DNN Dataflow Choice Is Overrated

no code implementations • 10 Sep 2018 • Xuan Yang, Mingyu Gao, Jing Pu, Ankita Nayak, Qiaoyi Liu, Steven Emberton Bell, Jeff Ou Setter, Kaidi Cao, Heonjae Ha, Christos Kozyrakis, Mark Horowitz

Many DNN accelerators have been proposed and built using different microarchitectures and program mappings.

Distributed, Parallel, and Cluster Computing

Programming Heterogeneous Systems from an Image Processing DSL

3 code implementations • 28 Oct 2016 • Jing Pu, Steven Bell, Xuan Yang, Jeff Setter, Stephen Richardson, Jonathan Ragan-Kelley, Mark Horowitz

We address this problem by extending the image processing language, Halide, so users can specify which portions of their applications should become hardware accelerators, and then we provide a compiler that uses this code to automatically create the accelerator along with the "glue" code needed for the user's application to access this hardware.

Software Engineering

A Systematic Approach to Blocking Convolutional Neural Networks

1 code implementation • 14 Jun 2016 • Xuan Yang, Jing Pu, Blaine Burton Rister, Nikhil Bhagdikar, Stephen Richardson, Shahar Kvatinsky, Jonathan Ragan-Kelley, Ardavan Pedram, Mark Horowitz

Convolutional Neural Networks (CNNs) are the state of the art solution for many computer vision problems, and many researchers have explored optimized implementations.

Blocking

EIE: Efficient Inference Engine on Compressed Deep Neural Network

4 code implementations • 4 Feb 2016 • Song Han, Xingyu Liu, Huizi Mao, Jing Pu, Ardavan Pedram, Mark A. Horowitz, William J. Dally

EIE has a processing power of 102GOPS/s working directly on a compressed network, corresponding to 3TOPS/s on an uncompressed network, and processes FC layers of AlexNet at 1. 88x10^4 frames/sec with a power dissipation of only 600mW.