Nektar issueshttps://gitlab.nektar.info/nektar/nektar/-/issues2021-07-16T13:58:40Zhttps://gitlab.nektar.info/nektar/nektar/-/issues/75Seg fault when using defintiion of quadrature points when 'c' direction is sp...2021-07-16T13:58:40ZSpencer SherwinSeg fault when using defintiion of quadrature points when 'c' direction is specified larger than 'a' and b' directionsI have noted some segmentation faults when I specify the quadrature points on prisms of the form:
<E COMPOSITE=“C[0]” BASISTYPE=“Modified_A,Modified_A,Modified_B” NUMMODES="3,3,3" POINTSTYPE="GaussLobattoLegendre,GaussLobattoLeg...I have noted some segmentation faults when I specify the quadrature points on prisms of the form:
<E COMPOSITE=“C[0]” BASISTYPE=“Modified_A,Modified_A,Modified_B” NUMMODES="3,3,3" POINTSTYPE="GaussLobattoLegendre,GaussLobattoLegendre,GaussRadauMAlpha1Beta0" NUMPOINTS="5,5,6" FIELDS="u,v,w" />
instead of
<E COMPOSITE=“C[0]" BASISTYPE="Modified_A,Modified_A,Modified_B" NUMMODES="3,3,3" POINTSTYPE="GaussLobattoLegendre,GaussLobattoLegendre,GaussRadauMAlpha1Beta0" NUMPOINTS="6,6,5" FIELDS="u,v,w" />
From the core dump it would appear to be arising the evaluation of the derivative factors within collections. I am not sure you should be able to specify the 'c' drection to be larger than the 'a' and 'b' directions but if not we should have an ASSERT in to stop it.BacklogSpencer SherwinSpencer Sherwinhttps://gitlab.nektar.info/nektar/nektar/-/issues/59Bug in GenerateElementVector2021-07-16T13:51:04ZEhsan KazemiBug in GenerateElementVectorThere is a bug in m_fields[0]->GenerateElementVector function member for parallel runs. The bug is isolated and illustrated in the enclosed solver. When we run the solver for the test case in the serial mode we get the value
area is 3.5...There is a bug in m_fields[0]->GenerateElementVector function member for parallel runs. The bug is isolated and illustrated in the enclosed solver. When we run the solver for the test case in the serial mode we get the value
area is 3.5
for area, but in the parallel run with 2 processors (np = 2) we get area is 1.5.
The solver is just ADRSolver in which m_fields[0]->GenerateElementVector is hard coded, with exactly the same test case.
[ADRSolverBug.zip](/uploads/25c9f5b5a3378acfb0389be0665d0c2b/ADRSolverBug.zip)BacklogChris CantwellChris Cantwellhttps://gitlab.nektar.info/nektar/nektar/-/issues/54Holes in isocontours when smoothing and interpolating to higher order2022-04-08T14:07:01ZAndrea CassinelliHoles in isocontours when smoothing and interpolating to higher orderAfter calculating QCriterion and smoothing with C0Projection, using the isocontour option with smoothing (smooth=1:globalcondense=1) works fine, but if one tries to interpolate to higher order the surfaces present holes.
This can be test...After calculating QCriterion and smoothing with C0Projection, using the isocontour option with smoothing (smooth=1:globalcondense=1) works fine, but if one tries to interpolate to higher order the surfaces present holes.
This can be tested with the following simple cylinder test case, where Q has already been computed and smoothed:
FieldConvert -n10 -m isocontour:fieldid=4:fieldvalue=1:smooth=1:globalcondense=1 cyl.xml q-C0.fld test.vtu
[test.vtu](/uploads/668ce59436713c436aede81fcd9df23f/test.vtu)
[q-C0.fld](/uploads/e16fc2fc24e23b6a06f6dad63448eb08/q-C0.fld)
[cyl.xml](/uploads/20588f150b208f994a85932d4d6b45c4/cyl.xml)Backloghttps://gitlab.nektar.info/nektar/nektar/-/issues/33Add support for high-order Gmsh prisms (orders 5+) and pyramids (orders 2+)2021-07-16T13:24:20ZDave MoxeyAdd support for high-order Gmsh prisms (orders 5+) and pyramids (orders 2+)After the recently merged !679 which adds support for input/output of most of the Gmsh high-order elements, we still need to add support for higher order prisms (5+) and pyramids (1+). The former can be done now with a little work to wor...After the recently merged !679 which adds support for input/output of most of the Gmsh high-order elements, we still need to add support for higher order prisms (5+) and pyramids (1+). The former can be done now with a little work to work out the ordering of interior volumes. The latter is more complex and requires the implementation of pyramids at a nodal shape level, possibly following on from !660.BacklogDave MoxeyDave Moxey