[Bug target/77438] MMX intrinsic on x86_64 generates bloated code

[glisse at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=77438

--- Comment #1 from Marc Glisse <glisse at gcc dot gnu.org> ---
Gcc does avoid using the stack when it is more efficient to do so (depends on
the -march setting).
Yes, using SSE would be better.
The general advice is to stop using MMX.
Using gcc's vector extension generates even worse code in this case:
  _8 = VIEW_CONVERT_EXPR<long unsigned int>(x_4(D));
  _9 = VIEW_CONVERT_EXPR<long unsigned int>(y_5(D));
  _10 = _8 ^ _9;
  _11 = _9 & 9187201950435737471;
  _12 = _8 & 9187201950435737471;
  _13 = _10 & 9259542123273814144;
  _14 = _11 + _12;
  _15 = _13 ^ _14;
  _6 = VIEW_CONVERT_EXPR<__m64>(_15);
(I think there is another PR asking that vector lowering allow the use of
larger vector types)