On Wed, Sep 20, 2017 at 2:06 PM, Richard Sandiford
<richard.sandif...@linaro.org> wrote:
> extract_range_from_binary_expr_1 had:
>
> if (range_int_cst_p (&vr0)
> && range_int_cst_p (&vr1)
> && TYPE_OVERFLOW_WRAPS (expr_type))
> ...
> ...
> extract_range_from_multiplicative_op_1 (vr, code, &vr0, &vr1);
>
> but extract_range_from_multiplicative_op_1 also requires range_int_cst_p.
> I think we should bail out if either range isn't a constant.
>
> This might only be theoretical with current sources, but it's needed
> once polynomial constants are added.
>
> Tested on aarch64-linux-gnu, x86_64-linux-gnu and powerpc64le-linus-gnu.
> OK to install?
extract_range_from_multiplicative_op_1 also allows anti-ranges but indeed
requires INTEGER_CSTs. Note it won't get anti-ranges because we handle
this case by ranges_from_anti_range.
Thus ok if you also remove the apparent anti-range handling from
extract_range_from_multiplicative_op_1 (in the assert).
Thanks,
Richard.
> Richard
>
>
> 2017-09-20 Richard Sandiford <richard.sandif...@linaro.org>
> Alan Hayward <alan.hayw...@arm.com>
> David Sherwood <david.sherw...@arm.com>
>
> gcc/
> * tree-vrp.c (extract_range_from_binary_expr_1): Don't call
> extract_range_from_multiplicative_op_1 if !range_int_cst_p.
>
> Index: gcc/tree-vrp.c
> ===================================================================
> --- gcc/tree-vrp.c 2017-08-30 12:18:46.631541422 +0100
> +++ gcc/tree-vrp.c 2017-09-20 13:04:20.397027078 +0100
> @@ -2462,9 +2462,14 @@ extract_range_from_binary_expr_1 (value_
> signop sign = TYPE_SIGN (expr_type);
> unsigned int prec = TYPE_PRECISION (expr_type);
>
> - if (range_int_cst_p (&vr0)
> - && range_int_cst_p (&vr1)
> - && TYPE_OVERFLOW_WRAPS (expr_type))
> + if (!range_int_cst_p (&vr0)
> + || !range_int_cst_p (&vr1))
> + {
> + set_value_range_to_varying (vr);
> + return;
> + }
> +
> + if (TYPE_OVERFLOW_WRAPS (expr_type))
> {
> typedef FIXED_WIDE_INT (WIDE_INT_MAX_PRECISION * 2) vrp_int;
> typedef generic_wide_int
