Matt Smith

@inducer Do you have a sense of how well-behaved the basis functions are when evaluated outside of the unit element coordinate range? I had suggested to @dshtey2 first using the...

> While I don't think that in general there's much of an expectation that evaluating functions outside of the reference element would yield meaningful results, in your case (given purely...

> Huh, interesting. That should be fixable. Could you put code to reproduce this somewhere? (Maybe a temporary PR) The code is in mirgecom at the moment, in @dshtey2's branch....

> Any idea of how much faster this is compared to #350? On my laptop, for a mesh with 1.2M elements: `main`: ``` 53.433 _get_box_mesh poiseuille-mpi.py:74 └─ 53.420 inner_wrapper pytools/__init__.py:243...

Still experimenting with this, but so far I'm getting the sense that it's something other than the indirection that's causing `resample_by_picking` to be slow. First I tried pre-sorting by `batch.to_element_indices`,...

> It's OK that that doesn't appear in the kernel. It gets used in the "wrapper" code that computes the OpenCL "grid" bounds. Right now, the (simple) work decomposition is...

FYI @kaushikcfd, while I was browsing through this code the other day trying to understand a warning that was being emitted (which turned into inducer/meshmode#453), I spotted a few opportunities...

Just linking my draft of this here for future reference (to be added together with grudge DOF array stuff, when that is available): [meshmode@production-dof-tagging](https://github.com/majosm/meshmode/tree/production-dof-tagging) [grudge@production-dof-tagging](https://github.com/majosm/grudge/tree/production-dof-tagging) [mirgecom@production-dof-tagging](https://github.com/majosm/mirgecom/tree/production-dof-tagging)