_mm_unpackhi_epi32
_mm_tmp1 = _mm_unpackhi_epi32(r0, r1); \
_mm_tmp3 = _mm_unpackhi_epi32(r2, r3); \
__m128 lo = (__m128)_mm_unpackhi_epi32(a, b); /* (a[2], b[2], ...) */
__m128 hi = (__m128)_mm_unpackhi_epi32(c, d); /* (c[2], d[2], ...) */
__m128 lo = (__m128)_mm_unpackhi_epi32(a, b); /* (..., a[3], b[3]) */
__m128 hi = (__m128)_mm_unpackhi_epi32(c, d); /* (..., c[3], d[3]) */