Expanding MML for dedicated multiple microprocessor applications

The development of dedicated multiple-microprocessor applications can be improved by explicit inclusion and utilization of relevant knowledge. For an integrated development tool featuring the retargetability to prototype different architectures as well as configuration and allocation control, a knowledge-based subsystem can potentially automate the resource allocation process. A system of this type, KMML, is described. The system features a planner and a layered allocator, with emphasis on the allocation of resources, aiming at efficient mapping of logical process-oriented description onto possible architectural configurations. The resource-allocation problem is characterized by a large solution space, high cost in evaluating partial allocations, the general lack of algorithms, and the need to optimize the functional specifications, within constraints that include speed and others. The MML source language is extended to embed strategies of allocation and configuration, to specify constraints, specifications and the control of organizational alternatives