CORDIS - EU research results

Architecting More Than Moore – Wireless Plasticity for Heterogeneous Massive Computer Architectures

Periodic Reporting for period 1 - WiPLASH (Architecting More Than Moore – Wireless Plasticity for Heterogeneous Massive Computer Architectures)

Reporting period: 2019-10-01 to 2020-09-30

The WiPLASH project focuses on the grand challenge of introducing diversification and specialization in heterogeneous processor architectures, while ensuring their generality and scalability. This problem is stalling the progress of computing systems all-pervasive in our society: those processing BigData, executing AI, or driving scientific computations in biology, finance, or complex systems. In this context, the WiPLASH project aims to take a radical step further in computing by designing a new breed of massive heterogeneous architectures at extreme scales. In particular, the long-term vision of WiPLASH is the realization of massive heterogeneous architectures with accelerator-like performance and without loss of purpose generality. Towards this end, WiPLASH will pioneer an on-chip wireless communication based on graphene terahertz nano-antennas able to provide architectural plasticity, reconfigurability and adaptation to the application requirements. The proposed breakthrough of the project is the experimental validation of the graphene-based tunable antennas, as well as their technological and architectural integration. To this end, the project aims to achieve three specific objectives:
- Prototype a miniaturized and tunable graphene antenna in the terahertz band
- Co-integrate graphene RF components with submillimeter-wave transceivers
- Demonstrate low-power reconfigurable wireless chip-scale networks

The culminating goal is to demonstrate that the wireless plane offers the plasticity required by future computing platforms by speeding up at least one key application by 10X over a state-of-the-art baseline. This will demonstrate that our approach can pave the way for a new generation of scalable, massively parallel biologically plausible AI processors. The project is expected to have a disruptive impact in the multi-billion-euro computer processor market, where Europe has traditionally been a secondary player. Given its disruptive nature, the wireless plasticity vision delivers competitive advantages to new and high-potential actors (young researchers, deep-tech SMEs) in a field that, traditionally, exhibits large barriers to entry. In particular, WiPLASH aims to prove the potential of its approach by developing a new accelerator for AI processing, with relevant applications in neuroscience, drug discovery, disease detection, among many other examples.
Towards its overarching and specific goals, the scientific work packages of WiPLASH focused on the following key actions during the first twelve months of execution:
1) Establishing the system specifications and the measurement setup. The specifications were decided based on current and future transceiver capabilities, as well as communication requirements from the architecture.
2) Prototyping the antenna. Experimental measurements are being performed on batches of metallic and graphene antennas. Theoretical analysis and simulations were used to anticipate the working point of the antennas depending on the graphene characteristics and antenna size, paving the way for a working antenna.
3) Modeling the wireless channel within the computing package. Simulations in the range of 60-140 GHz have been performed assuming a flip-chip package, and these are currently being extended to 240 GHz and extended to other chip packages. This allows us to perform accurate link budget analysis and other communication-related tasks.
4) Advancing in the development of heterogeneous architectures by enabling the integration in-memory cores into the design flows and simulation methods. This involves augmenting the full-system simulator of WiPLASH with ways to simulate these in-memory computing cores.
WiPLASH has already gone beyond the state of the art in a few areas as a result of the on-going efforts. More specifically, it:
- Laid down the multi-chip, wireless-enabled vision of massive heterogeneous architectures with system-level recofigurability that graphene antennas can provide. The vision goes beyond the state of the art in the field of wireless chip-scale networks that either consider single-chip architectures, lower frequencies, unrealistic specifications at the antenna and transceiver. Moreover, existing architectures do not exploit the system-level flexibility of wireless interconnects, as we aim to do in WiPLASH.
- Provided a methodology for the design, fabrication, transfer, and measurement of graphene antennas with photoconductive sources. While the prototype has not been experimentally validated yet, these steps imply going beyond the state of the art in the field because graphene antennas have been explored theoretically, but not measured experimentally.
- Studied the wireless channel at subTHz frequencies in realistic computer chip packages. These results suppose an advance with respect to the state of the art because existing channel models assume unpackaged chips or incorrect positioning of the antennas. For the first time, models with wire-bonding and interposer-based packages are studied.
- Advanced in the integration of in-memory computing cores and wireless interconnects into full-system simulators. Separately, those items may be modeled (partially) in specialized simulators, but never put together in the context of a full-system simulators or design space exploration tools. Therefore, WiPLASH progresses beyond existing approaches and enables the exploration of the architectural design space opened by the wireless technology, which was not possible to date.

During the rest of the project, we expect to:
- Provide a first graphene antenna prototype, demonstrating terahertz radiation from a photoconductive source.
- Advance in the potential integration of working graphene antennas into a complete circuit containing the necessary elements to generate and modulate the signals that will be fed the antenna.
- Design a full protocol stack of wireless communications, containing new hand-tuned protocols at the physical layer, link layer, and network layer.
- Develop a new wireless-enabled heterogeneous architecture capable of accelerating a variety of AI workloads by at least 10X with respect to conventional architectures.
- Deliver a full simulation tool integrating heterogeneous cores and wireless interconnect models, enabling full-system simulations with both elements, and extensive design space explorations towards determining how to best exploit the capabilities of wireless in multiple applications.

The results of this project are expected to have a disruptive impact in the multi-billion-euro computer processor market, where Europe has traditionally been a secondary player. In this context, WiPLASH represents a window of opportunity to take the lead with a disruptive approach that marks a big departure from more conventional roadmaps and help ensure the independence and sovereignity of Europe in this field. Spin-offs or SMEs can be created not only focusing on the creation of adaptive AI processors for multiple application areas, but also exploiting any of the partial results of the project. From the architectural standpoint, new flexible processor architectures can reach beyond AI, opening the door to faster and more efficient processing for applicaitons such as BigData, genomics, finance, or the automotive industry.