Copy a then insert 128 bits (composed of 4 packed single-precision (32-bit) floating-point elements) from b, at the location specified by imm8.