Potential-driven statistical ordering of transformations

ABSTRACT

Successive, well organized application of transformations has beenwidely recognized as an exceptionally effective, but complex anddifficult CAD task. We introduce a new potential-driven statisticalapproach for ordering transformations. Two new synthesis ideasare the backbone of the approach. The first idea is to quantifythe characteristics of all transformations and the relationship betweenthem based on their potential to reorganize a computationsuch that the complexity of the corresponding implementation isreduced. The second one is based on the observation that transformationsmay disable each other not only because they prevent theapplication of the other transformation, but also because both transformationstarget the same potential of the computation. These twoobservations drastically reduce the search space to find efficient andeffective scripts for ordering transformations. A key algorithmicnovelty is that both conceptual and optimization insights as well asall optimization algorithms are automatically derived by organizedexperimentation and statistical methods. On a large set of diversereal-life examples improvements in throughput, area, and power bylarge factors have been obtained. Both qualitative and quantitativestatistical analysis indicate effectiveness, high robustness, and consistencyof the new approach for ordering transformations.

REFERENCES

• 1: David F. Bacon , Susan L. Graham , Oliver J. Sharp, Compiler transformations for high-performance computing, ACM Computing Surveys (CSUR), v.26 n.4, p.345-420, Dec. 1994 [doi>10.1145/197405.197406]
• 2: Robert King Brayton , Alberto L. Sangiovanni-Vincentelli , Curtis T. McMullen , Gary D. Hachtel, Logic Minimization Algorithms for VLSI Synthesis, Kluwer Academic Publishers, Norwell, MA, 1984
• 3: A.R Chandrakasan, M. Potkonjak, R. Mehra, J. Rabaey, and R. Brodersen. Optimizing power using transformations. IEEE Transactions on CAD, 14(1):12, 1995.
• 4: Sujit Dey , Miodrag Potkonjak , Steven G. Rothweiler, Performance optimization of sequential circuits by eliminating retiming bottlenecks, Proceedings of the 1992 IEEE/ACM international conference on Computer-aided design, p.504-509, November 1992, Santa Clara, California, United States
• 5: Sujit Dey , Miodrag Potkonjak , Rabindra K. Roy, Exploiting hardware sharing in high-level synthesis for partial scan optimization, Proceedings of the 1993 IEEE/ACM international conference on Computer-aided design, p.20-25, November 07-11, 1993, Santa Clara, California, United States
• 6: B. Efron and R.J. Tibashirani. An introduction to the bootstrap. Chapman & Hall, New York, NY, 1993.
• 7: S. Kirkpatrick, C. Gelatt, and M. Vecchi. Optimization by simulated annealing. Science, 220(4598):671-680, 1983.
• 8: Henry Massalin, Superoptimizer: a look at the smallest program, Proceedings of the second international conference on Architectual support for programming languages and operating systems, p.122-126, October 1987, Palo Alto, California, United States
• 9: W. M. McKeeman, Peephole optimization, Communications of the ACM, v.8 n.7, p.443-444, July 1965 [doi>10.1145/364995.365000]
• 10: Christos H. Papadimitriou , Kenneth Steiglitz, Combinatorial optimization: algorithms and complexity, Prentice-Hall, Inc., Upper Saddle River, NJ, 1982
• 11: Miodrag Potkonjak , Jan Rabaey, Maximally fast and arbitrarily fast implementation of linear computations, Proceedings of the 1992 IEEE/ACM international conference on Computer-aided design, p.304-308, November 1992, Santa Clara, California, United States
• 12: J. M. Rabaey , C. Chu , P. Hoang , M. Potkonjak, Fast Prototyping of Datapath-Intensive Architectures, IEEE Design & Test, v.8 n.2, p.40-51, April 1991 [doi>10.1109/54.82037]
• 13: M. E. Wolf , M. S. Lam, A Loop Transformation Theory and an Algorithm to Maximize Parallelism, IEEE Transactions on Parallel and Distributed Systems, v.2 n.4, p.452-471, October 1991 [doi>10.1109/71.97902]

INDEX TERMS

Primary Classification:
  J. Computer Applications

Additional Classification:
  F. Theory of Computation
  F.1 COMPUTATION BY ABSTRACT DEVICES
      F.1.2 Modes of Computation
          Subjects: Probabilistic computation
  F.2 ANALYSIS OF ALGORITHMS AND PROBLEM COMPLEXITY
      F.2.1 Numerical Algorithms and Problems
          Subjects: Computation of transforms (e.g., fast Fourier transform)

General Terms:
Design, Experimentation, Measurement, Performance, Theory