Towards massively parallel computations in algebraic geometry (Q2040457)

Finding procedures in Algebraic Geometry that can be implemented in parallel computing is non-trivial, because some of the most popular algorithms for computations in polynomial rings, as those based on the determination of a Gröbner basis, have an intrinsic sequential nature. The authors describe some algebraic geometric questions that can be handled with parallel computing. The main problem that they address is the construction of a test for detecting the smoothness of algebraic varieties \(X\), following Hironaka's desingularisation approach which is based on the study of hypersurfaces of local maximal contact. The test splits \(X\) in a finite union of affine charts \(X_i\)'s embedded, with low codimension, in complete intersection smooth varieties. Then the test applies Hironaka's termination criterion and the Jacobian criterion to detect the smoothness of each \(X_i\). The algorithm has been implemented under the workflow management system GPI-Space, which controls the execution of many copies of the computer algebra system SINGULAR, for the study of the charts \(X_i\)'s. The authors show that their method works in some cases of surfaces of general type, too hard to be handled by a sequential approach.