Mesh-based Source Routines

[paulromano]

No description provided.

[paulromano]

Partially completed with openmc-dev/openmc#2334. Remaining to-dos on this include:

• Add support for sampling from other mesh types (Regular, Cylindrical, etc.) (Mesh Source Class openmc#2759)
• Add a MeshSource class that will handle user-defined angle/energy/time distributions that can be set either for the whole mesh or element-by-element (Mesh Source Class openmc#2759)

[paulromano]

Discussed this yesterday over a call. @pshriwise has a branch in the works extending the MeshSpatial class to cover all mesh types. One additional need that we discussed is for the case where we want to independent activate multiple materials that occur in the same mesh element. In that case, we would have a separate distribution for each material and would need to have rejection of each distribution within the mesh element. It's unclear how we would represent this from an API standpoint or how it would be implemented.

[eepeterson]

Do we need to add source biasing to the todo list on this as we mentioned it in the proposal. Source biasing beyond just setting relative strengths?

[pshriwise]

• Refactor source distributions to use distributions in global vectors so they can be shared across IndependentSource's.

Update: I've created a separate topic to track this in the project

[pshriwise]

Add a MeshSource class that will handle user-defined angle/energy/time distributions that can be set either for the whole mesh or element-by-element

@paulromano @eepeterson given our change in sampling order (underlying source first, then replace the spatial component with the sampled position in an element), do we still want to allow for global distributions across a mesh? The more global distributions we provide, the less it seems to make sense to have each element contain sources with their own distributions...

[paulromano]

With openmc-dev/openmc#2980, creating a mesh decay photon source for R2S is now agnostic to the underlying type of mesh. The last piece we need is ability to reject sampled positions in void, which will be enabled by openmc-dev/openmc#2916 after some refactoring.