AoS => SoA transformation for 3-component vectors

AoS2SoA.c

/*----------------------------------------------------------------------------------------------------------------------
input : srcStream => four 3-component vectors : x0y0z0 x1y1z1 x2y2z2 x3y3z3
output : dstStream => three 4-component vectors : x0x1x2x3 y0y1y2y3 z0z1z2z3
----------------------------------------------------------------------------------------------------------------------*/
void AoS2SoA(float * __restrict dstStream, float const * __restrict srcStream)
{
    __m128 x1_y1_z1_x2 = _mm_load_ps(srcStream + 0);
    __m128 y2_z2_x3_y3 = _mm_load_ps(srcStream + 4);
    __m128 z3_x4_y4_z4 = _mm_load_ps(srcStream + 8);

    __m128 y1_z1_y2_z2 = _mm_shuffle_ps(x1_y1_z1_x2, y2_z2_x3_y3, _MM_SHUFFLE(1, 0, 2, 1));
    __m128 x3_y3_x4_y4 = _mm_shuffle_ps(y2_z2_x3_y3, z3_x4_y4_z4, _MM_SHUFFLE(2, 1, 3, 2));

    __m128 x1_x2_x3_x4 = _mm_shuffle_ps(x1_y1_z1_x2, x3_y3_x4_y4, _MM_SHUFFLE(2, 0, 3, 0));
    __m128 z1_z2_z3_z4 = _mm_shuffle_ps(y1_z1_y2_z2, z3_x4_y4_z4, _MM_SHUFFLE(3, 0, 3, 1));
    __m128 y1_y2_y3_y4 = _mm_shuffle_ps(y1_z1_y2_z2, x3_y3_x4_y4, _MM_SHUFFLE(3, 1, 2, 0));

    _mm_store_ps(dstStream + 0, x1_x2_x3_x4);
    _mm_store_ps(dstStream + 4, y1_y2_y3_y4);
    _mm_store_ps(dstStream + 8, z1_z2_z3_z4);
}