This has the benefit of removing any SSE -> AVX penalty that may happen when the compiler emits VEX encoded instructions. Signed-off-by: James Almer <jamrial@gmail.com>