- removed comment
implement UIUC's speed-up in "evaluation" of spectral solution
Issue #1309
closed
The accompanying svn patch applies the recently released refinement of TwoPunctures by Vasileios Paschalidis and Zach Etienne at UIUC, greatly reducing the time taken to properly apply the solution of the spectral solve to all grid points.
The patch is relative to the ET_2012_11 release (though I suspect it would apply to the upcoming 2013_05 release without modification). It passes the TwoPunctures/test/bhns_eval test suite.
Keyword: twopunctures
Keyword: evaluation
Comments (9)
-
-
- changed status to open
- removed comment
-
- changed status to open
- removed comment
Patch looks fine. Only thing to '''possibly''' add is an option to switch from the old to the new method. If I remembers this all correctly then the results of the "fast" method should be bit-identical to those of the "slow" method since it is just a storing of intermediate results. Please apply.
-
- removed comment
Note: once someone comments on the added option, I am fine to apply this with proper "from..." comments (since Bernard most likely cannot, not can Vasilis).
-
- removed comment
If the results are the same, even within a tolerance (and this is tested by the existing test case, I think), then there should be no need to add the parameter.
I just looked at the test cases. There is no vacuum test that uses grid_setup_method = evaluation. I don't understand how the hydro initial data works in the BHNS tests in TwoPunctures/test; maybe these tests are sufficient to test the whole of TwoPunctures?
-
- removed comment
On second thought, I would be happier if the tests were testing more than just what happens on the x axis and in a norm. For something like this, I would expect several points tested in all directions. For example, 3D ascii output (probably downsampled from the full grid) of at least u and gxx.
-
- removed comment
It should be easy to set up a small 3D grid covering only a small region near a horizon, e.g. 4^3 grid points. Full 3D ASCII output should then work quite well.
-
- changed status to resolved
- removed comment
Applied as rev 133 of TwoPunctures. We do actually have a test that uses evaluation, in the bhns_eval test of TOVSolver. The results before and after the patch differ only in the last bit.
-
- edited description
- changed status to closed
- Log in to comment
- Assignee
- –
- Type
- Priority
- Status
- closed
- Component
- EinsteinToolkit thorn
- Milestone
- –
- Version
- ET_2012_11
- Votes
- 0
- Watchers
- 0
Thanks for doing the merge! Yes, there has only been one commit to TwoPunctures since the release, and this was to an unrelated file:
Since we are now in feature-freeze, this will have to wait until after the release, but it should be one of the first things that we review once that is done.