The goal of the workshop is to bring together researchers from multiple areas of expertise with an interest in developing hardware, software, and theory for trading or exploiting errors of various kinds for improved efficiency in computing systems. The workshop will explore both theory and hardware proposals for such platforms. The workshop will provide an opportunity to explore what it takes to demonstrate ideas in real systems, while fostering common terminology, software tools, and hardware evaluation and measurement tools.
To achieve the aims of the workshop, we have structured the workshop activities around five themes: Theme 1: Techniques that can be demonstrated today; Theme 2: Techniques that should be demonstrable today, but require hardware or software engineering effort; Theme 3: Techniques that require fundamentally new theoretical results and hardware technologies; Theme 4: Complete platform studies; Theme 5: Fundamental upper bounds on potential benefit of error-efficiency.
We expect the workshop to have impact in four ways: Impact 1: Common terminology for work on theory; Impact 2: Design objectives and guidelines for common hardware evaluation techniques; Impact 3: Identification of grand challenges across the hardware, software, and application stack for error-efficient computing systems; Impact 4: Target research problems identified by participants in each workshop theme.
The workshop will use a set of hardware kits as a initial common baseline for hands-on discussions during the workshop. Each kit consists of: Hardware kit item 1: One ARM Cortex-M0+ microcontroller evaluation board; Hardware kit item 2: one Lattice iCE40 FPGA evaluation board; Hardware kit item 3: One 96x64 pixel OLED display; Hardware kit item 4: One current measurement board; Other hardware kit items: USB cable, one mini breadboard, breadboard wires.
Time | Agenda | Room |
---|---|---|
08:30 – 09:30 | Welcome, workshop introduction, and participant introductions (lightning talks) | Salle Symphonie |
09:30 – 10:30 | Tutorial 1 Computation, Errors, and Efficiency: Terminology, History, and Bridges Phillip Stanley-Marbell, University of Cambridge, UK | Salle Symphonie |
10:30 – 11:00 | Coffee Break | Salle Symphonie |
11:00 – 12:30 | Parallel working sessions for themes 1 – 5 |
Theme 1 and 2: Parchet Theme 3 and 5: Lavaux A Theme 4: Lavaux B |
12:30 – 14:00 | Lunch | |
14:00 – 15:30 | Open Workshop Discussion Complete Platform Studies: Do We Need an End-to-End Demonstrator? |
Salle Symphonie |
15:30 – 16:00 | Coffee Break | Salle Symphonie |
16:00 – 17:30 |
Tutorial 2 Evolution of Stochastic Computing Armin Alaghi, University of Washington, USA |
Salle Symphonie |
17:30 – 18:30 | Theme discussion updates on themes 1 - 5 (status and goals) | Salle Symphonie |
18:30 – | Dinner |
Time | Agenda | Room |
---|---|---|
09:00 – 10:30 | Tutorial 3 Examining hardware-level knobs and their leverage for approximate computing Naveen Verma, Princeton University, USA |
Salle Symphonie |
10:30 – 11:00 | Coffee Break | Salle Symphonie |
11:00 – 12:30 | Parallel working sessions for themes 1 – 5 |
Theme 1 and 2: Parchet Theme 3 and 5: Lavaux A Theme 4: Lavaux B |
12:30 – 14:00 | Lunch | |
14:00 – 15:30 | Tutorial 4 (Topic: Type Systems and Programming Languages for Approximate Computing) Sasa Misailovic, University of Illinois at Urbana Champaign, USA Michael Carbin, Massachusetts Institute of Technology, USA |
Salle Symphonie |
15:30 – 16:00 | Coffee Break | Salle Symphonie |
16:00 – 17:30 | Tutorial 5 Hardware-level Challenges for Approximate Computing Andreas Gerstlauer, University of Texas at Austin, USA |
Salle Symphonie |
17:30 – 18:30 | Theme discussion updates on Themes 1 - 5 (status and updated goals) | Salle Symphonie |
18:30 – | Dinner |
Time | Agenda | Room |
---|---|---|
09:00 – 10:30 | Tutorial 6 Modern Coding Theory for Error-Efficient Computing Lara Dolecek, University of California at Los Angeles, USA |
Salle Symphonie |
10:30 – 11:00 | Coffee Break | Salle Symphonie |
11:00 – 12:30 | Parallel working sessions for themes 1 – 5 |
Theme 1 and 2: Parchet Theme 3 and 5: Lavaux A Theme 4: Lavaux B |
12:30 – 14:00 | Lunch | |
14:00 – 15:30 | Open Workshop Discussion: Roadblocks from Theory to Practice | Salle Symphonie |
15:30 – 16:00 | Coffee Break | Salle Symphonie |
16:00 – 17:30 | Tutorial 7 ACCEPT, an Open-Source Research Infrastructure for Approximate Computing Adrian Sampson, Cornell University, USA |
Salle Symphonie |
17:30 – 18:30 | Theme discussion updates on Themes 1 - 5 (preliminary reports) | Salle Symphonie |
19:00 – 19:30 | Meeting in front of the hotel for taxi transfer to restaurant for dinner | Front of hotel |
Time | Agenda | Room |
---|---|---|
09:00 – 10:30 | Tutorial 8 Displays, Perception, and Approximation Phillip Stanley-Marbell, University of Cambridge, UK |
Salle Symphonie |
10:30 – 11:00 | Coffee Break | Salle Symphonie |
11:00 – 12:30 | Parallel working sessions for themes 1 – 5 |
Theme 1 and 2: Parchet Theme 3 and 5: Lavaux A Theme 4: Lavaux B |
12:30 – 14:00 | Lunch | |
14:00 – 15:30 | Presentations on workshop themes 1 and 2 | Salle Symphonie |
15:30 – 16:00 | Coffee Break | Salle Symphonie |
16:00 – 18:30 | Presentations on worksop themes 3, 4 and 5 | Salle Symphonie |
18:30 – 19:00 | Summary of workshop insights and closing remarks | Salle Symphonie |
Sara Achour | MIT |
Armin Alaghi | University of Washington |
Mattia Cacciotti | EPFL |
Michael Carbin | MIT |
Alex Daglis | EPFL |
Eva Darulova | MPI-SWS |
Lara Dolecek | UCLA |
Mario Drumond | EPFL |
Natalie Enright Jerger | University of Toronto |
Babak Falsafi | EPFL |
Andreas Gerstlauer | The University of Texas at Austin |
Ghayoor Gillani | University of Twente |
Djordje Jevdjic | University of Washington |
Sasa Misailovic | UIUC |
Thierry Moreau | University of Washington |
Adrian Sampson | Cornell University |
Phillip Stanley-Marbell | University of Cambridge |
Radha Venkatagiri | UIUC |
Naveen Verma | Princeton University |
Marina Zapater | EPFL |
Damien Zufferey | MPI-SWS |