2005 · 89 citations · 21 references
EngineeringComputer EngineeringComputer ScienceComputer ArchitectureField-programmable Gate ArraysReconfigurable ComputingField-programmable Gate Array ArchitectureFpga ArchitectureHardware AccelerationMemory Subsystem DesignReconfigurable AcceleratorsReconfigurable System ArchitectureComputational Linear AlgebraData-intensive ComputingFloating-point ArithmeticMatrix MultiplicationReconfigurable Accelerator DesignReconfigurable ArchitecturesReconfigurable Array ArchitectureScientific Computing Acceleration
FPGAs are increasingly used for scientific computing, with vendors offering high‑performance reconfigurable systems that accelerate applications. The paper proposes a BLAS library tailored for state‑of‑the‑art reconfigurable systems. The library implements dot product, matrix‑vector multiply, and dense matrix multiply, using a single‑adder reduction circuit for I/O‑bound operations and a linear FPGA array for matrix multiplication to maximize memory bandwidth utilization. The implementation on a Cray XD1 shows that the designs exploit the memory hierarchy, achieving very low memory bandwidth requirements for Level 2 and Level 3 BLAS.
Field-Programmable Gate Arrays (FPGAs) have become an attractive option for scientific computing. Several vendors have developed high performance reconfigurable systems which employ FPGAs for application acceleration. In this paper, we propose a BLAS (Basic Linear Algebra Subprograms) library for state-of-the-art reconfigurable systems. We study three data-intensive operations: dot product, matrix-vector multiply and dense matrix multiply. The first two operations are I/O bound, and our designs efficiently utilize the available memory bandwidth in the systems. As these operations require accumulation of sequentially delivered floating-point values, we develop a high performance reduction circuit. This circuit uses only one floating-point adder and buffers of moderate size. For matrix multiply operation, we propose a design which employs a linear array of FPGAs. This design exploits the memory hierarchy in the reconfigurable systems, and has very low memory bandwidth requirements. To illustrate our ideas, we have implemented our designs for Level 2 and Level 3 BLAS on Cray XD1.
21
Numerical recipes in C: the art of scientific computing
Choice Reviews Online · 1993 · 18.1K citations · Full text
Numerical recipes in Pascal: the art of scientific computing
Choice Reviews Online · 1990 · 11.9K citations
A set of level 3 basic linear algebra subprograms
Jack Dongarra, Jeremy Du Croz, Sven Hammarling et al. · ACM Transactions on Mathematical Software · 1990 · 1.8K citations · Full text
Basic Linear Algebra Subprograms for Fortran Usage
C. L. Lawson, Richard Hanson, David R. Kincaid et al. · ACM Transactions on Mathematical Software · 1979 · 1.6K citations · Full text
Jiawei Hong, H. T. Kung · 1981 · 450 citations