It has been implemented in old BSD and is also available through the sum command line utility.
This algorithm is useless from a security perspective, and is weaker than the CRC-32 cksum for error detection.
[1][2] Below is the relevant part of the GNU sum source code (GPL licensed).
It computes a 16-bit checksum by adding up all bytes (8-bit words) of the input data stream.
As mentioned above, this algorithm computes a checksum by segmenting the data and adding it to an accumulator that is circular right shifted between each summation.