Automatic generation of application specific processors

ABSTRACT

An application-specific instruction-set processor (ASIP) is ideally suited for embedded applications that have demanding performance, size, and power requirements that cannot be satisfied by a general purpose processor. ASIPs also have time-to-market and programmability advantages when compared to custom ASICs. The AutoTIE system simplifies the creation of ASIPs by automatically enhancing a base processor with application specific instruction set architecture (ISA) extensions, including instructions, operations, and register files. The new instructions, operations, and register files are automatically recognized and exploited by the entire software tool chain, including the C/C++ compiler. Thus, taking advantage of the generated ASIP does not require any changes to the application or any assembly language coding. AutoTIE uses the C/C++ compiler to analyze an application, and based on the analysis generates thousands, or even millions, of possible ISA extensions for the application. AutoTIE then uses performance and hardware estimation techniques to combine the ISA extensions into a large number of potential ASIPs, and for a range of hardware costs, chooses the ASIP that provides the maximum performance improvement. For example, for an application performing a radix-4 FFT, AutoTIE considers over 34,000 potential sets of ISA extensions. For hardware costs ranging from 7800 gates to 128,000 gates, AutoTIE combines these extensions to form 31 ASIPs, which provide performance improvements ranging from a factor of 1.12 to a factor of 11.3 compared to a general-purpose processor.

REFERENCES

• 1: John L. Hennessy , David A. Patterson, Computer architecture: a quantitative approach, Morgan Kaufmann Publishers Inc., San Francisco, CA, 02
• 2: Ricardo E. Gonzalez, Xtensa: A Configurable and Extensible Processor, IEEE Micro, v.20 n.2, p.60-70, March 2000 [doi>10.1109/40.848473]
• 3: M. Lam, Software pipelining: an effective scheduling technique for VLIW machines, Proceedings of the ACM SIGPLAN 1988 conference on Programming Language design and Implementation, p.318-328, June 20-24, 1988, Atlanta, Georgia, United States
• 4: Shail Aditya , B. Ramakrishna Rau , Vinod Kathail, Automatic Architectural Synthesis of VLIW and EPIC Processors, Proceedings of the 12th international symposium on System synthesis, p.107, November 01-04, 1999
• 5: Michael Joseph Wolfe, Optimizing Supercompilers for Supercomputers, MIT Press, Cambridge, MA, 1990
• 6: Hans Zima , Barbara Chapman, Supercompilers for parallel and vector computers, ACM Press, New York, NY, 1990
• 7: Xtensa Instruction Set Architecture Reference Manual. Tensilica, Inc., Santa Clara, CA, 2002.
• 8: Armita Peymandoust, Laura Pozzi, Paolo Ienne, and Giovanni De Micheli. Automatic instruction-set extension and utilization for embedded processors. In 14th International Conference on Application-specific Systems, Architectures and Processors, June 2003.
• 9: Bernardo Kastrup , Arjan Bink , Jan Hoogerbrugge, ConCISe: A Compiler-Driven CPLD-Based Instruction Set Accelerator, Proceedings of the Seventh Annual IEEE Symposium on Field-Programmable Custom Computing Machines, p.92, April 21-23, 1999
• 10: Kubilay Atasu , Laura Pozzi , Paolo Ienne, Automatic application-specific instruction-set extensions under microarchitectural constraints, Proceedings of the 40th conference on Design automation, June 02-06, 2003, Anaheim, CA, USA [doi>10.1145/775832.775897]

INDEX TERMS

Primary Classification:
  C. Computer Systems Organization
  C.1 PROCESSOR ARCHITECTURES
      C.1.3 Other Architecture Styles

General Terms:
Algorithms, Design, Measurement, Performance

Keywords:
ASIPs, automatic instruction-set generation, configurable processors, extensible processors