A pipelined multimicroprocessor system for on-line compiling

Programming languages tend to be more and more frequently interpreted rather than compiled. This is mainly because hardware is becoming cheaper and newer hardware with a still higher degree of complexity is rapidly appearing.

The well-known separation of interpretation and execution is used together with an intermediate code to present a technique which achieves an important improvement in program execution speed. The solution given involves a highly interactive multiprocessor with extremely short parallel processes. A pipelining system at the level of the intermediate code together with a first-in first-out queuing buffer is used.

Our proposal amounts to combining a set of microprocessors into a multiprocessor system. The set consists of an interpretation processor and one or more execution processors. The execution processors run a code that is free of jumps. The performance thus reaches the level of compiled code. No restriction on intermediate codes is made. However, a complex an powerful code that is not storage consuming will generally speed up the type of machine presented.

The machine architecture together with the technological possibilities and the realization of the machine are emphasized in this paper.