Extra Methods for 2nd Derivative Penalties

[JaredCrean2]

There are a couple of functions that users of SummationByParts could use to make the viscous terms faster:

• computing element-level operations for only the nodes that are in the stencil of the interpolation operator
  • Ex. differentiateElementR that computes only the elements of uprime = Dx * u that will be used for volume to face interpolation
• A uniform way to get both the stencil size of the interpolation operator as well as the nodes involved (sorted in increasing order)
  • I think sbpface.perm might do this, but I'm not sure if that is correct for SparseFace. I'm also not sure if it is sorted.

I don't need these functions in the short term, this issue is for future reference.

[JaredCrean2]

Another useful feature would be to combine multiplication by a diagonal matrix and interpolation. Currently, the operation t2 = [Nx*R, Ny*R, Nz*R]*t1 has to be done in 2 steps: tmp = [Nx*R*t1, Ny*R*t1, Nz*R*t1]; t2 = R*tmp. Combining these into a single operation would remove the need for a temporary array tmp, which could be significant for the jacobian calculation, where tmp may be somewhat large.

[JaredCrean2]

It would also be helpful to have methods to zero out only the relevant entries of the 4D and 5D jacobians (for the face interpolation functions).