PyrGeom reference space to physical space (and inverse) mappings.
In NESO [1] we have SYCL implementations of the X maps from reference space to physical space and SYCL newton method implementations of X^-1. Our unit tests compare these mapping implementations against the Nektar++ versions to check they are consistent.
From 5.4.0 onwards we see inconsistency (seemingly only for Pyramids in the first instance) between the evaluation of X used by GetCoord in Nektar++ and the evaluation used in the Newton implementation of X^-1 in Nektar which uses the new m_isoParameter member variable. This inconsistency is order 0.01.
Here [2] should be a MFE demonstrating the issue. We see good agreement between the X implementations at the vertices but not in the interior (maybe some kind of collapsed vs non-collapsed coordinate issue?).
[1] https://github.com/ExCALIBUR-NEPTUNE/NESO [2] https://github.com/will-saunders-ukaea/pyr_mappings_mfe