On Fri, 11 Dec 2020, Jakub Jelinek wrote:
> Hi!
>
> As the testcase shows, for 32-bit word size we can end up with op1
> up to 0xffffffff (0x100000000 % 0xffffffff == 1 and so we use bit == 32
> for that), but the CONST_INT we got from caller is for DImode in that case
> and not valid for SImode operations.
>
> The following patch canonicalizes the two spots where the constant needs
> canonicalization.
>
> Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
Shouldn't we simply use gen_int_mode?
> 2020-12-11 Jakub Jelinek <ja...@redhat.com>
>
> PR rtl-optimization/98229
> * optabs.c (expand_doubleword_mod): Canonicalize op1 and
> 1 - INTVAL (op1) as word_mode constants when used in
> word_mode arithmetics.
>
> * gcc.c-torture/compile/pr98229.c: New test.
>
> --- gcc/optabs.c.jj 2020-12-02 11:40:47.574338088 +0100
> +++ gcc/optabs.c 2020-12-10 19:00:46.151434474 +0100
> @@ -1081,7 +1081,8 @@ expand_doubleword_mod (machine_mode mode
> return NULL_RTX;
> }
> }
> - rtx remainder = expand_divmod (1, TRUNC_MOD_EXPR, word_mode, sum, op1,
> + rtx op1w = GEN_INT (trunc_int_for_mode (INTVAL (op1), word_mode));
> + rtx remainder = expand_divmod (1, TRUNC_MOD_EXPR, word_mode, sum, op1w,
> NULL_RTX, 1, OPTAB_DIRECT);
> if (remainder == NULL_RTX)
> return NULL_RTX;
> @@ -1098,8 +1099,9 @@ expand_doubleword_mod (machine_mode mode
> if (mask == NULL_RTX)
> return NULL_RTX;
> }
> - mask = expand_simple_binop (word_mode, AND, mask,
> - GEN_INT (1 - INTVAL (op1)),
> + rtx oneminusop1
> + = GEN_INT (trunc_int_for_mode (1 - INTVAL (op1), word_mode));
> + mask = expand_simple_binop (word_mode, AND, mask, oneminusop1,
> NULL_RTX, 1, OPTAB_DIRECT);
> if (mask == NULL_RTX)
> return NULL_RTX;
> --- gcc/testsuite/gcc.c-torture/compile/pr98229.c.jj 2020-12-10
> 19:08:25.502328354 +0100
> +++ gcc/testsuite/gcc.c-torture/compile/pr98229.c 2020-12-10
> 19:08:10.062499943 +0100
> @@ -0,0 +1,7 @@
> +/* PR rtl-optimization/98229 */
> +
> +unsigned long long
> +foo (unsigned long long x)
> +{
> + return x % ~0U;
> +}
>
> Jakub
>
>
--
Richard Biener <rguent...@suse.de>
SUSE Software Solutions Germany GmbH, Maxfeldstrasse 5, 90409 Nuernberg,
Germany; GF: Felix Imendörffer; HRB 36809 (AG Nuernberg)
