Lab Home | Phone | Search
	Center for Nonlinear Studies
	Home  ▶ People   Current   Executive Committee   Postdocs   Visitors   Students   Research   Publications  ▶ Conferences   Workshops   Sponsorship  ▶ Talks   Seminars   Postdoc Seminars Archive   Quantum Lunch   Quantum Lunch Archive   P/T Colloquia   Archive   Ulam Scholar     Postdoc Nominations  ▶ Student Requests   Student Program  ▶ Visitor Requests   Description   Past Visitors  ▶ Services   General     History of CNLS     Maps, Directions   CNLS Office   T-Division   LANL				Monday, December 17, 201210:00 AM - 11:00 AMCNLS Conference Room (TA-3, Bldg 1690) Seminar Hash-based Algorithms for Massively Parallel Architectures Bob Robey and Jonathan RobeyEulerian Applications Group, XCP-2 and We have developed a set of hash-based algorithms with a base operational complexity of O(n) which can replace the typically tree-based O(n log n) algorithms. These hash-based algorithms are also suitable for the finer grain parallelism of the next generation of parallel computing. To take full advantage of this new algorithm, a whole class of mesh-based operations (or in more general terms, discretized data) including sorting, neighbor calculations, remapping and table look-ups are demonstrated. Speed-ups on the CPU over current algorithms can be as high as 50x. These algorithms are also much easier to implement on the GPU and other future architectures than current tree-based algorithms with additional speed-ups of 10x-50x. Host: Mikhail Shashkov, XCP-4 Methods and Algorithms, 667-4400