-
Kevin Modzelewski authored
Mostly related to making sure that some hot paths get inlined better, particularly for allocation/sweeping. These changes make a simple for-loop benchmark (which is just dependent on allocation speed) about 25% faster.
cf19a38c