Tuplets: Words for a Tuple Space Machine

Linda Gel85, CG89] is a coordination language providing generative communication via tuple spaces. Each tuple space is a global associative memory consisting of a bag (or multi-set) of tuples, where the tuples are variable-sized objects. Repeated insertions and removals of variable-sized tuples are liable to result in the memory containing several small areas of free memory to which larger, variable-sized tuples can not be allocated. This memory redundancy then has to be removed by compacting the memory to produce a large, contiguous area of free memory to which large, variable-sixed tu-ples can be allocated. Furthermore, when attempting to implement tuple spaces in content-addressable memory (CAM), variable-sized tuples require variable bucket sizes for hash-based schemes. This, along with the need to compare variable-sized blocks of memory when matching, result in implementation problems. In this report, tuplets are proposed as a solution to the problems of variable-sized tuples. Tuplets are xed-sized tuple components, with each tuple composed of a number of tuplets. Being xed-sized, tuplet storage behaviour is predictable and manageable. Therefore, only xed-sized hash buckets are required, and only xed numbers of words need be compared in hardware CAM schemes. The central design criterion for tuplets is to associate a key with each tuplet, identifying the tuple of which it is a component, and its location in that tuple. Various tuplet design and implementation possibilities are considered and examined. Example programs illustrate the implementation of the various tuplet design possibilities. Furthermore, the examples show that the tuplet operations are of comparable complexity to tuple operations, as well as showing a greater potential for parallel execution.