# Meeting notes Till & Enrique 30.11.2022
- Davids problem: Too aggressive inlining. Solution: more conservative inlining. David will try Till's fix. This should work for Graupel. Otherwise Scalar Op.
- FVM: Runtime performance. Recomputing ($\approx$ inlining) many times, due to chaining of neighbor reductions (2 orders of magnitude). Workaround: Split into smaller operators, call them on program level and create temporaries manually. ICON: Will suffer from same problem as soon as we fuse. Same for Ninas Tracer advection (most likely). Should ideally work start of Jan, almost har deadline mid on Jan.
- FVM: Reasonable performance end of Jan (with GPU support).
- FVM: End of Jan Scan for compiled backend usable from Python.