https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126914
--- Comment #8 from Jeffrey A. Law <law at gcc dot gnu.org> --- Yea. Interestingly enough I'm seeing a regression with the patches on RISC-V as well. In the RISC-V case the costing looks reasonable, but we get a regression because old, more expensive, sequence would combine with other nearby insns while the new, cheaper sequence will not easily combine. Pondering...
