On Sun, Sep 14, 2025 at 8:17 PM Andrew Pinski
<andrew.pin...@oss.qualcomm.com> wrote:
>
> This pattern shows up with some C++ code (std::vector<bool>) where we get:
> ```
> _9 = _201 - _36;
> _10 = (long unsigned int) _9;
> _11 = -_10;
> _12 = _201 + _11;
> ```
>
> In the original code it was `end - (end - begin)` but with inlined functions
> so it
> is not direct fixable.
>
> This patch adds a new amtch pattern like what was added in
> r0-128019-g77574c353464b3 (and then moved to match with
> r5-4705-ga499aac5dfa5d9).
>
> Bootstrapped and tested on x86_64-linux-gnu.
I wonder, if in general, -(ptr1 - ptr2) can be changed to (ptr2 -
ptr1). I think
this should be OK? Of course with the unsigned conversion in the way
it's (long unsigned int)(ptr2 - ptr2)?
Thanks,
Richard.
> PR tree-optimization/121921
> gcc/ChangeLog:
>
> * match.pd (`eptr - (eptr - bptr)`): New patterns.
>
> gcc/testsuite/ChangeLog:
> * gcc.dg/pr121921-1.c: New testcase.
>
> Signed-off-by: Andrew Pinski <andrew.pin...@oss.qualcomm.com>
> ---
> gcc/match.pd | 24 ++++++++++++++++++++++++
> gcc/testsuite/gcc.dg/pr121921-1.c | 14 ++++++++++++++
> 2 files changed, 38 insertions(+)
> create mode 100644 gcc/testsuite/gcc.dg/pr121921-1.c
>
> diff --git a/gcc/match.pd b/gcc/match.pd
> index dbbb7b51275..5dea74aabe2 100644
> --- a/gcc/match.pd
> +++ b/gcc/match.pd
> @@ -3047,6 +3047,30 @@ DEFINE_INT_AND_FLOAT_ROUND_FN (RINT)
> (if (TYPE_PRECISION (TREE_TYPE (@2)) >= TYPE_PRECISION (TREE_TYPE (@3)))
> (convert @1)))
>
> +/* Likewise but with an additional negate here
> + to swap the order of the subtraction. */
> +/* Pattern match
> + tem1 = (long) ptr1;
> + tem2 = (long) ptr2;
> + tem3 = tem1 - tem2;
> + tem4 = (unsigned long) tem3;
> + tem5 = -tem3;
> + tem5 = ptr1 + tem5;
> + and produce
> + tem5 = ptr2; */
> +(simplify
> + (pointer_plus @0 (negate (convert?@2 (minus@3 (convert @0) (convert @1)))))
> + /* Conditionally look through a sign-changing conversion. */
> + (if (TYPE_PRECISION (TREE_TYPE (@2)) == TYPE_PRECISION (TREE_TYPE (@3))
> + && ((GIMPLE && useless_type_conversion_p (type, TREE_TYPE (@1)))
> + || (GENERIC && type == TREE_TYPE (@1))))
> + @1))
> +/* ptr0 - (ptr0 - ptr1) -> ptr1 */
> +(simplify
> + (pointer_plus @0 (negate (convert?@2 (pointer_diff@3 @@0 @1))))
> + (if (TYPE_PRECISION (TREE_TYPE (@2)) >= TYPE_PRECISION (TREE_TYPE (@3)))
> + (convert @1)))
> +
> /* Pattern match
> tem = (sizetype) ptr;
> tem = tem & algn;
> diff --git a/gcc/testsuite/gcc.dg/pr121921-1.c
> b/gcc/testsuite/gcc.dg/pr121921-1.c
> new file mode 100644
> index 00000000000..b11afc819da
> --- /dev/null
> +++ b/gcc/testsuite/gcc.dg/pr121921-1.c
> @@ -0,0 +1,14 @@
> +/* { dg-do compile } */
> +/* { dg-options "-O -fdump-tree-cddce1" } */
> +
> +/* PR tree-optimization/121921 */
> +
> +int *
> +fx (int *b, int *e)
> +{
> + __SIZE_TYPE__ p = b - e;
> + /* The first forwprop pass should optimize this to return e; */
> + return b - p;
> +}
> +
> +/* { dg-final { scan-tree-dump "return e" "cddce1" } } */
> --
> 2.43.0
>
