[Bug target/66697] Feature request: -mstackrealign and force_align_arg_pointer for x86_64

["ubizjak at gmail dot com" <gcc-bugzilla at gcc dot gnu dot org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66697

--- Comment #2 from UroÅ Bizjak <ubizjak at gmail dot com> ---
IMO, if "outside" code misaligns stack, then these applications need support
for -mincoming-stack-boundary=3, which is currently limited on x86_64 to 4.

Documentation says:

'-mincoming-stack-boundary=NUM'
     Assume the incoming stack is aligned to a 2 raised to NUM byte
     boundary.  If '-mincoming-stack-boundary' is not specified, the one
     specified by '-mpreferred-stack-boundary' is used.

     [...]

Following patch:

--cut here--
Index: config/i386/i386.c
===================================================================
--- config/i386/i386.c  (revision 228460)
+++ config/i386/i386.c  (working copy)
@@ -5102,8 +5102,7 @@ ix86_option_override_internal (bool main_args_p,
   ix86_incoming_stack_boundary = ix86_default_incoming_stack_boundary;
   if (opts_set->x_ix86_incoming_stack_boundary_arg)
     {
-      int min = (TARGET_64BIT_P (opts->x_ix86_isa_flags)
-                ? (TARGET_SSE_P (opts->x_ix86_isa_flags) ? 4 : 3) : 2);
+      int min = TARGET_64BIT_P (opts->x_ix86_isa_flags) ? 3 : 2;

       if (opts->x_ix86_incoming_stack_boundary_arg < min
          || opts->x_ix86_incoming_stack_boundary_arg > 12)
--cut here--

will compile following test:

-- cut here--
typedef float v4sf __attribute__((vector_size(16)));

v4sf test (v4sf a, v4sf b)
{
  volatile v4sf z = a + b;
  return z;
}
--cut here--

using "-O2 -mincoming-stack-boundary=3" to:

0000000000000000 <test>:
   0:   4c 8d 54 24 08          lea    0x8(%rsp),%r10
   5:   0f 58 c8                addps  %xmm0,%xmm1
   8:   48 83 e4 f0             and    $0xfffffffffffffff0,%rsp
   c:   41 ff 72 f8             pushq  -0x8(%r10)
  10:   55                      push   %rbp
  11:   48 89 e5                mov    %rsp,%rbp
  14:   41 52                   push   %r10
  16:   0f 29 4d e0             movaps %xmm1,-0x20(%rbp)
  1a:   0f 28 45 e0             movaps -0x20(%rbp),%xmm0
  1e:   41 5a                   pop    %r10
  20:   5d                      pop    %rbp
  21:   49 8d 62 f8             lea    -0x8(%r10),%rsp
  25:   c3                      retq   

The stack realignment code ensures ABI-compliant stack alignment in all
functions - and is kind of punishment for rogue application.