On Thu, 31 Jul 2025, Jakub Jelinek wrote:
> Hi!
>
> The following testcase is miscompiled, because byte 0x20000000
> is bit 0x100000000 and ifcombine incorrectly combines the two loads
> into a BIT_FIELD_REF even when they are very far away.
> The problem is that gimple-fold.cc ifcombine uses get_best_mode heavily,
> and that function has just int bitsize and int bitpos arguments, so
> when called e.g. with
> if (get_best_mode (end_bit - first_bit, first_bit, 0, ll_end_region,
> ll_align, BITS_PER_WORD, volatilep, &lnmode))
> where end_bit - first_bit doesn't fit into int, it is silently truncated.
> If there was just a single problematic get_best_mode call, I would probably
> just check for overflows in the caller, but there are many.
> And the two arguments are used solely as arguments to
> bit_field_mode_iterator constructor which has HOST_WIDE_INT arguments,
> so I think the easiest fix is just make the get_best_mode arguments
> also HOST_WIDE_INT.
>
> Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk/15.2?
OK.
Richard.
> 2025-07-31 Jakub Jelinek <ja...@redhat.com>
>
> PR tree-optimization/121264
> * machmode.h (get_best_mode): Change type of first 2 arguments
> from int to HOST_WIDE_INT.
> * stor-layout.cc (get_best_mode): Likewise.
>
> * gcc.dg/tree-ssa/pr121264.c: New test.
>
> --- gcc/machmode.h.jj 2025-04-08 14:08:52.826259940 +0200
> +++ gcc/machmode.h 2025-07-30 14:03:25.468655462 +0200
> @@ -958,7 +958,8 @@ private:
>
> /* Find the best mode to use to access a bit field. */
>
> -extern bool get_best_mode (int, int, poly_uint64, poly_uint64, unsigned int,
> +extern bool get_best_mode (HOST_WIDE_INT, HOST_WIDE_INT,
> + poly_uint64, poly_uint64, unsigned int,
> unsigned HOST_WIDE_INT, bool, scalar_int_mode *);
>
> /* Determine alignment, 1<=result<=BIGGEST_ALIGNMENT. */
> --- gcc/stor-layout.cc.jj 2025-04-28 11:35:40.798888512 +0200
> +++ gcc/stor-layout.cc 2025-07-30 14:04:00.148195383 +0200
> @@ -3167,7 +3167,7 @@ bit_field_mode_iterator::prefer_smaller_
> decide which of the above modes should be used. */
>
> bool
> -get_best_mode (int bitsize, int bitpos,
> +get_best_mode (HOST_WIDE_INT bitsize, HOST_WIDE_INT bitpos,
> poly_uint64 bitregion_start, poly_uint64 bitregion_end,
> unsigned int align,
> unsigned HOST_WIDE_INT largest_mode_bitsize, bool volatilep,
> --- gcc/testsuite/gcc.dg/tree-ssa/pr121264.c.jj 2025-07-30
> 14:13:14.529877964 +0200
> +++ gcc/testsuite/gcc.dg/tree-ssa/pr121264.c 2025-07-30 14:15:47.050867722
> +0200
> @@ -0,0 +1,12 @@
> +/* PR tree-optimization/121264 */
> +/* { dg-do compile } */
> +/* { dg-options "-O2 -fdump-tree-optimized" } */
> +/* { dg-final { scan-tree-dump " \\\| " "optimized" } } */
> +
> +struct A { char b; char c[0x20000010]; } a;
> +
> +int
> +foo ()
> +{
> + return a.c[0x20000000] || a.c[1];
> +}
>
> Jakub
>
>
--
Richard Biener <rguent...@suse.de>
SUSE Software Solutions Germany GmbH,
Frankenstrasse 146, 90461 Nuernberg, Germany;
GF: Ivo Totev, Andrew McDonald, Werner Knoblich; (HRB 36809, AG Nuernberg)
