[Bug rtl-optimization/50728] Inefficient vector loads from aggregates passed by value

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50728

--- Comment #2 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-10-14 
13:40:31 UTC ---
typedef __complex__ float Value;
struct A {
  Value a[16 / sizeof (Value)];
} __attribute__ ((aligned(16)));

Value get(A a)
{
  return a.a[0];
}

_Z3get1A:
.LFB0:
        .cfi_startproc
        movq    %xmm0, -24(%rsp)
        movss   -24(%rsp), %xmm0
        movss   %xmm0, -32(%rsp)
        movss   -20(%rsp), %xmm0
        movss   %xmm0, -28(%rsp)
        movq    -32(%rsp), %xmm0
        ret

eh ...