iirc, we had expected a runtime speedup w/ inlining too - so might still be some opportunities there to optimize