I have idly been prototyping hardware implementations of a hash table lookup. The scenario imagined is where messages arrive containing a textual identifier, and the processing of that message require the retrieval of some corresponding state from memory. Some messages cause a new entry in the store, whilst others are terminal for the associated state, ie. cause the entry to be removed from the table. Example: The message is delivered as a sequence of bytes: offset purpose 0-7 sequence number 8-15 identifier 16 state change type 17 new state Byte 16 has the following enumeration: 0: new message, 1: change to existing state, 2: terminal state reached. The state storage is a fixed number of bytes for each entry. |