Design and Performance Analysis of Reconfigurable Hybridized Macro Pipeline Multiprocessor System (HMPM)

Multiprocessor systems’ speed and throughput has surpassed the superscalar uniprocessor, making the attentions drawn towards new ways of increasing the efficiency and speed of the multiprocessor. Software developmental approaches are now being tailored towards multiprocessing by exploiting parallel programming technique. Compiler developers are not left behind, new and existing compilers are shaping and reshaping for parallel programming. If multiple processing elements are applied to execution of a single program, they can either be used concurrently, that is, the program must be structured into several concurrent subtasks. Alternatively may be executed sequentially, in a pipeline fashion, where each processing element handlees an independent subtask at any point in time. Many researchers had worked on different architecture to develop multiprocessor systems. Some of these are HYDRA (developed in Stanford University), an on-core multiprocessors architecture, which uses parallel execution technique for multiprocessing. MaRs is another multiprocessors system which uses macro pipeline technique for multiprocessing, PMMLA, a reconfigurable multiprocessors system is another example. However, most of these systems are application specific. Some are good for executing parallel tasks while some are good for sequential task, depends on the technique used by the architecture to perform multiprocessing. Hybridized Macro Pipeline Multiprocessor System (HMPM) is another multiprocessors system, which hybridized the two multiprocessing techniques. The aim of this paper is to develop a hybridize multiprocessor system and a heuristic for reconfiguring the HMPM design in order to enforce load balance, and to detemine the performance of the system by modeling its energy consumption and memory accessing speed. Keywords—Reconfigurable, Energy Consumption, Macro Pipeline, Parallelism, Multiprocessor