Is there anyone on the compiler team that also plays a role in llvm development and optimizations? Or is there anyone on the llvm core team that cares about optimizations that would directly benefit Swift? If there's no one available that either can or is willing to optimize llvm, this feels like a worst case scenario. I'm curious what steps need to be taken to see actual compiler performance gains as it relates to llvm.

Interesting that the two highest functions in that trace relate to Bitcode serialization and GlobalISel. Why is Swift serializing Bitcode if it is also running ISEL in the same compile?