A scalable sparse matrix-vector multiplication kernel for energy-efficient sparse-blas on FPGAs

Sparse Matrix-Vector Multiplication (SpMxV) is a widely used mathematical operation in many high-performance scientific and engineering applications. In recent years, tuned software libraries for multi-core microprocessors (CPUs) and graphics processing units (GPUs) have become the status quo for computing SpMxV. However, the computational throughput of these libraries for sparse matrices tends to be significantly lower than that of dense matrices, mostly due to the fact that the compression formats required to efficiently store sparse matrices mismatches traditional computing architectures. This paper describes an FPGA-based SpMxV kernel that is scalable to efficiently utilize the available memory bandwidth and computing resources. Benchmarking on a Virtex-5 SX95T FPGA demonstrates an average computational efficiency of 91.85%. The kernel achieves a peak computational efficiency of 99.8%, a >50x improvement over two Intel Core i7 processors (i7-2600 and i7-4770) and showing a >300x improvement over two NVIDA GPUs (GTX 660 and GTX Titan), when running the MKL and cuSPARSE sparse-BLAS libraries, respectively. In addition, the SpMxV FPGA kernel is able to achieve higher performance than its CPU and GPU counterparts, while using only 64 single-precision processing elements, with an overall 38-50x improvement in energy efficiency.

[1]  Junqing Sun,et al.  Mapping Sparse Matrix-Vector Multiplication on FPGAs , 2007 .

[2]  Timothy A. Davis,et al.  The university of Florida sparse matrix collection , 2011, TOMS.

[3]  Phillip H. Jones,et al.  An I/O Bandwidth-Sensitive Sparse Matrix-Vector Multiplication Engine on FPGAs , 2012, IEEE Transactions on Circuits and Systems I: Regular Papers.

[4]  Endong Wang,et al.  Intel Math Kernel Library , 2014 .

[5]  David Gregg,et al.  FPGA Based Sparse Matrix Vector Multiplication using Commodity DRAM Memory , 2007, 2007 International Conference on Field Programmable Logic and Applications.

[6]  Philip Heng Wai Leong,et al.  A Model for Matrix Multiplication Performance on FPGAs , 2011, 2011 21st International Conference on Field Programmable Logic and Applications.

[7]  Georgi Kuzmanov,et al.  Reconfigurable sparse/dense matrix-vector multiplier , 2009, 2009 International Conference on Field-Programmable Technology.

[8]  Samuel Williams,et al.  Optimization of sparse matrix-vector multiplication on emerging multicore platforms , 2007, Proceedings of the 2007 ACM/IEEE Conference on Supercomputing (SC '07).

[9]  Nectarios Koziris,et al.  Understanding the Performance of Sparse Matrix-Vector Multiplication , 2008, 16th Euromicro Conference on Parallel, Distributed and Network-Based Processing (PDP 2008).

[10]  Viktor K. Prasanna,et al.  Sparse Matrix-Vector multiplication on FPGAs , 2005, FPGA '05.

[11]  Michael Garland,et al.  Implementing sparse matrix-vector multiplication on throughput-oriented processors , 2009, Proceedings of the Conference on High Performance Computing Networking, Storage and Analysis.

[12]  Yan Zhang,et al.  FPGA vs. GPU for sparse matrix vector multiply , 2009, 2009 International Conference on Field-Programmable Technology.

[13]  Eric S. Chung,et al.  Towards a Universal FPGA Matrix-Vector Multiplication Architecture , 2012, 2012 IEEE 20th International Symposium on Field-Programmable Custom Computing Machines.