

Attachments (8)

Comments

No comments.