On 01/06/2018 11:58 PM, Prathamesh Kulkarni wrote:
[ Snip ]
>> I think with those changes we're probably in good shape. But please
>> repost for final approval.
> I have the updated the attached version with your suggestions.
> Does it look OK ?
> Bootstrap+test passes on x86_64-unknown-linux-gnu.
>
> Thanks,
> Prathamesh
>> jeff
>
> pr82665-9.diff
>
>
> diff --git a/gcc/vr-values.c b/gcc/vr-values.c
> index 794b4635f9e..41a4a0b041f 100644
> --- a/gcc/vr-values.c
> +++ b/gcc/vr-values.c
> @@ -793,6 +793,39 @@ vr_values::extract_range_from_binary_expr (value_range
> *vr,
>
> extract_range_from_binary_expr_1 (vr, code, expr_type, &vr0, &vr1);
>
> + /* Set value_range for n in following sequence:
> + def = __builtin_memchr (arg, 0, sz)
> + n = def - arg
> + Here the range for n can be set to [0, PTRDIFF_MAX - 1]. */
> +
> + if (vr->type == VR_VARYING
> + && code == POINTER_DIFF_EXPR
> + && TREE_CODE (op0) == SSA_NAME
> + && TREE_CODE (op1) == SSA_NAME)
> + {
> + tree op0_ptype = TREE_TYPE (TREE_TYPE (op0));
> + tree op1_ptype = TREE_TYPE (TREE_TYPE (op1));
> + gcall *call_stmt = NULL;
> +
> + if (TYPE_MODE (op0_ptype) == TYPE_MODE (char_type_node)
> + && TYPE_PRECISION (op0_ptype) == TYPE_PRECISION (char_type_node)
> + && TYPE_MODE (op1_ptype) == TYPE_MODE (char_type_node)
> + && TYPE_PRECISION (op1_ptype) == TYPE_PRECISION (char_type_node)
Note that while the operands of POINTER_DIFF_EXPR can be pointers to
different types, we do require that they have the same mode and
precision. So the tests of TYPE_MODE and TYPE_PRECISION for op1_ptype
are not needed.
OK with those two checks removed.
Jeff
ps. FWIW, the close a stage in our development cycle is a patch
submission deadline. So if a patch is submitted prior to the deadline,
then it can move forward, even if review/approval happens after the
deadline.
