Latest loopy produces many warnings

[connorjward]

Since updating our loopy fork we are seeing a lot of warnings like:

  /home/connor/Code/firedrake-dev1/src/loopy/loopy/schedule/__init__.py:2255: LoopyWarning: in ke
rnel wrap_form0_cell_integral: Falling back to a slow scheduler implementation due to: cannot sch
edule kernels with priority dependencies between sibling loop nests (add 'v1_scheduler_fallback' 
to silenced_warnings kernel attribute to disable)

and

tests/slate/test_slate_mixed_direct.py::test_slate_mixed_matrix[nest]
tests/slate/test_slate_mixed_direct.py::test_slate_mixed_matrix[aij]
  /home/connor/Code/firedrake-dev1/src/loopy/loopy/schedule/tools.py:839: UserWarning: Cannot enf
orce the constraint: i_0 to be nested within i, as i_0 is either a parallel loop or not an iname.
    cannot_satisfy_callback(f"Cannot enforce the constraint:"

Do you have any advice as to how we should go about resolving these? I assume there must be some new metadata associated with domains, or assignments or similar that we are not specifying?

Thanks in advance.

[inducer]

Converting to discussion since IIUC everything is operating as intended.

The warning concerns a fast path for loopy's "scheduler" (loop-nest-order-er). This was added in #350, and the fast path has asymptotically lower cost than the old version. The warning says that the fast path can't be used because there is a loop nesting constraint (via prioritize_loops) between sibling (i.e. non-nested) loops. This is not a sensible request; hence the fast path refuses to proceed. Removing the offending nesting constraints might make the fast path usable (and may even lead to improved code gen times). HTH.

[connorjward]

Yes this is definitely a discussion not an issue. Thank you for reassigning it.

This is useful information thanks. I will attempt to address the warnings and report back.

[connorjward]

These warnings seem to come from some unnecessary calls to prioritize_loops as suggested (here and here).

Thank you for your help!