MICROCOMPUTER WITH DISCONNECTED, OPEN, INDEPENDENT, BIMEMORY ARCHITECTURE, ALLOWING LARGE INTERACTING, INTERCONNECTED MULTI-MICROCOMPUTER PARALLEL SYSTEMS ACCOMMODATING MULTIPLE LEVELS OF PROGRAMMER DEFINED HIERARCHY
The first basic problem of present multiprocessor systems is that when more than two levels of hierarchy are included or more than about twenty processors are interconnected, the hardware and software overhead requirements become so burdensome that additional levels of hierarchy and additional processors have not increased the overall effectiveness or through put of the multiprocessor system. Many of the present multiprocessor systems use one interconnecting circuit to carry data information some of the time and use the same interconnecting circuit to carry address information at other times. This dual use of one circuit to carry both data information and address information, in present multiprocessor systems, requires many additional timing circuits, additional switching circuits, additional arbitration logic circuits and additional error checking logic circuits, and a very large number of additional interrupts when circulating Pierce loops are also used, between the various interconnected processors. This additional complexity, in the interconnecting circuits, is a very large part of the reason that more than two levels of hierarchy and more than about twenty processors being interconnected have not increased the overall effectiveness or through put of the multiprocessor system.
The second basic problem of present multiprocessor systems is most multiprocessor systems are designed with essentially permanent connections between the various clusters of processors, creating one basic system of mechanical connections between the processors. To mimic human thinking to the desired degree requires the ability to use the human thinking of the field application system designers and the programmers to decide which sets of billions of different optional mechanical and logical combinations of interconnections between the individual processors in a multiprocessor system are most effective for a given problem at a given moment in time, based on the actual data flow in the problem, Cat computer speeds. None of the present multiprocessor systems permit the field application system designers and the programmers to arbitrarily change the mechanical interconnections between the various processors in the billions of different ways the human field application system designers and programmers think the processors should be interconnected.
The human programmers and field application system designers think the processors should be interconnected this way today and change their mind and think they should be interconnected in a different way tomorrow. Sometimes the human programmers think the individual processors should be interconnected one way and the human field application system designers think the individual processors should be interconnected in a different way. Only actual, real time, testing in the field of the various sets of billions of mechanical and logical interconnections will make the multicomputer system "be capable of mimicking human thinking" to the required degree.
The third basic problem of present multiprocessor systems is the inability to make existing multiprocessor systems work as truly "parallel processors". The Strategic Defense Initiative Organization problem and the World Climate Prediction problem probably both require very highly parallel architectures to be effective. Over 200 billion interacting floating point calculations are required per second on the World Climate Prediction problem as a minimum, and the SDIO requirement is very much larger. The SDIO problem could probably use 20 trillion interacting floating point calculations per second.
The fourth basic problem of present multiprocessor systems is removing individual processors that have failed or have been damaged by enemy action from a running system without causing further failure of the overall program logic of the balance of the remaining running system.
The fifth basic problem of present multiprocessor systems is adding individual processors to a running system without causing failure of the overall program logic of the balance of the running system.
The sixth basic problem of present multiprocessor systems is some of the switching circuits and some of the arbitration logic circuits and some of the error checking logic circuits needed to effectively interconnect several thousand processors, are being placed outside of the individual processors. These different external logic circuits and BLOCK/SHORT circuits in circulating Pierce loop circuits, preclude using one standard group of circuits made up of simple, dedicated purpose, single line conductors, where the standard group of circuits can be of various random lengths -- less than a maximum length -- to interconnect any two processors.
The seventh basic problem of present multiprocessor systems is that more than one standard group of circuits are used in the multiprocessor system to interconnect processors on different hierarchy levels. For example, some present multiprocessor systems, like the "Restructurable Integrated Circuit For Implementing Digital Systems" as described by Rob Budzinski, John Linn, and Satish Thatte, COMPUTER, March 1982, pages 43 through 54, use the concept of external coordination of microprocessors as developed by R. G. Arnold and E. W. Page, Proc. 3rd Ann. Symp. Computer architecture 1976, pages 40-45, where both time-shared busses referred to above, and "circulating" Pierce Loop busses are used on each level of hierarchy. The highest hierarchy level of busses (Master Control Group) use "circulating" busses without special "BLOCK/SHORT" logic circuit modules. Each additional level of hierarchy (Control Group), below the Master Control Group, requires an additional set of busses with special "BLOCK/SHORT" logic circuit modules.
This requirement of an additional set of busses (Control Group) for each level of hierarchy caused Arnold and Page to state in their OBSERVATIONS AND CONCLUSIONS, "As is evident from the examples, only a limited number of Control Groups are likely to be used". When one thinks of one hundred levels of hierarchy requiring one hundred separate sets of (Control Group) busses interconnecting each of several thousand processors, it is easy to understand Arnold and Page's above stated conclusions.
The Bimemory Independent Central Processing Unit microcomputer (BICPU microcomputer) invention, as herein taught, solves all seven of these basic problems. The BICPU microcomputer is actually capable of mimicking the human thinking of the field application system designers and the human programmers involved in the system. If the human programmer can think of a solution, or the human field application system designer can think of a solution, several thousand BICPU microcomputers, mechanically interconnected in one of several billion different unique systems, where the programmers, at computer speeds, can logically connect and logically disconnect the thousands of BICPU microcomputers in several billion additional unique ways, using logical bimemory "S" hookups and logical bimemory "Y" hookups, probably will "be capable of mimicking human thinking" to the required degree.
