On Thu, 17 Jul 2025, Jakub Jelinek wrote:
> Hi!
>
> The following testcase ICEs because SCALAR_INT_TYPE_MODE of course
> doesn't work for large BITINT_TYPE types which have BLKmode.
> native_encode* as well as e.g. r14-8276 use in cases like these
> GET_MODE_SIZE (SCALAR_INT_TYPE_MODE ()) and TREE_INT_CST_LOW (TYPE_SIZE_UNIT
> ()) for the BLKmode ones.
> In this case, it wants bits rather than bytes, so I've used
> GET_MODE_BITSIZE like before and TYPE_SIZE otherwise.
>
> Furthermore, the patch only computes encoding_size for big endian
> targets, for little endian we don't really adjust anything, so there
> is no point computing it.
>
> Bootstrapped/regtested on x86_64-linux and i686-linux, ok for
> trunk/15.2/14.4?
OK.
Thanks,
Richard.
> 2025-07-17 Jakub Jelinek <ja...@redhat.com>
>
> PR tree-optimization/121131
> * gimple-fold.cc (fold_nonarray_ctor_reference): Use
> TREE_INT_CST_LOW (TYPE_SIZE ()) instead of
> GET_MODE_BITSIZE (SCALAR_INT_TYPE_MODE ()) for BLKmode BITINT_TYPEs.
> Don't compute encoding_size at all for little endian targets.
>
> * gcc.dg/bitint-124.c: New test.
>
> --- gcc/gimple-fold.cc.jj 2025-06-30 10:59:37.553460624 +0200
> +++ gcc/gimple-fold.cc 2025-07-17 16:09:34.993333798 +0200
> @@ -9916,10 +9916,17 @@ fold_nonarray_ctor_reference (tree type,
> {
> if (BYTES_BIG_ENDIAN != WORDS_BIG_ENDIAN)
> return NULL_TREE;
> - const unsigned int encoding_size
> - = GET_MODE_BITSIZE (SCALAR_INT_TYPE_MODE (TREE_TYPE (cfield)));
> if (BYTES_BIG_ENDIAN)
> - inner_offset += encoding_size - wi::to_offset (field_size);
> + {
> + tree ctype = TREE_TYPE (cfield);
> + unsigned int encoding_size;
> + if (TYPE_MODE (ctype) != BLKmode)
> + encoding_size
> + = GET_MODE_BITSIZE (SCALAR_INT_TYPE_MODE (ctype));
> + else
> + encoding_size = TREE_INT_CST_LOW (TYPE_SIZE (ctype));
> + inner_offset += encoding_size - wi::to_offset (field_size);
> + }
> }
>
> return fold_ctor_reference (type, cval,
> --- gcc/testsuite/gcc.dg/bitint-124.c.jj 2025-07-17 16:17:52.412798465
> +0200
> +++ gcc/testsuite/gcc.dg/bitint-124.c 2025-07-17 16:21:03.803284269 +0200
> @@ -0,0 +1,30 @@
> +/* PR tree-optimization/121131 */
> +/* { dg-do run { target bitint } } */
> +/* { dg-options "-O2" } */
> +
> +#if __BITINT_MAXWIDTH__ >= 156
> +struct A { _BitInt(156) b : 135; };
> +
> +static inline _BitInt(156)
> +foo (struct A *x)
> +{
> + return x[1].b;
> +}
> +
> +__attribute__((noipa)) _BitInt(156)
> +bar (void)
> +{
> + struct A a[] = { 1, 1, -13055525270329736316393717310914023773847wb,
> + 1, 1, 1, 1, 1, 1, 1, 1, 1 };
> + return foo (&a[1]);
> +}
> +#endif
> +
> +int
> +main ()
> +{
> +#if __BITINT_MAXWIDTH__ >= 156
> + if (bar () != -13055525270329736316393717310914023773847wb)
> + __builtin_abort ();
> +#endif
> +}
>
> 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)
