Copy a, then insert 128 bits (composed of 2 packed double-precision (64-bit) floating-point elements) from b at the location specified by imm8.