Parallel Factorization of Boolean Polynomials

Kulkarni, V and Emelyanov, P and Ponomaryov, D and Krishna, M and Raha, S and Nandy, SK (2019) Parallel Factorization of Boolean Polynomials. In: 12th International Andrei P. Ershov Informatics Conference, PSI 2019; Novosibirsk; Russian Federation, 2-5 July 2019, Novosibirsk; Russian Federation, pp. 80-94.

PDF nikolaj.pdf - Published Version Restricted to Registered users only Download (17MB) | Request a copy

Abstract

Polynomial factorization is a classical algorithmic problem in algebra, which has a wide range of applications. Of special interest is factorization over finite fields, among which the field of order two is probably the most important one due to the relationship to Boolean functions. In particular, factorization of Boolean polynomials corresponds to decomposition of Boolean functions given in the Algebraic Normal Form. It has been also shown that factorization provides a solution to decomposition of functions given in the full DNF (i.e., by a truth table), for positive DNFs, and for cartesian decomposition of relational datatables. These applications show the importance of developing fast and practical factorization algorithms. In the paper, we consider some recently proposed polynomial time factorization algorithms for Boolean polynomials and describe a parallel MIMD implementation thereof, which exploits both the task and data level parallelism. We report on an experimental evaluation, which has been conducted on logic circuit synthesis benchmarks and synthetic polynomials, and show that our implementation significantly improves the efficiency of factorization. Finally, we report on the performance benefits obtained from a parallel algorithm when executed on a massively parallel many core architecture (Redefine).

Item Type:	Conference Paper
Publication:	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Publisher:	Springer
Additional Information:	Copyright of this article belongs to Elsevier B.V.
Keywords:	Boolean functions; Computer circuits; Polynomial approximation; Reconfigurable architectures, Boolean polynomial; Data-level parallelism; Decomposition of functions; Experimental evaluation; Factorization algorithms; Many-core architecture; Polynomial factorization; Reconfigurable computing, Factorization
Department/Centre:	Division of Interdisciplinary Sciences > Computational and Data Sciences
Date Deposited:	28 Jan 2020 07:19
Last Modified:	17 Feb 2020 17:38
URI:	http://eprints.iisc.ac.in/id/eprint/64384

Actions (login required)

View Item