[Bug middle-end/18041] OR of two single-bit bitfields is inefficient

["rguenth at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=18041

--- Comment #7 from Richard Biener <rguenth at gcc dot gnu.org> ---
GCC 6 at -O2 on x86_64 produces

foo:
.LFB0:
        .cfi_startproc
        movzbl  (%rdi), %eax
        movl    %eax, %edx
        shrb    %dl
        orl     %eax, %edx
        andl    $-2, %eax
        andl    $1, %edx
        orl     %edx, %eax
        movb    %al, (%rdi)
        ret

If we lower bitfield accesses to BIT_FIELD_REF / BIT_INSERT_EXPR we get

  <bb 2>:
  _7 = b_5(D)->D.1753;
  _1 = BIT_FIELD_REF <_7, 1, 0>;
  _2 = BIT_FIELD_REF <_7, 1, 1>;
  _3 = _1 | _2;
  _8 = BIT_INSERT_EXPR <_7, _3, 0 (1 bits)>;
  b_5(D)->D.1753 = _8;
  return;

if we lower to shifts/masks then

  <bb 2>:
  _7 = b_5(D)->D.1753;
  _11 = _7 >> 1;
  _1 = _7 | _11;
  _14 = _7 & 254;
  _10 = _1 & 1;
  _16 = _10 | _14;
  b_5(D)->D.1753 = _16;
  return;

it fails to notice that the & 254 is not necessary.  That's easier to
grasp from the first lowering.