{"title":"Polynomial homotopies on multicore workstations","authors":"J. Verschelde, G. Yoffe","doi":"10.1145/1837210.1837230","DOIUrl":"https://doi.org/10.1145/1837210.1837230","url":null,"abstract":"Homotopy continuation methods to solve polynomial systems scale very well on parallel machines. We examine its parallel implementation on multiprocessor multicore workstations using threads. With more cores we speed up pleasingly parallel path tracking jobs. In addition, we compute solutions more accurately in about the same amount of time with threads, and thus achieve quality up. Focusing on polynomial evaluation and linear system solving (key ingredients of Newton's method) we can double the accuracy of the results with the quad doubles of QD-2.3.9 in less than double the time, if all available eight cores are used.","PeriodicalId":123389,"journal":{"name":"International Workshop on Parallel Symbolic Computation","volume":"76 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2010-07-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115875032","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"A complete modular resultant algorithm targeted for realization on graphics hardware","authors":"Pavel Emeliyanenko","doi":"10.1145/1837210.1837219","DOIUrl":"https://doi.org/10.1145/1837210.1837219","url":null,"abstract":"This paper presents a complete modular approach to computing bivariate polynomial resultants on Graphics Processing Units (GPU). Given two polynomials, the algorithm first maps them to a prime field for sufficiently many primes, and then processes each modular image individually. We evaluate each polynomial at several points and compute a set of univariate resultants for each prime in parallel on the GPU. The remaining \"combine\" stage of the algorithm comprising polynomial interpolation and Chinese remaindering is also executed on the graphics processor. The GPU algorithm returns coefficients of the resultant as a set of Mixed Radix (MR) digits. Finally, the large integer coefficients are recovered from the MR representation on the host machine. With the approach of displacement structure [16] and efficient modular arithmetic [8] we have been able to achieve more than 100x speed-up over a CPU-based resultant algorithm from Maple 13.","PeriodicalId":123389,"journal":{"name":"International Workshop on Parallel Symbolic Computation","volume":"6 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2010-07-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115005580","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Parallel computation of the minimal elements of a poset","authors":"C. Leiserson, M. M. Maza, Liyun Li, Yuzhen Xie","doi":"10.1145/1837210.1837221","DOIUrl":"https://doi.org/10.1145/1837210.1837221","url":null,"abstract":"Computing the minimal elements of a partially ordered finite set (poset) is a fundamental problem in combinatorics with numerous applications such as polynomial expression optimization, transversal hypergraph generation and redundant component removal, to name a few. We propose a divide-and-conquer algorithm which is not only cache-oblivious but also can be parallelized free of determinacy races. We have implemented it in Cilk++ targeting multicores. For our test problems of sufficiently large input size our code demonstrates a linear speedup on 32 cores.","PeriodicalId":123389,"journal":{"name":"International Workshop on Parallel Symbolic Computation","volume":"116 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2010-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116268349","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Roomy: a system for space limited computations","authors":"D. Kunkle","doi":"10.1145/1837210.1837216","DOIUrl":"https://doi.org/10.1145/1837210.1837216","url":null,"abstract":"There are numerous examples of problems in symbolic algebra in which the required storage grows far beyond the limitations even of the distributed RAM of a cluster. Often this limitation determines how large a problem one can solve in practice. Roomy provides a minimally invasive system to modify the code for such a computation, in order to use the local disks of a cluster or a SAN as a transparent extension of RAM.\u0000 Roomy is implemented as a C/C++ library. It provides some simple data structures (arrays, unordered lists, and hash tables). Some typical programming constructs that one might employ in Roomy are: map, reduce, duplicate elimination, chain reduction, pair reduction, and breadth-first search. All aspects of parallelism and remote I/O are hidden within the Roomy library.","PeriodicalId":123389,"journal":{"name":"International Workshop on Parallel Symbolic Computation","volume":"15 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2010-06-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123862407","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Exact sparse matrix-vector multiplication on GPU's and multicore architectures","authors":"Brice Boyer, J. Dumas, Pascal Giorgi","doi":"10.1145/1837210.1837224","DOIUrl":"https://doi.org/10.1145/1837210.1837224","url":null,"abstract":"We propose different implementations of the sparse matrix-dense vector multiplication (SpMV) for finite fields and rings Z /m Z. We take advantage of graphic card processors (GPU) and multi-core architectures. Our aim is to improve the speed of SpMV in the LinBox library, and henceforth the speed of its black-box algorithms. Besides, we use this library and a new parallelisation of the sigma-basis algorithm in a parallel block Wiedemann rank implementation over finite fields.","PeriodicalId":123389,"journal":{"name":"International Workshop on Parallel Symbolic Computation","volume":"290 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2010-04-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123109174","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Vincent Danjean, Roland Gillard, S. Guelton, Jean-Louis Roch, Thomas Roche
{"title":"Adaptive loops with kaapi on multicore and grid: applications in symmetric cryptography","authors":"Vincent Danjean, Roland Gillard, S. Guelton, Jean-Louis Roch, Thomas Roche","doi":"10.1145/1278177.1278185","DOIUrl":"https://doi.org/10.1145/1278177.1278185","url":null,"abstract":"The parallelization of two applications in symmetric cryptography is considered: block ciphering and a new method based on random sampling for the selection of basic substitution boxes (S-box) with good algebraic properties. While both consists mainly in loops with independent computations and possibly early termination, they are subject to changing computation loads and processor speeds which can be managed by distributed workstealing. To take benefit of workstealing, we propose in this paper a generic way to rewrite loops in a recursive way, involving three complementary levels of parallelism. Dealing with early termination is performed by an amortized control, original to our knowledge. Those schemes have been embedded in STL-like parallel algorithms implemented on top of Kaapi library that provides distributed workstealing on a wide range of platforms. Experiments and performances are reported on SMP (up to 16 processors) and grid architectures (up to 2120 processors) for benchmarks (e.g. STL find if) and for the two target cryptography applications. These experiments exhibit the stability of the library and its usability by external users for effective applications.","PeriodicalId":123389,"journal":{"name":"International Workshop on Parallel Symbolic Computation","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2007-07-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130482561","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"LinBox and future high performance computer algebra","authors":"B. Char, B. D. Saunders, Bryan S. Youse","doi":"10.1145/1278177.1278197","DOIUrl":"https://doi.org/10.1145/1278177.1278197","url":null,"abstract":"Computer chip design is entering an era in which further increases in computational power will come by increased on-chip parallelism through multi-core architectures rather than by increasing clock speed. If high performance computer algebra tools are to be offered, they must keep pace with this reality. LinBox is a library for exact linear algebra computation with integer matrices and matrices over finite fields. We discuss how LinBox design can be adapted for distributed and multi-core computation.","PeriodicalId":123389,"journal":{"name":"International Workshop on Parallel Symbolic Computation","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2007-07-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130320549","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"A comparative analysis of parallel disk-based Methods for enumerating implicit graphs","authors":"E. Robinson, D. Kunkle, G. Cooperman","doi":"10.1145/1278177.1278190","DOIUrl":"https://doi.org/10.1145/1278177.1278190","url":null,"abstract":"It is only in the last five years that researchers have begun to use disk-based search techniques on a large scale. The primary examples of its use come from symbolic algebra and from artificial intelligence. In the field of parallel search, disk-based search has been forced on researchers because the historical growth in the amount of RAM per CPU core has now stopped. Indeed, the current trend toward multi-core CPUs now threatens to take us backwards. This article makes an original contribution to the design of disk-based parallel search algorithms. It presents a survey of disk-based techniques side-by-side, for the first time. This allows researchers to choose from a menu of techniques, and also to create new hybrid algorithms from the building blocks presented here.","PeriodicalId":123389,"journal":{"name":"International Workshop on Parallel Symbolic Computation","volume":"48 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2007-07-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128384304","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Computing monodromy via parallel homotopy continuation","authors":"A. Leykin, F. Sottile","doi":"10.1145/1278177.1278195","DOIUrl":"https://doi.org/10.1145/1278177.1278195","url":null,"abstract":"Numerical homotopy continuation gives a powerful tool for the applied scientist who seeks solutions to a system of polynomial equations. Techniques from numerical homotopy continuation can also be useful in pure mathematical research. We discuss applications of a particular homotopy continuation idea that leads to probabilistic numerical algorithms for construction of monodromy groups. One such application is used to analyze positive-dimensional solutions of polynomial systems. It is called the monodromy breakup method and partitions a witness set representing a positive-dimensional solution into irreducible components. The first author in collaboration with Jan Verschelde has implemented two parallel versions of this algorithm which show good speedup.\u0000 We use numerical homotopy continuation to compute Galois groups of certain enumerative geometric problems coming from Schubert calculus.The basic idea is similar: given a parametric family of 0-dimensional polynomial systems, we construct loops in the parameter space, follow the solution paths along these loops to obtain a permutation of the set of the solutions.These permutations are used to compute the subgroup of the full symmetric group that they generate.","PeriodicalId":123389,"journal":{"name":"International Workshop on Parallel Symbolic Computation","volume":"63 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2007-07-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127012495","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"KAAPI: A thread scheduling runtime system for data flow computations on cluster of multi-processors","authors":"T. Gautier, Xavier Besseron, L. Pigeon","doi":"10.1145/1278177.1278182","DOIUrl":"https://doi.org/10.1145/1278177.1278182","url":null,"abstract":"The high availability of multiprocessor clusters for computer science seems to be very attractive to the engineer because,at a first level, such computers aggregate high performances. Nevertheless, obtaining peak performances on irregular applications such as computer algebra problems remains a challenging problem. The delay to access memory is non uniform and the irregularity of computations requires to use scheduling algorithms in order to automatically balance the workload among the processors.\u0000 This paper focuses on the runtime support implementation to exploit with great efficiency the computation resources of a multiprocessor cluster. The originality of our approach relies on the implementation of an efficient work-stealing algorithm for a macro data flow computation based on minor extension of POSIX thread interface.","PeriodicalId":123389,"journal":{"name":"International Workshop on Parallel Symbolic Computation","volume":"76 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2007-07-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127537121","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}