https://issues.dlang.org/show_bug.cgi?id=20126

          Issue ID: 20126
           Summary: bad codegen for return statement in function with asm
                    code
           Product: D
           Version: D2
          Hardware: All
                OS: All
            Status: NEW
          Severity: regression
          Priority: P1
         Component: dmd
          Assignee: nobody@puremagic.com
          Reporter: turkeyman@gmail.com

On my local 2.087, this code works, but in master, it generates bad code:

asmtest.d
---------

extern(C) float floop(float* r, float x)
{
  asm
  {
    mov EAX, x;
    mov RCX, r;
    xchg [RCX], EAX;
    mov x, EAX;
  }
  return x;
}


> dmd asmtest.d -c -m64

Dumping this emits:

0000000000000000 <floop>:
   0:   55                      push   %rbp
   1:   48 8b ec                mov    %rsp,%rbp
   4:   48 83 ec 10             sub    $0x10,%rsp
   8:   48 89 7d f0             mov    %rdi,-0x10(%rbp)
   c:   f3 0f 11 45 f8          movss  %xmm0,-0x8(%rbp)

  11:   8b 45 f8                mov    -0x8(%rbp),%eax
  14:   48 8b 4d f0             mov    -0x10(%rbp),%rcx
  18:   87 01                   xchg   %eax,(%rcx)
  1a:   89 45 f8                mov    %eax,-0x8(%rbp)

  1d:   c9                      leaveq
  1e:   c3                      retq


It's missing this before the return:
  movss 0x8(%rbp),%xmm0

You can see where it loads from the argument in xmm0 to -0x8(%rbp) to transfer
to RAX in the asm code, then the asm code as written.
It writes the modified value back to the stack, but the `return x;` statement
should load `x` back into xmm0 to return the value, but it just skips that.

I presume that it assumes xmm0 was not clobbered by the asm block and elides the return code, but the variable was written...

This fails on x64 in master on both windows and linux, but it works correctly
for -m32 (does emit an fld before return).
It also used to work on x64 in 2.087.0 (which i'm running locally)

--