From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gcc-bugzilla@gcc.gnu.org>
Received: by sourceware.org (Postfix, from userid 48)
 id 3D7BB387087A; Sat, 13 Feb 2021 21:55:26 +0000 (GMT)
DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 3D7BB387087A
From: "barry.revzin at gmail dot com" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug c++/99091] New: constexpr variable evaluated at runtime
Date: Sat, 13 Feb 2021 21:55:26 +0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: new
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: c++
X-Bugzilla-Version: 10.0
X-Bugzilla-Keywords: 
X-Bugzilla-Severity: normal
X-Bugzilla-Who: barry.revzin at gmail dot com
X-Bugzilla-Status: UNCONFIRMED
X-Bugzilla-Resolution: 
X-Bugzilla-Priority: P3
X-Bugzilla-Assigned-To: unassigned at gcc dot gnu.org
X-Bugzilla-Target-Milestone: ---
X-Bugzilla-Flags: 
X-Bugzilla-Changed-Fields: bug_id short_desc product version bug_status
 bug_severity priority component assigned_to reporter target_milestone
Message-ID: <bug-99091-4@http.gcc.gnu.org/bugzilla/>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
X-Bugzilla-URL: http://gcc.gnu.org/bugzilla/
Auto-Submitted: auto-generated
MIME-Version: 1.0
X-BeenThere: gcc-bugs@gcc.gnu.org
X-Mailman-Version: 2.1.29
Precedence: list
List-Id: Gcc-bugs mailing list <gcc-bugs.gcc.gnu.org>
List-Unsubscribe: <https://gcc.gnu.org/mailman/options/gcc-bugs>,
 <mailto:gcc-bugs-request@gcc.gnu.org?subject=unsubscribe>
List-Archive: <https://gcc.gnu.org/pipermail/gcc-bugs/>
List-Post: <mailto:gcc-bugs@gcc.gnu.org>
List-Help: <mailto:gcc-bugs-request@gcc.gnu.org?subject=help>
List-Subscribe: <https://gcc.gnu.org/mailman/listinfo/gcc-bugs>,
 <mailto:gcc-bugs-request@gcc.gnu.org?subject=subscribe>
X-List-Received-Date: Sat, 13 Feb 2021 21:55:26 -0000

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=3D99091

            Bug ID: 99091
           Summary: constexpr variable evaluated at runtime
           Product: gcc
           Version: 10.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: barry.revzin at gmail dot com
  Target Milestone: ---

Consider the following (link to compiler explorer:
https://godbolt.org/z/jWsxsK)

#include <array>

static constexpr auto doy =3D []{
    std::array<int, 13> month_offset{};
    for (int m=3D1; m<=3D12;++m) {
        month_offset[m] =3D (153*(m > 2 ? m-3 : m+9) + 2)/5;
    }
    return month_offset;
}();

auto foo(int m) -> int {
    #ifdef LOCAL
    constexpr auto doy =3D []{
        std::array<int, 13> month_offset{};
        for (int m=3D1; m<=3D12;++m) {
            month_offset[m] =3D (153*(m > 2 ? m-3 : m+9) + 2)/5;
        }
        return month_offset;
    }();
    #endif

    return doy[m];
}

This is a fragment of code that does some calculation to figure out the dat=
e,
which isn't super important. If LOCAL is not defined (i.e. we declare the a=
rray
as a namespace-scope constexpr), the -O3 codegen of 'foo' is:

foo(int):
        movsx   rdi, edi
        mov     eax, DWORD PTR doy[0+rdi*4]
        ret

However, if we want to move the declaration of doy to be more local to the
calculation and compile with -DLOCAL, the codegen instead is (on -O3):

foo(int):
        pxor    xmm0, xmm0
        mov     rax, QWORD PTR .LC0[rip]
        movsx   rdi, edi
        mov     DWORD PTR [rsp-24], 275
        movaps  XMMWORD PTR [rsp-72], xmm0
        movdqa  xmm0, XMMWORD PTR .LC1[rip]
        mov     QWORD PTR [rsp-68], rax
        movaps  XMMWORD PTR [rsp-56], xmm0
        movdqa  xmm0, XMMWORD PTR .LC2[rip]
        movaps  XMMWORD PTR [rsp-40], xmm0
        mov     eax, DWORD PTR [rsp-72+rdi*4]
        ret

This can be alleviated by marked the local variable doy as being static
constexpr. Except that this prevents 'foo' from being a constexpr function
(can't have static variables).=20

This just seems like bad codegen, I'm not sure there's any reason that the
compiler needs to do work here. It would be nice if the codegen with a local
constexpr variable were the same as if it were a non-local constexpr variab=
le.=