[Bug c++/95383] New: Poor codegen when constructing a trivial Optional

[Bug c++/95383] New: Poor codegen when constructing a trivial Optional

[barry.revzin at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95383

            Bug ID: 95383
           Summary: Poor codegen when constructing a trivial Optional
           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: ---

Here is a complete example:

struct nullopt_t {} inline constexpr nullopt{};

template <typename T>
struct Optional {
    struct Empty { };
    union {
        Empty _;
        T value;
    };
    bool engaged;

    Optional(nullopt_t) : _(), engaged(false) { }
    Optional(T v) : value(v), engaged(true) { }
};

Optional<int> foo(bool b) {
    if (b) {
        return 42;
    }
    return nullopt;
}

Optional here is a valid implementation strategy for trivial types, like int.
You can see some codegen examples here: https://godbolt.org/z/KwuWzB

gcc 10.1 -O2 emits:

foo(bool):
        test    dil, dil
        mov     eax, 0
        movabs  rdx, 4294967338
        cmovne  rax, rdx
        ret

gcc 10.1 -O3 is worse:

foo(bool):
        xor     eax, eax
        mov     ecx, 42
        test    dil, dil
        cmovne  rdx, rcx
        mov     ecx, 1
        cmovne  rax, rcx
        movabs  rcx, -1095216660481
        and     rdx, rcx
        sal     rax, 32
        or      rax, rdx
        ret

gcc trunk (as of today) -O2 is the same as this bad -O3 version.

clang 10, on the other hand, on -O2 or -O3, emits:

foo(bool):                                # @foo(bool)
        shl     rdi, 32
        lea     rax, [rdi + 42]
        ret

which is much better. 

Using std::optional instead of this Optional (https://godbolt.org/z/By-fYx) for
comparison, clang emits the same code as above. gcc 10 -O3 emits a branch:

foo(bool):
        xor     eax, eax
        test    dil, dil
        je      .L2
        mov     DWORD PTR [rsp-8], 42
        mov     eax, 1
.L2:
        mov     BYTE PTR [rsp-4], al
        mov     rax, QWORD PTR [rsp-8]
        ret

[Bug c++/95383] Poor codegen when constructing a trivial Optional

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95383

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jamborm at gcc dot gnu.org,
                   |                            |rguenth at gcc dot gnu.org
           Keywords|                            |missed-optimization
   Last reconfirmed|                            |2020-05-29
             Status|UNCONFIRMED                 |NEW
     Ever confirmed|0                           |1

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
I see the cited -O3 code-gen with -O2 on trunk.  We expand from

foo (bool b)
{
  struct Optional D.2251;
  unsigned char _7;

  <bb 2> [local count: 1073741824]:
  if (b_2(D) != 0)
    goto <bb 3>; [50.00%]
  else
    goto <bb 4>; [50.00%]

  <bb 3> [local count: 536870913]:
  D.2251.D.2152.value = 42;

  <bb 4> [local count: 1073741824]:
  # _7 = PHI <1(3), 0(2)>
  MEM <unsigned char> [(struct Optional *)&D.2251 + 4B] = _7;
  return D.2251;

where the main issue is likely we fail to elide D.2251 because of the
aggregate return (and us not exposing the ABI, etc.).

Note above D.2251 is partly initialized which makes it a bit more difficult
to optimize.  SRA might try to handle these cases by rematerializing the
aggregate just before the return.  We can then hope for RTL expansion
to do "magic".

[Bug middle-end/95383] Poor codegen when constructing a trivial Optional

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95383

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|c++                         |middle-end
           See Also|                            |https://gcc.gnu.org/bugzill
                   |                            |a/show_bug.cgi?id=95405,
                   |                            |https://gcc.gnu.org/bugzill
                   |                            |a/show_bug.cgi?id=101326
             Blocks|                            |101926
           Severity|normal                      |enhancement


Referenced Bugs:

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101926
[Bug 101926] [meta-bug] struct/complex argument passing and return should be
improved

[Bug middle-end/95383] Poor codegen when constructing a trivial Optional

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95383

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|---                         |DUPLICATE

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Dup.

*** This bug has been marked as a duplicate of bug 95405 ***