Re: Don't treat variable-length vectors as VLAs during gimplification

[Richard Biener <richard.guenther@gmail.com>]

On Wed, Sep 18, 2019 at 8:53 AM Richard Sandiford
<richard.sandiford@arm.com> wrote:
>
> Source-level SVE vectors should be gimplified in the same way
> as normal fixed-length vectors rather than as VLAs.
>
> This is tested by later SVE patches.
>
> Tested on aarch64-linux-gnu with SVE (with and without follow-on
> patches) and x86_64-linux-gnu.  OK to install?

OK.

Richard.

> Richard
>
>
> 2019-09-18  Richard Sandiford  <richard.sandiford@arm.com>
>
> gcc/
>         * gimplify.c (gimplify_decl_expr): Use poly_int_tree_p instead
>         of checking specifically for INTEGER_CST.
>
> Index: gcc/gimplify.c
> ===================================================================
> --- gcc/gimplify.c      2019-08-08 18:11:51.411313290 +0100
> +++ gcc/gimplify.c      2019-09-18 07:52:22.799034800 +0100
> @@ -1754,11 +1754,12 @@ gimplify_decl_expr (tree *stmt_p, gimple
>        tree init = DECL_INITIAL (decl);
>        bool is_vla = false;
>
> -      if (TREE_CODE (DECL_SIZE_UNIT (decl)) != INTEGER_CST
> +      poly_uint64 size;
> +      if (!poly_int_tree_p (DECL_SIZE_UNIT (decl), &size)
>           || (!TREE_STATIC (decl)
>               && flag_stack_check == GENERIC_STACK_CHECK
> -             && compare_tree_int (DECL_SIZE_UNIT (decl),
> -                                  STACK_CHECK_MAX_VAR_SIZE) > 0))
> +             && maybe_gt (size,
> +                          (unsigned HOST_WIDE_INT) STACK_CHECK_MAX_VAR_SIZE)))
>         {
>           gimplify_vla_decl (decl, seq_p);
>           is_vla = true;