* Use FxHashMap instead of HashMap for better performance * Replace the binary search in the coalescing pass with a FxHashMap This speeds up coalescing by up to 16% and overall compilation by 9%
This fixes the next part of #287.