On Tue, Sep 03, 2024 at 10:42:34AM +0200, Richard Biener wrote:
> The following moves the assert on NUM_POLY_INT_COEFFS != 1 after
> INTEGER_CST processing.
>
> Bootstrap and regtest running on x86_64-unknown-linux-gnu, pushed
> as obvious after getting into stage3.
>
> * fold-const.cc (poly_int_binop): Move assert on
> NUM_POLY_INT_COEFFS after INTEGER_CST processing.
> ---
> gcc/fold-const.cc | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/gcc/fold-const.cc b/gcc/fold-const.cc
> index 2ada59f712b..70db16759d0 100644
> --- a/gcc/fold-const.cc
> +++ b/gcc/fold-const.cc
> @@ -1241,7 +1241,6 @@ poly_int_binop (poly_wide_int &res, enum tree_code code,
> const_tree arg1, const_tree arg2,
> signop sign, wi::overflow_type *overflow)
> {
> - gcc_assert (NUM_POLY_INT_COEFFS != 1);
> gcc_assert (poly_int_tree_p (arg1) && poly_int_tree_p (arg2));
Even the other assert is just a waste of time in the
if (TREE_CODE (arg1) == INTEGER_CST && TREE_CODE (arg2) == INTEGER_CST)
case.
> if (TREE_CODE (arg1) == INTEGER_CST && TREE_CODE (arg2) == INTEGER_CST)
> @@ -1254,6 +1253,8 @@ poly_int_binop (poly_wide_int &res, enum tree_code code,
> return true;
> }
>
> + gcc_assert (NUM_POLY_INT_COEFFS != 1);
> +
> switch (code)
> {
> case PLUS_EXPR:
> --
> 2.43.0
Jakub
