| A | B | C | D | E | F | |
|---|---|---|---|---|---|---|
1 | Section | Time (EST) | Title | Speaker | Session Chair | |
2 | Opening | 9:00AM - 9:10AM | Opening and Announcements | Amir Yazdanbakhsh (Google) | ||
3 | Keynote Speaker | 9:10AM - 9:40AM | Full-Stack Deep Learning | Kurt Keutzer (Berkeley) | Amir Yazdanbakhsh (Google) | |
4 | Best Paper Nominees | 9:40AM - 9:55AM | Towards GPU-Driven Code Execution for Distributed Deep Learning | Changho Hwang | Amir Yazdanbakhsh (Google) | |
5 | 9:55AM - 10:10AM | Autophase V2: Towards Function Level Phase Ordering Optimization | Mohammed Almakki / Ayman Izzeldin | |||
6 | 10:10AM - 10:15AM | Best Paper Announcement | N/A | |||
7 | 10:15AM - 10:45AM | Break | ||||
8 | Invited Talks | 10:45AM -11:10AM | Learning Architecture and Mapping Space for DNN Accelerators | Ananda Samajdar (IBM) | Suvinay Subramanian (Google) | |
9 | 11:10AM - 11:35AM | Learning A Continuous and Reconstructible Latent Space for Hardware Accelerator Design | Qijing Huang (Nvidia) | |||
10 | Keynote Speaker | 11:35AM - 12:00PM | Learning Semantic Representations of Hardware Designs | Shobha Vasudevan (Google) | ||
11 | 12:00PM - 1:00PM | Lunch | ||||
12 | Keynote Speaker | 1:00PM - 1:30PM | Repairing Code with Deep Learning | Miltos Allamanis | TBD | |
13 | Invited Talk | 1:30PM - 2:00PM | LoopStack: a Lightweight Tensor Algebra Compiler Stack | Bram Wasti (Meta) & Aleks Zlateski (Meta) | Chris Cummins (Meta) | |
14 | Keynote Speaker | 2:00PM - 2:30PM | Tensor Query Processing: How to Leverage AI Investments to Accelerate Databases, Classical ML Inference, and More | Carlo Curino (Microsoft) | Neeraja Yadwadkar (UT Austin) | |
15 | 2:30PM - 2:45PM | Break | ||||
16 | Keynote Speaker | 2:45PM - 3:15PM | Using Machine Learning to Automate Compiler Design | Zheng Wang (University of Leeds) | Chris Cummins (Meta) | |
17 | Scheduling and Performance Optimization | 3:15PM - 3:30PM | Exploring Efficient ML-based Scheduler for Microservices in Heterogenous Clusters | Rohan Mahapatra | Chris Cummins (Meta) | |
18 | 3:30PM - 3:45PM | DiffTune Revisited: A Simple Baseline for Evaluating Learned llvm-mca Parameters | Andreas Abel | |||
19 | 3:45AM - 4:00PM | Unveiling Source of Performance Variance on Searching-based Compiler Optimization | Pranati Modumudi | |||
20 | 4:00PM - 4:05PM | Short Break Before Panel | ||||
21 | Panel Discussion | 4:05PM - 4:55PM | ML for Computer Architecture and Systems Panel Discussion | Kurt Keutzer (Berkeley) / Cliff Young (Google Research) / Carlo Curino (Microsoft) / Martin Maas (Google Research) | Chris Cummins / Amir Yazdanbakhsh / TBD | |
22 | Closing Remarks | 4:55PM - 5:00PM | Closing Remarks | Amir Yazdanbakhsh (Google Research) | ||