Todo List

Member ConjugateGradient::get_step (PatchData &pd, double f0, int &j, MsqError &err)

Michael NOTE: ConjugateGradient::get_step's int &j is only to remain consisitent with CUBIT for an initial test. It can be removed.

Michael NOTE: ConjugateGradient::get_step's int &j is only to remain consisitent with CUBIT for an initial test. It can be removed.

Class CornerJacobianQualityMetric

MB: CornerJacobianQualityMetric is currently not being evaluated using the corner jacobian method (instead it is using the area functions from MsqMeshEntity... needs to be modified to do so.

MB: CornerJacobianQualityMetric is currently not being evaluated using the corner jacobian method (instead it is using the area functions from MsqMeshEntity... needs to be modified to do so.

Member GridConversion::ExampleFunction (const std::string &instring)

Nothing to do on this example function.

Page How To Make A New IR Project

LICENSE

AUTHORS

SVN Repository Structure

Member LaplacianSmoother::optimize_vertex_positions (PatchData &pd, MsqError &err)

Michael: optimize_vertex_position is probably not implemented in an optimal way. We used to use all of the vertices in the patch as 'adjacent' vertices. Now we call get_adjacent_vertex_indices. We could use a VERTICES_ON_VERTEX type of patch or a global patch?

Michael: optimize_vertex_position is probably not implemented in an optimal way. We used to use all of the vertices in the patch as 'adjacent' vertices. Now we call get_adjacent_vertex_indices. We could use a VERTICES_ON_VERTEX type of patch or a global patch?

Class LPtoPTemplate

MB. Suggestions made by Todd Munson: a) There is an inconsistent use of fabs. The hessian evaluation when using the one norm does not take the absolute value, while the gradient does. b) The analytic gradient and hessian evaluations are incorrect when the quality metric changes sign due to taking the absolute value. The negative of the element gradient and hessian also needs to be taken. c) Done. The analytic gradient and hessian evaluations are incorrect when the negate flag is set to -1. The negative of the element gradient and hessian also needs to be taken in this case. d) The malloc in the concrete_eval routine should be removed.

MB. Suggestions made by Todd Munson: a) There is an inconsistent use of fabs. The hessian evaluation when using the one norm does not take the absolute value, while the gradient does. b) The analytic gradient and hessian evaluations are incorrect when the quality metric changes sign due to taking the absolute value. The negative of the element gradient and hessian also needs to be taken. c) Done. The analytic gradient and hessian evaluations are incorrect when the negate flag is set to -1. The negative of the element gradient and hessian also needs to be taken in this case. d) The malloc in the concrete_eval routine should be removed.

Member LVQDTargetCalculator::compute_target_matrices (PatchData &pd, PatchData &ref_pd, MsqError &err)

Michael: this function should be protected, but it is used in QualityMetricTest to avoid creating the MeshSet in the unit tests.

Michael: this function should be protected, but it is used in QualityMetricTest to avoid creating the MeshSet in the unit tests.

Member SmartLaplacianSmoother::optimize_vertex_positions (PatchData &pd, MsqError &err)

Michael: optimize_vertex_position is probably not implemented in an optimal way. We used to use all of the vertices in the patch as 'adjacent' vertices. Now we call get_adjacent_vertex_indices. We could use a VERTICES_ON_VERTEX type of patch or a global patch?

Michael: optimize_vertex_position is probably not implemented in an optimal way. We used to use all of the vertices in the patch as 'adjacent' vertices. Now we call get_adjacent_vertex_indices. We could use a VERTICES_ON_VERTEX type of patch or a global patch?

Member TerminationCriterion::get_current_function_value ()

Michael: this function is not reliable. It needs to be more robust. How do we know whether currentOFValue got updated or not? We may want to make sure that all the criteria get checked.

Michael: this function is not reliable. It needs to be more robust. How do we know whether currentOFValue got updated or not? We may want to make sure that all the criteria get checked.

Member UntangleBetaQualityMetric::evaluate_element (PatchData &pd, MsqMeshEntity *element, double &fval, MsqError &err)

This function needs to be modified so that it no longer uses compute_weighted_jacobian. It also needs to set an error whenever sent a 2D element and the surface normal information is not available.

This function needs to be modified so that it no longer uses compute_weighted_jacobian. It also needs to set an error whenever sent a 2D element and the surface normal information is not available.