Electrical computers and digital processing systems: processing – Processing architecture – Array processor
Reexamination Certificate
2000-02-24
2003-06-10
Ellis, Richard L. (Department: 2183)
Electrical computers and digital processing systems: processing
Processing architecture
Array processor
C712S017000, C712S019000
Reexamination Certificate
active
06578133
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates to digital circuit implementation and more particularly relates to a new silicon device architecture for quick and efficient design and implementation of digital integrated circuits.
BACKGROUND OF THE INVENTION
The art of digital integrated circuit implementation in silicon is well established. There are numerous technologies such as Application Specific Integrated Circuits (ASIC), Field Programmable Gate Arrays (FPGAs) and Complex Programmable Logic Devices (CPLD) which implement digital circuits by means of interconnected cells such as gates, flip-flops and other digital elements. The biggest advantage of these technologies is fast operation caused by parallel activity of all digital elements.
However, the present technologies also have a number of shortcomings that are becoming more noticeable as the cell geometry shrinks and design size increases. For example, the place and route time for such devices increases rather exponentially. The present technologies also require cumbersome and labor intensive critical path timing analysis of the routed designs. This invention eliminates entirely the timing analysis and the design compilation time is linearly proportional to the design size.
The current device architectures that use deep sub-micron silicon technologies require complex analysis of cells and their connections to determine the overall design performance. New physical phenomena are playing larger role at higher cell densities, making the layout analysis continuously more complex. It is becoming now quite apparent that some form of incremental compilation will be necessary for the layout of high density deep sub-micron devices. However, such incremental silicon compilation would require a substantial human effort and involvement, which will slow even more the design process. It is thus another object of this invention to eliminate the device layout, requiring incremental compilation altogether.
Efficient testing of complex devices requires placing additional boundary-scan circuits in the silicon. This makes the design more complex and lowers the effective utilization of silicon surface area. However, since there is no other good way to test the silicon, this process is widely applied to ASIC devices in excess of 100,000 gates. It is yet another object of this invention to provide for effective device testing without the need for any additional boundary scan or similar circuits.
Due to a random nature of cell utilization, relatively large areas of the silicon are set aside to facilitate connections between cells in gate arrays, CPLDs and FPGAs. This lowers the effective utilization of the silicon. The devices built per my invention are based on highly regular memory architecture and do not require design dependent interconnect areas, thus improving the silicon surface area utilization.
The current technologies dissipate a large amount of heat because all circuits operate in parallel. This limits the design size that can be placed on the silicon. This invention describes a parallel-serial circuit operation, which lowers power dissipation and allows considerably higher circuit densities.
The current design methodologies are based more on art than strict mathematical algorithms. As a result, designers must manually tweak some of the circuits for better performance or improved area utilization. This requires high level of expertise, constant employee education and trial and error approach for best results. Another object is to eliminate manual tweaking of designs and lower the expertise demanded from a designer. Since this invention is based on mathematical algorithms, it provides a fully automated design environment, which eliminates manual tweaking of designs and lowers the level of expertise demanded from the designer.
Since the current design tool technology is tuned to the physical phenomena in silicon, designers continuously need to buy newer and more advanced software as technology changes. This invention isolates the designer from changes in the silicon technologies so that one tool set will be able to handle all future silicon process enhancements. Another object is to isolate designs from changes in silicon technology by handling. all future silicon process enhancements with one tool set.
The current silicon production growth outstrips design tool capabilities by a large margin. While the silicon technologies allow building devices with tens of millions of transistors, the current commercial design tools limit that number to around ten million transistors. Because of that, the design tools are a major drag on further spread of silicon device applications. However, the fault is not with the design tools but with the way digital circuits are implemented in silicon. The currently used silicon architectures make design tools unnecessarily slow, cumbersome and difficult to use. It is thus the purpose of this invention to provide such silicon architecture that would facilitate fast and simple design tools, requiring minimum learning and no expertise in device architecture.
Another object of this invention is to eliminate the device layout for each new design, requiring tedious and time consuming incremental compilation.
It is yet another object of this invention to provide for effective device testing without the need for any additional boundary scan or similar circuits.
The devices built per the invention are based on highly regular memory architecture and do not require design dependent interconnect areas, thus improving the silicon surface area utilization.
This invention describes a parallel-serial circuit operation, which lowers power dissipation and allows considerably higher circuit densities.
Another object is to eliminate manual tweaking of designs and lower the expertise demanded from a designer.
Another object is to isolate a design from the applied silicon technologies by handling all future silicon process enhancements with one tool set.
Another object of the invention is to provide silicon architecture that facilitates fast and simple design tools to minimize learning and require no expertise in device architecture.
BRIEF DESCRIPTION OF THE INVENTION
The purpose of this invention is to better utilize the silicon surface area by using highly regular memory structure for logic applications instead of such cells as gates and flip-flops. Sequencers are logic processors that do not have jump, branch or similar instructions that change the order of instruction execution. This invention describes how a set of synchronized sequencers can be used for quick and efficient processing of system level designs.
First, a system level description in the form of schematics and VHDL, Verilog or other hardware description language files is converted into an equivalent set of Boolean logic equations. Next, these equations are mathematically optimized, converted into binary instruction code and distributed between sequencers according to an algorithm that simplifies communication between design sections located in different sequencers. For optimum utilization of silicon resources, approximately the same number of instructions is assigned to each sequencer.
Each sequencer uses an associated program memory for storing binary instructions executing the assigned logic equations and data memory for storing the results of logic equation processing. During device programming the associated program memory for each sequencer is loaded with its own set of machine instructions representing the associated logic equations. Since all sequencers and the associated program memories are driven preferably by a single common address register, they operate and execute logic equations synchronously with each other.
Because of the synchronous operation of sequencers, the compiler that distributes logic equations between them can calculate in advance at what time the result of each logic operation will be provided on the selected sequencer output. A compiler then arranges instructions in the interconnected neighboring sequencer so that it will be re
Ellis Richard L.
O'Reilly David
LandOfFree
MIMD array of single bit processors for processing logic... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with MIMD array of single bit processors for processing logic..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and MIMD array of single bit processors for processing logic... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3103829