Computer-aided design and analysis of circuits and semiconductor – Nanotechnology related integrated circuit design
Reexamination Certificate
1999-06-17
2002-04-23
Smith, Matthew (Department: 2825)
Computer-aided design and analysis of circuits and semiconductor
Nanotechnology related integrated circuit design
Reexamination Certificate
active
06378122
ABSTRACT:
FIELD OF THE INVENTION
The present invention generally relates to genetic algorithms and evolvable hardware, and more particularly to a method for evolving configuration bitstreams for programmable logic devices.
BACKGROUND
Many conventional design methodologies are based on a structured design approach. That is, high-level requirements are partitioned, perhaps hierarchically, into lower-level requirements. Teams of engineers are assigned to create designs and sub-designs that meet the requirements at the different levels. The structured design methodology is advantageous because it supports building on past experiences in addressing similar types of requirements, thereby promoting development of an effective design in an efficient manner. However, a drawback to traditional design methods is that the same experiences that promote quickly satisfying a requirement may blind engineers to alternative, and perhaps better, solutions.
Work is presently underway to fundamentally change the way in which designs are created. The new methodology uses principles of natural selection from the biological world to create electronic hardware designs. The process is often characterised as “evolvable hardware” or using genetic algorithms to create hardware designs. In an example process of evolving a design, a population of designs is first randomly created, tested, and scored based on the suitability to meet the design requirements. Then based on natural selection principles, certain ones of the designs in the population are selected to “reproduce,” that is, used to create new designs for the population. The process of testing, scoring, and reproducing is then repeated until a suitable design has evolved.
The hardware used in evolving designs is typically some type of programmable logic device. SRAM FPGAs are suitable because they can be programmed, tested, and reprogrammed many times in evolving a design. The XC6200 FPGA from XILINX has been often used for the additional reason that the architecture does not allow contention in the device, no matter what the bitstream is.
A problem often encountered with evolving designs for other SRAM FPGA architectures is that the objective designs are usually intended to be digital solutions for digital devices (e.g., a bitstream for an FPGA), but an evolved design can cause a digital device to exhibit unintended asynchronous behavior and even contain contentions for resources capable of destroying the device. For example, widely used commercial parts, such as the XC4000EX/XL and Virtex FPGAs from XILNX, can be damaged if multiple signals are allowed to drive the same wire. Thus, such parts have been avoided for evolving hardware.
At present, testing the evolving circuits is very time consuming. Many hours are spent reading and writing configuration bitstreams and testing the fitness of the bitstreams. Some implementations use dedicated input/output ports for probing the evolving designs. However, dedicating input/output ports to testing limits the portability of the design. Further, setting up the hardware for a test procedure is expensive. A test system board must include several IC devices for performing the tests, and these IC devices may not be configurable. If these non-configurable IC devices do not perform properly with an evolved design, the tests may not be valid or useful, and the hardware may have to be changed.
Today's evolved circuits tend to be suitable for implementation on specific devices. Irregularities in device fabrication, operating temperature, operating voltage, and other environmental factors can affect a circuit's performance, and thus, its fitness. Therefore, an evolved circuit is rarely suitable for implementation on a variety of devices.
A method that address the aforementioned problems, as well as other related problems, is therefore desirable.
SUMMARY OF THE INVENTION
The invention provides a method and apparatus for allowing multiple populations of configuration bitstreams to evolve in parallel. The method comprises creating respective configuration bitstreams on a data processing system. To better understand the parallel aspect of the evolution, if the bitstreams are evolved using a Java programming language (Java is a Trademark of Sun Microsystems, Inc.), a Java class defines the kind of circuit to be evolved to perform the desired function. Multiple objects of a class are instantiated; that is multiple populations of the desired circuit are set up and are evolved using genetic algorithms. Each of these genetic algorithm classes is threaded so that the microprocessor can perform parallel evolution without the classes getting mixed up. Evolution of each bitstream continues until a predetermined criterion is met. The configuration bitstreams are downloaded from the data processing system to several devices, where data are generated. Each bitstream is downloaded to a separate FPGA or to a separate area of an FPGA, and data are collected from each area for evaluation. The data for the several devices are then evaluated to determine suitability of each of the bitstreams to implement the desired function.
To implement the method, chromosome data structures are established having data associated with programming resources of the area of the programmable logic device. Respective configuration bitstreams are created from data of the chromosome data structures, and the data in the chromosome data structures are evolved, based on the relative suitability, until at least one predetermined criterion is met. The configuration bitstreams are downloaded to several devices. The respective configuration bitstreams are evaluated for relative suitability to meet predetermined criteria when deployed on a programmable logic device
An apparatus for evolving the configuration bitstreams comprises: means for establishing chromosome data structures having data associated with programming resources of the area of the programmable logic device; means for creating respective configuration bitstreams from data of the chromosome data structures; means for evolving data in the chromosome data structures based on the relative suitability until at least one predetermined criterion is met; means for downloading the configuration bitstreams to the device; and means for evaluating the respective configuration bitstreams for relative suitability to meet predetermined criteria when deployed on a programmable logic device.
The above summary of the present invention is not intended to describe each disclosed embodiment of the present invention. The figures and detailed description that follow provide additional example embodiments and aspects of the present invention.
REFERENCES:
patent: 5784577 (1998-07-01), Jacobson et al.
patent: 5794033 (1998-08-01), Aldebert et al.
patent: 5864486 (1999-01-01), Deming et al.
patent: 5995744 (1999-11-01), Guccione
patent: 6023565 (2000-02-01), Lawman et al.
patent: 6035106 (2000-03-01), Carruthers et al.
patent: 6052720 (2000-04-01), Traversat et al.
patent: 6078735 (2000-06-01), Baxter
patent: 6078736 (2000-06-01), Guccione
patent: 6099583 (2000-08-01), Nag
patent: 6102963 (2000-08-01), Agrawal
patent: 6134516 (2000-10-01), Wang et al.
patent: 6161125 (2000-12-01), Traversat et al.
patent: 6167364 (2000-12-01), Stellenberg et al.
patent: 6216258 (2001-04-01), Mohan et al.
patent: 6226776 (2001-05-01), Panchul et al.
patent: 6237029 (2001-05-01), Master
patent: 6243851 (2001-06-01), Hwang et al.
patent: 6272669 (2001-08-01), Anderson et al.
patent: 6279146 (2001-08-01), Evans et al.
patent: 6301695 (2001-10-01), Burnham et al.
patent: 0 657 832 (1995-06-01), None
patent: WO 98 19256 (1998-05-01), None
patent: WO 00 38087 (2000-06-01), None
Adrian Thompson, “Silicon Evolution,” Proceedings of Genetic Programming 1996 (GP96), J.R. Koza et al. (Eds), pp. 444-452, MIT Press 1996.
Levi D and Guccione S. A.: “GeneticFPGA: A Java-Based Tool for Evolving Stable Circuits”, Part of the SPIE Conference on Reconfigurable Technology: FPGAs for Computing Applications, vol. 3844, Sep. 1999 pp. 114-121, Boston USA.
Kitano H. et
Guccione Steven A.
Levi Delon
Do Thuan
Xilinx , Inc.
Young Edel M.
LandOfFree
Method and apparatus for evolving a plurality of versions of... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and apparatus for evolving a plurality of versions of..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and apparatus for evolving a plurality of versions of... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2881122