Broadcast 128 bits from memory (composed of 4 packed single-precision (32-bit) floating-point elements) to all elements. This effectively duplicates the 128-bit vector.