On Mon, Nov 15, 2021 at 12:16 PM Richard Biener via Gcc-patches
<gcc-patches@gcc.gnu.org> wrote:
>
> This avoids doing aforementioned canoncalization when -ftrapping-math
> is in effect and we honor NaNs.
>
> Bootstrapped and tested on x86_64-unknown-linux-gnu.
>
> OK?
Ping.
>
> Thanks,
> Richard.
>
> 2021-11-15 Richard Biener <rguent...@suse.de>
>
> PR middle-end/103193
> * match.pd: Avoid canonicalizing (le/ge @0 @0) to (eq @0 @0)
> with NaNs and -ftrapping-math.
> ---
> gcc/match.pd | 5 ++++-
> 1 file changed, 4 insertions(+), 1 deletion(-)
>
> diff --git a/gcc/match.pd b/gcc/match.pd
> index a319aefa808..a7f1e56fe2f 100644
> --- a/gcc/match.pd
> +++ b/gcc/match.pd
> @@ -4629,7 +4629,10 @@ DEFINE_INT_AND_FLOAT_ROUND_FN (RINT)
> (if (! FLOAT_TYPE_P (TREE_TYPE (@0))
> || ! HONOR_NANS (@0))
> { constant_boolean_node (true, type); }
> - (if (cmp != EQ_EXPR)
> + (if (cmp != EQ_EXPR
> + /* With -ftrapping-math conversion to EQ loses an exception. */
> + && (! FLOAT_TYPE_P (TREE_TYPE (@0))
> + || ! flag_trapping_math))
> (eq @0 @0)))))
> (for cmp (ne gt lt)
> (simplify
> --
> 2.31.1
