> > (define_expand "umulqihi3" > > [(set (match_operand:HI 0 "register_operand") > > (mult:HI (zero_extend:HI (match_operand:QI 1 "register_operand")) > > (zero_extend:HI (match_operand:QI 2 "register_operand"))))] > > "0" > > "" > > ) > > Just delete it?
No, we actually have that insn. It's the 0 that needs to be deleted. > It sure looks like 99% of the *_real and *_virt insn patterns should > have "*" names so that they don't generate unused expanders. Turns out it was 100%. > > #define FAILED > > #define MAYBE_OK(insn) if (insn_ok_now (insn)) return; > > That non-fallback FAILED probably still ought to be gcc_unreachable. I added that.
