[Bug c/40120] New: aliasing of __m128i is not allowed

[kretz at kde dot org]

emmintrin.h says
/* The Intel API is flexible enough that we must allow aliasing with other
   vector types, and their scalar components.  */
typedef long long __m128i __attribute__ ((__vector_size__ (16),
__may_alias__));

but the following testcase fails:

#include <emmintrin.h>
#include <stdio.h>

int main()
{
    const __m128i d = _mm_set1_epi32(1);
    int i;
    for (i = 0; i < 4; ++i) {
        if (((const int *)&d)[i] != 1) {
            printf("%d %d %d %d\n",
                    ((const int *)&d)[0],
                    ((const int *)&d)[1],
                    ((const int *)&d)[2],
                    ((const int *)&d)[3]);
            return 1;
        }
    }

    return 0;
}

It does not fail when compiled with -fno-strict-aliasing. It also does not fail
when using __m128, _mm_set1_ps and casts to float *.

According to my tests the above does not fail with gcc 4.2, and still fails
with 4.4.0.


-- 
           Summary: aliasing of __m128i is not allowed
           Product: gcc
           Version: 4.3.2
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: kretz at kde dot org
 GCC build triplet: x86_64-unknown-linux-gnu
  GCC host triplet: x86_64-unknown-linux-gnu
GCC target triplet: x86_64-unknown-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40120