On Fri, Jul 26, 2024 at 2:00 PM Jeff Law <jeffreya...@gmail.com> wrote: > On 7/24/24 12:00 PM, Raphael Moreira Zinsly wrote: > ... > > diff --git a/gcc/config/riscv/riscv.md b/gcc/config/riscv/riscv.md > > index 46c46039c33..5780c5abacf 100644 > > --- a/gcc/config/riscv/riscv.md > > +++ b/gcc/config/riscv/riscv.md > > @@ -3969,7 +3969,7 @@ > > (unspec:BLK [(match_operand:X 0 "register_operand" "r") > > (match_operand:X 1 "register_operand" "r")] > > UNSPEC_TIE))] > > - "" > > + "!rtx_equal_p (operands[0], operands[1])" > > "" > > [(set_attr "type" "ghost") > > (set_attr "length" "0")] > I'm guessing this condition to to prevent subsequent passes from combing > the stack tie with something like a reg->reg copy resulting in a stack > tie with the same source/dest register?
Correct, specifically it can result in a stack tie with sp as both the src and dest registers and removing the S0 move needed in the vector probe loop. > Or are we trying to make sure the users of this insn never generate that > scenario to begin with? > > jeff > -- Raphael Moreira Zinsly
