Proposed Runs for 3D Mach Stems

2-clump runs

We vary the separation and Mach numbers.

Run Separation (rclump) Mfast, Mslow
A 2.5 10, 7.32
B 4.5 10, 7.32
C 6.5 10, 7.32
D 2.5 5, 3.66
E 4.5 5, 3.66
F 6.5 5, 3.66

On bluehive, I used about 3800 SUs per run for 40 cells/rclump resolution. Adding a level of refinement to get 80 cells/rclump will increase SUs by approximately a factor of 16 (a little less due to filling fractions). Accounting for the fact that Stampede cpus are slightly faster than bluehive (2.7 GHz vs. 2.4 GHz) leads to approximately 52,000 SUs per run. For 6 runs, this is a total of 312,000 SUs.


3-clump runs

Clumps stay at M = 2.5, 5, 10 for all runs. The M = 2.5 and M = 5 clumps remain in same plane at a fixed separation distance, and the M = 10 clump is positioned at some angle and distance from the M = 2.5 clump.

Looking down the barrel, the 3-clump setup is like this:

Run Separation (rclump) 3rd Clump Position (deg)
G 2.5 30
H 2.5 60
I 2.5 90
J 5 30
K 5 60
L 5 90

We would get longer run times because filling fractions are higher with 3 clumps. My best guess is that these runs could take about 77,000 SUs each. A total of 462,000 SUs for 6 runs.


10-clump run

Clumps at random positions with varying separations and velocities, but all moving in same general direction.

With many clumps on a large grid, this could be a very expensive run. I'm guessing it would be at least 100,000 SUs if we still want 80 cells/rclump.

In total, my crude estimates give on the order of 900,000 SUs for the entire study. Again, we're playing with a factor of ~16 here if we want the increased resolution. So if we keep the resolution at 40 cells/rclump, then we can do all these runs with less than 60,000 SUs. Maybe we only do the 2-clump runs at the increased resolution and the rest at the original, lower resolution.

Attachments (1)

Download all attachments as: .zip

Comments

No comments.