Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
MeshAdapt via Banded Interface (#788)
* Resolved logging and simmetrix compilation * LOG: Turn on logging for the anisotropic mesh before and after * FUNC: started adding in a hook for true model reconstruction capability * USR: Added conditional to ensure model reconstruction only when given input flag * ERM: Cleaned up some aspects of anisotropy. * API: temporary fix for model reconstruction * Anisotropy, changed to constrained volume and compute size scale and frame in same loop * API: reconstruction function incorporated * MAINT: cleaned out some debug statements * ANISO: cleaned up some code for anisotropic size field * BLD: update hashstack for updated scorec lib * FIX: useModel incorporate into PUMIDomain * RECONSTRUCTION: Fixed bug in loading a reconstructed mesh regarding m�material arrays by adding a necessary input flag into MeshAdaptPUMIDrvr * RECONSTRUCTION: turn off old reconstruction call in NumericalSolution and set useModel variable if not in inputs * RECONSTRUCTION: enable 2D reconstruction, generalization of data structures for 2D, and add default behavior to useModel flag * RECONSTRUCTION: change the flag needed to start model reconstruction * STACK: changed default stack hash to get updated scorec pkgs * Update .hashstack_default * ANISO: added maximum aspect ratio option, defaults to 100 * MA: remove input for forced adaptation * ANISO: preserve size frame and scales post adapt * minor typo fix in doc string * save the meshes before and after adapt with step number * adapt based on solve step * Removed clamping from anisotropy, gradeMesh instead of smoothing * Changed how the adapt inputs are passed in to preserve input fields * Added feature to search for error near interface when no error target is supplied, turned clamping back on * anisotropic grading added through aspect ratio scaling. * Use the minimum height function and change the error to size relationship * error-to-size relationship fix for isotropic, setup framework for anisotropic logic * FIX: fix some errors with API * FIX: resolve test for gmshCheck by adding in targetError * API: added several inputs/parameters to MeshAdaptPUMIDrvr including hPhi, numAdaptSteps (for controlling adaptation) in NumericalSolution * FIX: avoid double counting of solve step when counting for adaptivity * API: enable transfer of epsFact into size field considerations * STATE: save state of initial implementation of parallel gradation algorithm * FEATURE: implementation of parallel gradation * FIX/API: require the use of context in order to get domain and physics * FIX: fix merge conflict * FIX: fix conflicts from merge * FIX:avoid node material conflicts by creating a field beforehand and using that as the source of classification * MAINT: modified some pList references to use context, ct // turned off goodQuality input * FIX: resolve parallel assignment of node materials through parallel communication * FIX/MAINT: cleanup of debug statements, added description, simplify nodeMaterial procedure to resolve bug * OPT: move mesh derivation routine to immediately after mesh generation * ENH: improve reliability of updateMaterial from a derived model by taking the minimum material tag * MAINT: for more consistent behavior on HPC, use Zoltan only for now * MAINT: turn off mass check since not all applications have the requisite fields * FIX: correct numbering of logging meshes to prevent overlapping names in parallel cases * MAINT: removed some old code/clutter * ENH: Added edge-based band intersection detection for interface based refinement, accounting for parallel cases as well * MAINT/DOC: cleaned up debug statements * MAINT/ENH: Added initial adapt before solve and moved some code into a function * MAINT: turn adapt back on, remove unnecessary statements after adapt in initial adapt * SAVE: state of numericalsolution with list of needed data structures to transfer/rebuild * FIX: ensure consistency in adapt workflow by setting free DOFs from RD and MCorr * FIX: matched shock capturing for VOF by transporting old dofs and performing computations * Save state for debugging, single phase vbdf time integration with adaptivity workflow is consistent * FIX: generalized postAdapt functions as member functions of relevant classes,can handle multiphase 2D vbdf case with systemStepExact=True. * Turned off vof shock capturing lagging. Need to store u^{n-2} solution to get it working properly * ENH/FIX: Added secondary initial adapt and corrected how nUpdatesTimeHistoryCalled is updated after adapt * FIX: avoid memory error by properly destorying marker field * TRY/FIX: force shock capturing lagging post adapt based on incorrect numDiff last and rectify subgrid error array for lagging * UPDATE: hard-coded parameter for interface band size * FIX: wrap initial adapt in sfconfig dependent conditional * TESTS: fix tests that require epsFact information to be passed in * TEST: didn't finish fixing test from previous commit * FIX: needed to destroy some entities to clear memory * VMS: finish parallel VMS implementation with simple all_reduce sum. Also extension of centroid computation to 3D. * FIX: Add definition of postAdaptUpdate for backward euler cfl * MAINT: add warning if users ever call postAdaptUpdate() without having a proper definition * FIX: shock capturing depends on subgrid error so sge needs to be in loop when recovering the shock capturing * FIX: resolved issues with lagging with multiple models by defining appropriate stages that represent the model stagger. Also ensured adaptive time stepping is accounted for with the dtLast modification. * MAINT: consistency in function def * FIX: MCorr no longer needs to be corrected since LS ebqe array was fixed * FIX: account for situation when dtLast does not exist * FIX: is not instead of != * FIX: populate the lagged eddy-viscosity array. This for some reason is not sufficient to yield exactly the same results between pseudoadapt and non-adaptive case. 1e-12 error. * TST: added pseudo option for adapt to enable ease of testing * TST: added gauge comparison test for post-adapt restart/solution transfer via 2D dambreak case * TST: added gauge compare test to setup.py * TEST: hack to get really crude test of gauge comparison running. will need to reformat into template of other tests * FIX/TST: gauge compare test not being run by Travis * FIX/TST: change the install path for gauge_compare test to fix Travis error * DOC: add some documentation strings to the grading functions * DOC: added some more documentation strings
- Loading branch information