Closed
Description
They're all run in parallel now, and they're incredibly slow because there is a massive amount of contention inside LLVM.
I have a patch to fix this and am attempting to get it into LLVM upstream. In the meantime, if we upgrade LLVM we should include this patch.