June 16-20, 2013

Leipzig, Germany

Presentation Details

Name: VLI – A Library for High Precision Integer & Polynomial Arithmetic
Time: Thursday, June 20, 2013
9:00 AM - 9:20 AM
Room:   Hall 5
CCL - Congress Center Leipzig
Speakers:   Timothée Ewart, University of Geneva
Abstract:   We present a high-performance C++ library for high but fixed precision (128 to 512 bit) integer arithmetic and symbolic polynomial computations with assembly language kernels (x86-64, Power64). While the large integer and polynomial computation parts of the library can be used independently optimized kernels for symbolic polynomials with large integer coefficients are provided. We also supply a tunable hybrid CPU/GPU inner product function for vectors of polynomials using OpenMP and NVIDIA CUDA with inline PTX assembly for our main target application: the high-temperature series expansions. We have a maximum speed-up of 38x compared to a GNU Multiple Precision Arithmetic Library (GMP) solution, and we attain 49% of the peak performance of the NVIDIA Kepler.

Paper authors:
Timothee Ewart1, Andreas Hehn2, Matthias Troyer2
1Université de Geneve, Switzerland
2Eidgenössische Technische Hochschule Zürich, Switzerland
Program may be subject to changes.