High-Performance Designs for Linear Algebra Operations on Reconfigurable Hardware

Abstract

Abstract-Numerical linear algebra operations are key primitives in scientific computing. Performance optimizations of such operations have been extensively investigated. With the rapid advances in technology, hardware acceleration of linear algebra applications using field-programmable gate arrays (FPGAs) has become feasible. In this paper, we propose FPGA-based designs for several basic linear algebra operations, including dot product, matrix-vector multiplication, matrix multiplication, and matrix factorization. By identifying the parameters for each operation, we analyze the trade-offs and propose a high-performance design. In the implementations of the designs, the values of the parameters are determined according to the hardware constraints, such as the available chip area, the size of available memory, the memory bandwidth, and the number of I/O pins. The proposed designs are implemented on Xilinx Virtex-II Pro FPGAs. Experimental results show that our designs scale with the available hardware resources. Also, the performance of our designs compares favorably with that of general-purpose processor-based designs. We also show that, with faster floating-point units and larger devices, the performance of our designs increases accordingly.

The Implementation of Texture-Based Video Up-Scaling on Coarse-Grained Reconfigurable Architecture

“…[21] proposes a new architecture of bicubic interpolation and implemented it on FPGA. [22] proposes a linear algebra implementations on FPGA, the authors utilizes the overlap technique between I/O and execution time to increase computing speed. [23] shows the procedure of mapping a Jacobi Iterative Solver on FPGA.…”

FPGA accelerator for floating-point matrix multiplication

“…Although the PE connection pattern in the form of a tree is also possible [13], the linear list has the advantage of a much more regular structure, which allows simpler routing between PEs and consequently the higher clock frequency. After the initial latency, a list of n PEs multiply two nelement vectors in one clock cycle, or two square matrices of order n in n …”

Scalable Hybrid Designs for Linear Algebra on Reconfigurable Computing Systems

“…The main focus of that work was to examine the potential capacity of FPGAs in performing BLAS operations. The only work that has implemented linear algebra applications on the reconfigurable computing systems is [21]. However, it only employs the FPGAs in the systems.…”