David Patterson
Google Fellow and UC Berkeley professor emeritus; co-author of Google TPU papers and MLPerf Training
- Google Fellow, Google1
- Pardee Professor of Computer Science, Emeritus, University of California, Berkeley21
David Patterson is a Google Fellow and the Pardee Professor of Computer Science, Emeritus, at UC Berkeley, and Google's research profile says Patterson worked at Berkeley for four decades and at Google for one decade. Patterson co-authored the 2017 paper evaluating Google's Tensor Processing Unit in production datacenters and later TPU papers, as well as the MLPerf Training Benchmark paper. Berkeley's faculty page describes Patterson's Berkeley projects, including RISC, RAID, and Networks of Workstations.1234
Work
MLPerf Training BenchmarkPublished research
Co-author of the MLPerf Training Benchmark paper (arXiv 2019; MLSys 2020), which Berkeley's faculty page lists among Patterson's selected publications.42
Tensor Processing Unit (TPU) papers (external site: arxiv.org)Published research
Co-author of In-Datacenter Performance Analysis of a Tensor Processing Unit (2017), which evaluates a custom ASIC deployed in Google datacenters since 2015 for neural network inference, and of later papers on TPU v4 and Google's training supercomputers.321
This profile covers public professional work only, from the sources below. It is not an endorsement or a ranking. To correct or remove it, use the correction form.