On Fri, 7 Nov 2025, Avinash Jayakar wrote:
> Hi,
>
> I have incorporated the changes suggested in version 1 and added 2 tests for
> vec_set and vec_extract, I am not sure if a test for vec_extract is needed.
> Bootstrapped and regtested on powerpc64le and x86 with no regressions.
> Ok for trunk? Should this go in 13/14/15 as well?
>
> Change from v1:
> - Added tests.
> - Use poly comparison instead of convert and test.
>
> Thanks and regards,
> Avinash Jayakar
>
> The function gimple_expand_vec_set_expr in the isel pass, converted
> VIEW_CONVERT_EXPR to VEC_SET_EXPR without checking the bounds on the index,
> which cause ICE on targets that supported VEC_SET_EXPR like x86 and powerpc.
> This patch adds a bound check on the index operand and rejects the conversion
> if index is out of bound.
>
> 2025-11-06 Avinash Jayakar <[email protected]>
>
> gcc/ChangeLog:
> PR tree-optimization/122126
> * gimple-isel.cc (gimple_expand_vec_set_extract_expr): Add bound
> check.
>
> gcc/testsuite/ChangeLog:
> PR tree-optimization/122126
> * gcc.dg/pr122126_vextr.c: New test.
> * gcc.dg/pr122126_vset.c: New test.
> ---
> gcc/gimple-isel.cc | 10 ++++++++++
> gcc/testsuite/gcc.dg/pr122126_vextr.c | 9 +++++++++
> gcc/testsuite/gcc.dg/pr122126_vset.c | 9 +++++++++
> 3 files changed, 28 insertions(+)
> create mode 100644 gcc/testsuite/gcc.dg/pr122126_vextr.c
> create mode 100644 gcc/testsuite/gcc.dg/pr122126_vset.c
>
> diff --git a/gcc/gimple-isel.cc b/gcc/gimple-isel.cc
> index b5dc579ff46..e745608904e 100644
> --- a/gcc/gimple-isel.cc
> +++ b/gcc/gimple-isel.cc
> @@ -102,6 +102,16 @@ gimple_expand_vec_set_extract_expr (struct function *fun,
> tree pos = TREE_OPERAND (ref, 1);
>
> tree view_op0 = TREE_OPERAND (op0, 0);
> +
> + tree idx = TREE_OPERAND (ref, 1);
> + // if index is a constant, then check the bounds
> + poly_uint64 idx_poly;
> + if (poly_int_tree_p (idx, &idx_poly))
> + {
> + poly_uint64 nelts = TYPE_VECTOR_SUBPARTS (TREE_TYPE (view_op0));
> + if (known_gt (idx_poly, nelts))
> + return false;
> + }
> machine_mode outermode = TYPE_MODE (TREE_TYPE (view_op0));
> machine_mode extract_mode = TYPE_MODE (TREE_TYPE (ref));
>
> diff --git a/gcc/testsuite/gcc.dg/pr122126_vextr.c
> b/gcc/testsuite/gcc.dg/pr122126_vextr.c
> new file mode 100644
> index 00000000000..d25555b0aec
> --- /dev/null
> +++ b/gcc/testsuite/gcc.dg/pr122126_vextr.c
> @@ -0,0 +1,9 @@
> +/* { dg-do compile } */
> +/* { dg-options "-O2" } */
> +/* { dg-additional-options "-mavx2" { target x86_64-*-* i?86-*-* } } */
Please use { target avx2 } instead in both testcases.
OK with those changes.
Richard.
> +
> +#define vect16 __attribute__((vector_size(16)))
> +void ub_set() {
> + volatile vect16 unsigned BS_VAR_0;
> + unsigned a = BS_VAR_0[12];
> +}
> \ No newline at end of file
> diff --git a/gcc/testsuite/gcc.dg/pr122126_vset.c
> b/gcc/testsuite/gcc.dg/pr122126_vset.c
> new file mode 100644
> index 00000000000..a398b484a36
> --- /dev/null
> +++ b/gcc/testsuite/gcc.dg/pr122126_vset.c
> @@ -0,0 +1,9 @@
> +/* { dg-do compile } */
> +/* { dg-options "-O2" } */
> +/* { dg-additional-options "-mavx2" { target x86_64-*-* i?86-*-* } } */
> +
> +#define vect16 __attribute__((vector_size(16)))
> +void ub_set() {
> + volatile vect16 unsigned BS_VAR_0;
> + BS_VAR_0[12] = 4;
> +}
> \ No newline at end of file
>
--
Richard Biener <[email protected]>
SUSE Software Solutions Germany GmbH,
Frankenstrasse 146, 90461 Nuernberg, Germany;
GF: Jochen Jaser, Andrew McDonald, Werner Knoblich; (HRB 36809, AG Nuernberg)
