Keynotes
Industrializing an asynchronous technology: Tiempo from 2007 to 2025
by Marc Renaudin (Qualcomm Technologies Inc.)
This presentation traces the history of Tiempo, from its founding in 2007 to its acquisition by Qualcomm in January 2025. It highlights the major phases of the company’s development, the evolution of its business model, as well as the target markets, products, and their adoption by the industry. At the core of Tiempo’s success is an innovative hardware technology based on asynchronous circuits. This technology combines ASC, a dedicated asynchronous standard‑cell library complemented by a synchronous one, with the ACC synthesis tool, which maps SystemVerilog designs onto logic gates. Integrated into a complete design flow, these technologies enabled the development of Common Criteria–certified programmable platforms for smart cards and security enclaves, including both hardware and software IP.

Towards Local Learning and Asynchronous Silicon: A Co-Design Strategy for Neuromorphic Edge AI by Federico Corradi (Technical University of Eindhoven)
Deep Spiking Neural Networks (SNNs) are limited by the high memory costs and global dependencies of standard backpropagation. We present a transition to local activity-based learning via Forward Propagation Through Time (FPTT) and Trace Propagation (TP), which reduce memory complexity to linear O(H * L) by eliminating temporal unrolling and spatial weight transport. While these algorithms theoretically enable efficient edge intelligence, their transition to silicon raises a fundamental challenge: at what point does the super-linear overhead of global clock distribution and activity gating become a “power wall” that asynchronous logic could breach? What are neuromorphic architectures that exploit asynchronous and synchronous logic for edge AI? To address these questions, we will examine state-of-the-art neuromorphic architectures across digital and mixed-signal domains. We will discuss the inherent benefits and drawbacks of mapping sparse, event-driven algorithms to both clocked and asynchronous substrates. Finally, we will explore the critical bottlenecks, such as hardware rigidity, process variations, and toolchain friction, to understand why continuous, on-chip learning is not yet fully exploited at the extreme edge, and outline the necessary steps to overcome these barriers.

Program
1-2 June Summer School
3. June
| Time | Event |
|---|---|
| 08:00 | Registration & breakfast snacks |
| 09:00 | Welcome |
| 09:15 | Session: Asynchronous circuit Synthesis (1h30)
|
| 10:45 | Break (coffee & snacks) |
| 11:10 | Keynote Industrializing an asynchronous technology: Tiempo from 2007 to 2025 by Marc Renaudin |
| 12:10 | Lunch in the Faculty Club |
| 13:10 | Session: Test of the asynchronous circuits (1h50)
|
| 15:00 | Break (coffee & snacks) |
| 15:30 | Demo Session (1h)
|
| 16:30 | Head to Copenhagen Center & short optional city walk |
| 19:00 | Conference Dinner |
| Time | Event |
|---|---|
| 08:00 | Breakfast snacks & coffee |
| 09:20 | Fresh idea session (1h)
|
| 10:20 | Break (coffee & snacks) |
| 10:40 | Session: Asynchronous circuit applications and models (1h30)
|
| 12:10 | Lunch in the Faculty Club |
| 13:10 | Keynote 2 by Federico Corradi |
| 14:10 | Short break |
| 14:25 | Panel: “Open challenges for asynchronous digital integrated circuits in neuronal network hardware”
Moderator: Ole Richter |
| 15:10 | Break |
| 15:30 | Session: Efficient bundled-data and NEMS circuits
|
| 16:30 | Closing session and Best Paper Award |
| 16:50 | Move to Copenhagen Center – visit Streetfood market Reffen (alternative program for bad weather) |
| ~ 9:15 till 14:00 (Kopenhagen) | Trip to Viking Museum In Roskilde, with tour (https://www.vikingeskibsmuseet.dk/en) |
