perlfu wrote: > we probably could run branch relaxation before the post RA scheduler Presumably this would cause problems if new hazard related instructions pushed a branch over the representable size? https://github.com/llvm/llvm-project/pull/109127