Abstract not found

Though there is increased activity in the implementation of asymptotically fast polynomial arithmetic, little is reported on the details of such effort. In this paper, we discuss how we achieve high performance in implementing some well-studied fast algorithms for polynomial arithmetic in two high-level programming environments, AXIOM and Aldor. Two approaches are investigated. With Aldor we rely only on high-level generic code, whereas with AXIOM we endeavor to mix high-level, middle-level and low-level specialized code. We show that our implementations are satisfactory compared with other known computer algebra systems or libraries such as Magma v2.11-2 and NTL v5.4. Categories and Subject Descriptors:

We discuss changing the variable ordering for a regular chain in positive dimension. This quite general question has applications going from implicitization problems to the symbolic resolution of some systems of differential algebraic equations. We propose a modular method, reducing the problem to dimension zero and using Newton-Hensel lifting techniques. The problems raised by the choice of the specialization points, the lack of the (crucial) information of what are the free and algebraic variables for the new ordering, and the efficiency regarding the other methods are discussed. Strong hypotheses (but not unusual) for the initial regular chain are required. Change of ordering in dimension zero is taken as a subroutine.

The computation of triangular decompositions involves two fundamental operations: polynomial GCDs modulo regular chains and regularity test modulo saturated ideals. We propose new algorithms for these core operations based on modular methods and fast polynomial arithmetic. We rely on new results connecting polynomial subresultants and GCDs modulo regular chains. We report on extensive experimentation, comparing our code to pre-existing Maple implementations, as well as more optimized Magma functions. In most cases, our new code outperforms the other packages by several orders of magnitude.

Abstract. We introduce the concept of comprehensive triangular decomposition (CTD) for a parametric polynomial system F with coefficients in a field. In broad words, this is a finite partition of the the parameter space into regions, so that within each region the “geometry ” (number of irreducible components together with their dimensions and degrees) of the algebraic variety of the specialized system F (u) is the same for all values u of the parameters. We propose an algorithm for computing the CTD of F. It relies on a procedure for solving the following set theoretical instance of the coprime factorization problem. Given a family of constructible sets A1,..., As, compute a family B1,..., Bt of pairwise disjoint constructible sets, such that for all 1 ≤ i ≤ s the set Ai writes as a union of some of the B1,..., Bt. We report on an implementation of our algorithm computing CTDs, based on the RegularChains library in MAPLE. We provide comparative benchmarks with MAPLE implementations of related methods for solving parametric polynomial systems. Our results illustrate the good performances of our CTD code. 1

We discuss the parallelization of algorithms for solving polynomial systems symbolically by way of triangular decompositions. We introduce a component-level parallelism for which the number of processors in use depends on the geometry of the solution set of the input system. Our long term goal is to achieve an efficient multi-level parallelism: coarse grained (component) level for tasks computing geometric objects in the solution sets, and medium/fine grained level for polynomial arithmetic such as GCD/resultant computation within each task.

Abstract not found

Symbolic methods are powerful tools in scientific computing. The implementation of symbolic solvers is, however, a highly difficult task due to the extremely high time and space complexity of the problem. In this thesis, we study and apply fast algorithms, modular methods, parallel approaches and software engineering techniques to improve the efficiency of symbolic solvers for computing triangular decomposition, one of the most promising methods for solving non-linear systems of equations symbolically. We first adapt nearly optimal algorithms for polynomial arithmetic over fields to direct products of fields for polynomial multiplication, inversion and GCD compu-tations. Then, by introducing the notion of equiprojectable decomposition, a sharp modular method for triangular decompositions based on Hensel lifting techniques is obtained. Its implementation also brings to the Maple computer algebra system a unique capacity for automatic case discussion and recombination. A high-level categorical parallel framework is developed, written in the Al-dor language, to support high-performance computer algebra on symmetric multi-

Abstract not found

Triangular decompositions are one of the major tools for solving polynomial systems. For systems of algebraic equations, they provide a convenient way to describe complex solutions and a step toward isolation of real roots or decomposition into irreducible components. Combined with other techniques, they are used for these purposes by several computer algebra systems. For systems of partial differential equations, they provide the main practicable way for determining a symbolic description of the solution set. Moreover, thanks to Rosenfeld’s Lemma, techniques from the algebraic case apply to the differential one [3]. Research in this area is following the natural cycle: theory, algorithms, implementation, which will be the main theme of this tutorial. We shall also concentrate on the algebraic case and mention the differential one among the applications.