10/10/2012
The new set of simulations for the matched case has been a bit difficult getting off the ground. First, I was unable to get frame 0, due to refinement to highest level being triggered in ambient, ostensibly due to the default gradient in phi refinement criteria. By adding clearAllRefinements to grid, I was able to run the sim.
This past week I learned why the simulation was then getting to frame 13 and slowing down A LOT. It was because a much larger filling fraction of the amr mesh was being triggered due to density gradients forming in the ambient medium:
Now, this triggered refinement may be due to boundary effects, and thus the gradients are extraneous information that isn't important. Or, they may be physically meaningful, and something we may want to actually resolve. If this is the case, then I will have to see about moving to much more processors/memory to carry out the calculation further. For now, the fast method to get data is to de-refine outside of the BE radius.
What I'd like to accomplish with my refinement: -derefining outside of the sphere on everything -refining inside of the sphere on 2 things: jeans and density gradients
-The code I used was:
CALL AddRefinementCriterion(Mass_Field) CALL AddRefinementThreshold(JeansLength_Field,LESSTHAN, (/(4d0*levels(i)%dx, i=0, MaxLevel)/)) CALL CreateRefinement(Refinement) CALL CreateShape(Refinement%Shape) Refinement%Shape%size_param=(/1d0, 1d0, 1d0/) CALL SetShapeBounds(Refinement%Shape) Refinement%BufferCells=4 Refinement%tolerance=-1d0
I noticed 2 things of interest:
1) When I do not add the de-refine lines (from the 3rd call, on), the entire ambient was being refined to level 1, whereas the max level was being reached only inside of the sphere:
This seems to make sense if that is being triggered due to Jeans' criteria, which I am asking the code to use with the lines:
CALL AddRefinementCriterion(Mass_Field) CALL AddRefinementThreshold(JeansLength_Field,LESSTHAN, (/(4d0*levels(i)%dx, i=0, MaxLevel)/))
A quick calculation shows that, yes, the Jeans length is < 4*dx in the ambient, so this makes sense:
2) When I add the de-refine outside of the BE sphere, the ambient is coarse, and only the sphere is refined maximally:
I am happy with this refinement for now and am running in the AFRANK queue now. The higher levels of refinement inside of the BE sphere should be restricted to Jeans, by adjusting the qtolerance to have low tolerance for levs 0-5, and high levels for 6-7.
Questions
- How does cells/jeans work with ref subroutine?
- How does refVarFactor “ “ “ “
Attachments (5)
- forwikiJeans.png (34.3 KB) - added by 12 years ago.
- JeansMatched.png (23.2 KB) - added by 12 years ago.
- mesh1.png (15.6 KB) - added by 12 years ago.
- mesh2.png (12.3 KB) - added by 12 years ago.
- meshOct10.gif (130.2 KB) - added by 12 years ago.
Download all attachments as: .zip
Comments
No comments.