https://gcc.gnu.org/bugzilla/show_bug.cgi?id=108248
--- Comment #4 from Jeffrey A. Law <law at gcc dot gnu.org> --- Yea, thinking about our uarch, we're going to need finer control as well. There's a subset that ought to execute in any ALU, but there's another subset that are bound to a specific ALU and are potentially multi-cycle latency.