Efficient Quantized Sparse Matrix Operations on Tensor Cores

Shigang Li, Kazuki Osawa, Torsten Hoefler
Department of Computer Science, ETH Zurich
arXiv:2209.06979 [cs.DC], (14 Sep 2022)




   author={Li, Shigang and Osawa, Kazuki and Hoefler, Torsten},

   keywords={Distributed, Parallel, and Cluster Computing (cs.DC), Machine Learning (cs.LG), FOS: Computer and information sciences, FOS: Computer and information sciences, C.1.4; I.2.11},

   title={Efficient Quantized Sparse Matrix Operations on Tensor Cores},



   copyright={arXiv.org perpetual, non-exclusive license}


Download Download (PDF)   View View   Source Source   Source codes Source codes




The exponentially growing model size drives the continued success of deep learning, but it brings prohibitive computation and memory cost. From the algorithm perspective, model sparsification and quantization have been studied to alleviate the problem. From the architecture perspective, hardware vendors provide Tensor cores for acceleration. However, it is very challenging to gain practical speedups from sparse, low-precision matrix operations on Tensor cores, because of the strict requirements for data layout and lack of support for efficiently manipulating the low-precision integers. We propose Magicube, a high-performance sparse-matrix library for low-precision integers on Tensor cores. Magicube supports SpMM and SDDMM, two major sparse operations in deep learning with mixed precision. Experimental results on an NVIDIA A100 GPU show that Magicube achieves on average 1.44x (up to 2.37x) speedup over the vendor-optimized library for sparse kernels, and 1.43x speedup over the state-of-the-art with a comparable accuracy for end-to-end sparse Transformer inference.
No votes yet.
Please wait...

* * *

* * *

HGPU group © 2010-2024 hgpu.org

All rights belong to the respective authors

Contact us: