Skip to main content

Technology and hardware for neuromorphic computing

Periodic Reporting for period 1 - TEMPO (Technology and hardware for neuromorphic computing)

Reporting period: 2019-05-01 to 2020-04-30

Massive adoption of computing in all aspects of human activity has led to unprecedented growth in the amount of data generated. Machine learning has been employed to classify and infer patterns from this abundance of raw data, at various levels of abstraction. Among the algorithms used, brain-inspired, or “neuromorphic”, computation provides a wide range of classification and/or prediction tools. Additionally, certain implementations come about with a significant promise of energy efficiency: ranging from highly optimized Deep Learning engines to sparse, event based and efficient Spiking Neural Networks (SNN). Given the slowdown of silicon-only scaling, it is important to extend the roadmap of neuromorphic implementations by leveraging fitting technology innovations. Along these lines, the current project aims to sweep technology options, covering emerging memories and 3D integration, and attempt to pair them with contemporary Deep Learning (DL) and exploratory (SNN) neuromorphic computing paradigms.
In the first project period, the project has focussed on the design of test vehicles and process flows in the RTOs to enable the process technology pathfinding work later in the project to optimally leverage embedded Non-Volatile Memories for Neuromorphic and AI applications and on the design of core building blocks and accelerator architectures targeted to leverage the memory technologies in application demonstrators. Basic neuromorphic building blocks were investigated with a focus on the development of neuromorphic –ready NVM blocks, the modelling and simulation of eNVM, the quantification of the technology features and neuromorphic implementation of eNVM. Additionally, features of embedded memory for Neuromorphic Accelerators, have been addressed, such as multi-level memory, synapticity/plasticity of the memories. 3D specs suited for DNN accelerators have been defined and a design flow to be able to quantify performance and energy impact of 3D interconnect has been set-up. Design and architecture exploration, specification and design of critical building blocks to enable full accelerator IP blocks later in the project has been done.
The consortium will work further on these results to enable demonstration of energy efficient accelerators for the different use cases defined in the project. Willing to address the needs of end-users applicative sectors (aviation, automotive, etc.), the TEMPO project has integrated the main European actors of each sector to participate to the specification of needs data set definition. This allows the TEMPO partners to complement each other in a near-optimal way so as to provide Europe with a substantiated competitive advantage and a faster time-to-market opportunity in roll-out of the technology roadmap of neuromorphic implementations throughout the different sectors involved.