Expand all overlapping 4-bit subsequences of a circular 16-bit sequence

b24_expand.c

static uint64_t
b24_expand (uint16_t seq)
{
	#define seq_prep(x) 0, (uint32_t)((x) << 12U | \
	                       (x) >> 4U) << 16U | (x)
	__m128i k = _mm_shuffle_epi8(
		_mm_set_epi64x(seq_prep(seq)),
		_mm_set_epi64x(UINT64_C(0x0303030301010101), \
		               UINT64_C(0x0202020200000000)));
	#undef seq_prep

	#define pext_mask UINT64_C(0x783c1e0f783c1e0f)
	return _pext_u64((uint64_t)_mm_extract_epi64(k, 1), pext_mask) << 32U
	     | _pext_u64((uint64_t)_mm_extract_epi64(k, 0), pext_mask);
	#undef pext_mask
}