[Bug target/126589] [13/14/15/16 Regression] Complex-mul vectorisation not honoring signed zeros

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126589

--- Comment #5 from GCC Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Tamar Christina <[email protected]>:

https://gcc.gnu.org/g:d9b294889b14069d6f6c5ce985f17bf8bbf34894

commit r17-3228-gd9b294889b14069d6f6c5ce985f17bf8bbf34894
Author: Tamar Christina <[email protected]>
Date:   Wed Aug 12 14:15:43 2026 +0100

    vect: Support another form of COMPLEX_FMS which the testcases point out
[PR126589]

    This adds support for matching FMS where the reassoc has left the top level
    operation as a PLUS_MINUS.

    i.e. matching

           c.real + (a.imag * b.imag - a.real * b.real)
           c.imag - (a.real * b.imag + a.imag * b.real)

    which happens when the accumulator is not the same as the destination.
    i.e.

    d[i] = c[i] - (a[i] * b[i])

    vs

    c[i] -= a[i] * b[i]

    gcc/ChangeLog:

            PR tree-optimization/126589
            * tree-vect-slp-patterns.cc (complex_fms_pattern::matches): Check
            another form of FMS.

    gcc/testsuite/ChangeLog:

            PR tree-optimization/126589
            * gfortran.target/aarch64/pr122408_1.f90: update scans to check for
FMS.
            * gcc.dg/vect/complex/complex-mul2.c: New test.
            * gcc.dg/vect/complex/fast-math-complex-mls-manual-invalid-float.c:
New test.
            * gcc.dg/vect/complex/fast-math-complex-mls-outofplace-double.c:
New test.
            * gcc.dg/vect/complex/fast-math-complex-mls-outofplace-float.c: New
test.