search

inducer / modepy

Modes and nodes for high-order discretizations
Other
16 stars 7 forks source link

`submesh_for_shape` should support generic tensor products #49

Closed inducer closed 2 years ago

inducer commented 2 years ago

(This'll be needed/become possible after #41 merges.)

cc @alexfikl

alexfikl commented 2 years ago

Does this require any special support? I was hoping it would just work™ based on the node tuples it gets (should test that!).

inducer commented 2 years ago

I assumed it didn't since submesh_for_shape was specialized for Hypercube. Maybe just broadening it to TensorProductShape suffices?

inducer commented 2 years ago

Closing here, adding to my review.

alexfikl commented 2 years ago

Actually, just looked over it a bit and it should fail for line x triangle because of 

    vertex_node_tuples = [nt[::-1] for nt in gnitb(2, dims)]

which assumes hypercubes. Should be possible to fix that by taking the vertex_node_tuples to be the first order node tuples for that tensor product shape.