2004 · 34 citations · 6 references
EngineeringComputer EngineeringComputer ScienceParallel ComputingReconfigurable ComputingScientific ComputingData ParallelismForce FieldsHardware AccelerationMolecular DynamicsMolecular SimulationReconfigurable AcceleratorsAccelerated Molecular DynamicsField-programmable Gate Array AccelerationField-programmable Gate Array AcceleratorsHigh PerformanceLennard-jones PotentialMolecular Dynamics Force FieldsMolecular Simulation AlgorithmsReconfigurable Hardware Architecture
Abstract — Technological advances have made FPGAs an attractive platform for the acceleration of complex scientific applications. These applications demand high performance and highprecision floating point arithmetic. In this paper, we present a design for calculating the Lennard-Jones potential and force as is done in molecular dynamics simulations. This architecture employs IEEE 754 double precision floating point units, including a square root unit developed for this kernel. The design presented is a modular, very deeply pipelined architecture that exploits the fine-grained parallelism of the calculations. With the Xilinx Virtex-II Pro as a target device, an implementation using two pipelines operating in parallel achieves 3.9 GFLOPS. Index Terms — molecular dynamics, square root, supercomputing I.
6
Analysis of high-performance floating-point arithmetic on FPGAs
Gokul Govindu, Ling Zhuo, Seonil Choi et al. · 2004 · 138 citations
Makoto Taiji, Tetsu Narumi, Yousuke Ohno et al. · 2003 · 79 citations
Natural Sciences, High-performance Computing, Cluster Computing +7