Spring 2017 — Monday, 12:30pm — CSE503

Subscribe to the calendar: iCal or Google Calendar.

Date	Discussion Leader	Paper
Mar 27	Everyone	Assign Papers
Apr 3	Kendall & Bill	ESE: Efficient Speech Recognition Engine with Compressed LSTM on FPGA
Apr 10	Meghan	Black-box Concurrent Data Structures for NUMA Architectures
Apr 17	Vincent	Automatically Scheduling Halide Image Processing Pipelines
Apr 24	Amrita	Evaluating Programmable Architectures for Imaging and Vision Applications
May 1	Carlo	In-Datacenter Performance Analysis of a Tensor Processing Unit
May 8	Thierry & Emily	Can FPGAs Beat GPUs in Accelerating Next-Generation Deep Neural Networks?
May 15	Sam	TriCheck: Memory Model Verification at the Trisection of Software, Hardware, and ISA
May 22	Eddie	Tetris: Scalable and Efficient Neural Networks with 3D Memory
May 29	-	No meeting (Memorial Day)

Logistics

• Traditional paper discussions in addition to discussions of perspectives or visionary talks from people in our field
• Using the entire slide deck or most of the slide deck used by the original authors verbatim is prohibited
• It is okay to borrow some good slides from the original presentation but they should mostly be limited to diagrams, plots, etc.
• 2 original slides for discussion

Paper Suggestions

• Can FPGAs Beat GPUs in Accelerating Next-Generation Deep Neural Networks? - Appears in FPGA '17 (Intel's Position Paper)
• ESE: Efficient Speech Recognition Engine with Compressed LSTM on FPGA - Appears in FPGA '17 (Best Paper Award) (Stanford, the EIE people)
• Accelerating Binarized Convolutional Neural Networks with Software-Programmable FPGAs - Appears in FPGA '17
• FINN: A Framework for Fast, Scalable Binarized Neural Network Inference - Appears in FPGA '17
• Proprietary versus Open Instruction Sets - Appears in IEEE Micro '17
• Framework for Quantifying and Managing Accuracy in Stochastic Computing - Appears in DATE '17
• Automatically Scheduling Halide Image Processing Pipelines - Appears in SIGGRAPH '16
• The Rise of Mobile Visual Computing Systems - Appears in IEEE Pervasive Computing '16
• Rigel: Flexible Multi-rate Image Processing Hardware - Appears in SIGGRAPH '16
• Tetris: Scalable and Efficient Neural Networks with 3D Memory - Appears in ASPLOS '17 (Mark Horowitz et al.)
• Moonwalk: NRE Optimization in ASIC Clouds - Appears in ASPLOS '17 (Michael Taylor et al.)
• TriCheck: Memory Model Verification at the Trisection of Software, Hardware, and ISA (ASPLOS 2017)
• Determining Application-specific Peak Power and Energy Requirements for Ultra-low-power Processors, Hari Cherupalli (University of Minnesota); Henry Duwe (University of Illinois); Weidong Ye (University of Illinois); Rakesh Kumar (University of Illinois); John Sartori (University of Minnesota)
• Prophet: Precise QoS Prediction on Non-Preemptive Accelerators to Improve Utilization in Warehouse Scale Computers, Quan chen (Shanghai Jiao Tong University); Hailong Yang (Beihang University); Minyi Guo (Shanghai Jiao Tong University); Ram Srivatsa Kannan (University of Michigan - Ann Arbor); Jason Mars; Lingjia Tang (University of Michigan - Ann Arbor)
• Mallacc: Accelerating Memory Allocation, Svilen Kanev; Sam Likun Xi; Gu-Yeon Wei; David Brooks (Harvard University)
• REDSPY: Exploring Value Locality in Software, Shasha Wen (College of William and Mary); Milind Chabbi (Hewlett-Packard Enterprise); Xu Liu (College of William and Mary)
• Translation-Triggered Prefetching, Abhishek Bhattacharjee (Rutgers University)
• Typed Architectures: Architectural Support for Lightweight Scripting, Channoh Kim (Sungkyunkwan University); Jaehyeok Kim (Sungkyunkwan University); Sungmin Kim (Sungkyunkwan University); Dooyoung Kim (Sungkyunkwan University); Namho Kim (Seoul National University); Gitae Na (Sungkyunkwan University); Young H. Oh (Sungkyunkwan University); Hyeon Gyu Cho (Sungkyunkwan University); Jae W. Lee (Seoul National University)
• Black-box Concurrent Data Structures for NUMA Architectures, Irina Calciu (VMware Research Group); Siddhartha Sen (Microsoft Research); Mahesh Balakrishnan (Yale University); Marcos K. Auguilera (VMware Research Group)