[Bug rtl-optimization/103006] [9/10/11/12 Regression] wrong code at -O1 or -O2 on x86_64-linux-gnu

[pinskia at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103006

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|middle-end                  |rtl-optimization
           Keywords|                            |ra

--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
This looks like a Register allocator issue.
IN GCC 6 we have:
.L8:
        movq    %rsp, b(%rip)
        movq    %rsp, 48(%rsp)
        call    _Z2ffv
        movq    112(%rsp), %rax
        addq    $160, %rsp
        .cfi_def_cfa_offset 8
        ret

While in GCC 7+ we get:
        movl    $2, f(%rip)
        movl    $6, e(%rip)
        movq    %rsp, b(%rip)
        movq    %rsp, 48(%rsp)
        call    _Z2ffv
        movl    48(%rsp), %eax
        addq    $96, %rsp
        .cfi_def_cfa_offset 8
        ret