On Tue, Mar 28, 2023 at 3:11 PM Alexander Monakov via Gcc-patches
<gcc-patches@gcc.gnu.org> wrote:
>
> Do not attempt to use a plain subtraction for generating a three-way
> comparison result in autopref_rank_for_schedule qsort comparator, as
> offsets are not restricted and subtraction may overflow.  Open-code
> a safe three-way comparison instead.

OK.

> gcc/ChangeLog:
>
>         PR rtl-optimization/109187
>         * haifa-sched.cc (autopref_rank_for_schedule): Avoid use of 
> overflowing
>         subtraction in three-way comparison.
>
> gcc/testsuite/ChangeLog:
>
>         PR rtl-optimization/109187
>         * gcc.dg/pr109187.c: New test.
> ---
>
> I think I can commit this as obvious if no comment in a day, but explicit ack
> is always appreciated.
>
> Alexander
>
>  gcc/haifa-sched.cc              | 2 +-
>  gcc/testsuite/gcc.dg/pr109187.c | 8 ++++++++
>  2 files changed, 9 insertions(+), 1 deletion(-)
>  create mode 100644 gcc/testsuite/gcc.dg/pr109187.c
>
> diff --git a/gcc/haifa-sched.cc b/gcc/haifa-sched.cc
> index 4efaa9445..e11cc5c35 100644
> --- a/gcc/haifa-sched.cc
> +++ b/gcc/haifa-sched.cc
> @@ -5686,7 +5686,7 @@ autopref_rank_for_schedule (const rtx_insn *insn1, 
> const rtx_insn *insn2)
>
>        if (!irrel1 && !irrel2)
>         /* Sort memory references from lowest offset to the largest.  */
> -       r = data1->offset - data2->offset;
> +       r = (data1->offset > data2->offset) - (data1->offset < data2->offset);
>        else if (write)
>         /* Schedule "irrelevant" insns before memory stores to resolve
>            as many producer dependencies of stores as possible.  */
> diff --git a/gcc/testsuite/gcc.dg/pr109187.c b/gcc/testsuite/gcc.dg/pr109187.c
> new file mode 100644
> index 000000000..1ef14a73d
> --- /dev/null
> +++ b/gcc/testsuite/gcc.dg/pr109187.c
> @@ -0,0 +1,8 @@
> +/* { dg-do compile } */
> +/* { dg-options "-O2 --param sched-autopref-queue-depth=1" } */
> +
> +void f(int *a)
> +{
> +  for (;;)
> +    asm("" :: "r"(a[-0x10000000]), "r"(a[0x10000000]), "r"(a[0]) : "memory");
> +}
> --
> 2.39.1
>

Reply via email to