Future Scaling of Memory Hierarchy for Tensor Cores and Eliminating Redundant Shared Memory Traffic Using Inter-Warp Multicasting-Reference-Cited by-同舟云学术

Future Scaling of Memory Hierarchy for Tensor Cores and Eliminating Redundant Shared Memory Traffic Using Inter-Warp Multicasting

Author:

Lee Sunjung¹^ORCID,Hwang Seunghwan¹^ORCID,Kim Michael Jaemin¹^ORCID,Choi Jaewan¹,Ahn Jung Ho¹^ORCID

Affiliation:

1. Department of Intelligence and Information, Seoul National University (SNU), South Korea

Funder

Samsung Advanced Institute of Technology, Samsung Electronics Co., Ltd.

Engineering Research Center Program through the NRF of Korea funded by the Korean Government MSIT

IC Design Education Center (IDEC), Korea

Publisher

Institute of Electrical and Electronics Engineers (IEEE)

Subject

Computational Theory and Mathematics,Hardware and Architecture,Theoretical Computer Science,Software

Link

Reference45 articles.

2. GPU performance analysis and optimization,2012

3. DEVELOPING CUDA KERNELS TO PUSH TENSOR CORES TO THE ABSOLUTE LIMIT ON NVIDIA A100,2020

Cited by 2 articles. 订阅此论文施引文献订阅此论文施引文献，注册后可以免费订阅5篇论文的施引文献，订阅后可以查看论文全部施引文献

2. Accelerating Transformer Networks through Recomposing Softmax Layers;2022 IEEE International Symposium on Workload Characterization (IISWC);2022-11