SPADES: A 0.54-GFLOPS/W Sparse Matrix Vector Multiplication Accelerator Featuring On-the-Fly GZIP Decompression for 3.36X Reduction in Off-Chip Data Movement
Open Access
- 5 August 2024
- conference paper
- Published by Association for Computing Machinery (ACM)
- p. 1-6
Abstract
We propose SPADES, the first sparse matrix vector multiplication (SpMV) accelerator incorporating online GZIP decompression. The goal of the decompression is to reduce off-chip data movement, which has become a major source of energy consumption in SpMV accelerators. Our proposed GZIP-SpMV flow achieves an average compression ratio of 3.36 for the sparse matrix CSC data, reducing the off-chip data traffic by 70%. We fabricated SPADES in TSMC 28-nm with a die area of 2.47 mm2. Compared with the prior SpMV accelerator, SPADES achieves 2.32X better energy efficiency.Keywords
Funding Information
- NSF (National Science Foundation) (1840763)
This publication has 9 references indexed in Scilit:
- Memory Bandwidth Conservation for SpMV Kernels Through Adaptive Lossy Data Compression2023
- Data Compression Accelerator on IBM POWER9 and z15 Processors : Industrial Product2020
- ALRESCHA: A Lightweight Reconfigurable Sparse-Computation Accelerator2020
- A 1.4GHz 20.5Gbps GZIP decompression accelerator in 14nm CMOS featuring dual-path out-of-order speculative Huffman decoder and multi-write enabled register file array2019
- 34.4Mbps 1.56Tbps/W DEFLATE Decompression Accelerator Featuring Block-Adaptive Huffman Decoder in 14nm Tri-Gate CMOS for IoT Platforms2018
- A 190GFLOPS/W DSP for energy-efficient sparse-BLAS in embedded IoT2016
- Accelerating SpMV on FPGAs by Compressing Nonzero Values2015
- 1.1 Computing's energy problem (and what we can do about it)2014
- The university of Florida sparse matrix collectionACM Transactions on Mathematical Software, 2011