[Bug tree-optimization/96692] New: Failure to optimize xor+or+xor to andnot+xor

[Bug tree-optimization/96692] New: Failure to optimize xor+or+xor to andnot+xor

[gabravier at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96692

            Bug ID: 96692
           Summary: Failure to optimize xor+or+xor to andnot+xor
           Product: gcc
           Version: 11.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: gabravier at gmail dot com
  Target Milestone: ---

int f(int a, int b, int c)
{
    return (a ^ b) ^ (a | c);
}

This can be optimized to `return (c & ~a) ^ b;`. This transformation is done by
LLVM, but not by GCC.

[Bug rtl-optimization/96692] Failure to optimize xor+or+xor to andnot+xor

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96692

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|normal                      |enhancement
          Component|tree-optimization           |rtl-optimization

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
This transformation should only happen if there is an andnot instruction.
For an example if there is no andnot instruction, there will be 3 dependent
instruction while in the first case, there are two independent instructions
followed by one dependent ones.

[Bug rtl-optimization/96692] Failure to optimize xor+or+xor to andnot+xor

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96692

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2020-08-25
     Ever confirmed|0                           |1
             Status|UNCONFIRMED                 |NEW

[Bug rtl-optimization/96692] Failure to optimize xor+or+xor to andnot+xor

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96692

--- Comment #2 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Roger Sayle <sayle@gcc.gnu.org>:

https://gcc.gnu.org/g:d458c53a6f37c8c49aa854d12e6867b4d914555f

commit r13-1507-gd458c53a6f37c8c49aa854d12e6867b4d914555f
Author: Roger Sayle <roger@nextmovesoftware.com>
Date:   Tue Jul 5 17:55:53 2022 +0100

    PR rtl-optimization/96692: ((A|B)^C)^A using andn with -mbmi on x86.

    This patch addresses PR rtl-optimization/96692 on x86_64, by providing
    a set of combine splitters to convert the three operation ((A|B)^C)^D
    into a two operation sequence using andn when either A or B is the same
    register as C or D.  This is essentially a reassociation problem that's
    only a win if the target supports an and-not instruction (as with -mbmi).

    Hence for the new test case:

    int f(int a, int b, int c)
    {
        return (a ^ b) ^ (a | c);
    }

    GCC on x86_64-pc-linux-gnu wth -O2 -mbmi would previously generate:

            xorl    %edi, %esi
            orl     %edx, %edi
            movl    %esi, %eax
            xorl    %edi, %eax
            ret

    but with this patch now generates:

            andn    %edx, %edi, %eax
            xorl    %esi, %eax
            ret

    2022-07-05  Roger Sayle  <roger@nextmovesoftware.com>
                Uroš Bizjak  <ubizjak@gmail.com>

    gcc/ChangeLog
            PR rtl-optimization/96692
            * config/i386/i386.md (define_split): Split ((A | B) ^ C) ^ D
            as (X & ~Y) ^ Z on target BMI when either C or D is A or B.

    gcc/testsuite/ChangeLog
            PR rtl-optimization/96692
            * gcc.target/i386/bmi-andn-4.c: New test case.

[Bug rtl-optimization/96692] Failure to optimize xor+or+xor to andnot+xor

[gabravier at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96692

--- Comment #3 from Gabriel Ravier <gabravier at gmail dot com> ---
This seems to be fixed on trunk now, I think ?

[Bug rtl-optimization/96692] Failure to optimize xor+or+xor to andnot+xor

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96692

--- Comment #4 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Gabriel Ravier from comment #3)
> This seems to be fixed on trunk now, I think ?

On x86_64-linux-gnu yes but on aarch64 it is not optimized just yet:
f(int, int, int):
        eor     w1, w0, w1
        orr     w0, w0, w2
        eor     w0, w1, w0
        ret
f1(int, int, int):
        bic     w0, w2, w0
        eor     w0, w0, w1
        ret