DS0708 -

Heterogeneous Component Design Technology – HELICITY

HELICITY

The HELICITY industrial research project proposes to develop a new design flow for analog and heterogeneous systems by targeting design space extraction, design automation and behavioral verification of heterogeneous systems.

Objectives

The main objectives are:<br />Objective 1: to develop a complete hierarchical modeling approach as well as a multi-level design flow allowing reuse, architectural exploration and verification at the system level, and based on an industrially accepted modeling language.<br />Objective 2: to extend a graph-based design approach to multi-physical components, and use them for the rapid generation of predictive models for heterogeneous components.<br />Objective 3: to extend model generation approaches based on Pareto edge interpolation to predict both performance indicators and sizing parameters of multi-physical components and continuous-discrete systems over multiple hierarchical levels.<br />Objective 4: To build a hierarchy of data-based performance models for a multi-sensor system, and use it to demonstrate the power of the approach over manual design both in terms of design efficiency and 'in terms of final performance improvement.

WP1 Virtual Machine
WP2 Top-down model design / model generation methods
WP3 Bottom-up model design / model data generation
WP4 Virtual demonstrator

WP1. Virtual machine
Implementation of a clear and well-defined design flow, installed identically on the local IT resources of each partner.

WP3. Bottom-up model design / model data generation
During this period, INL-CNRS and ASYGN focused on accelerating the exploration of design space at the system level, by combining the Pareto fronts extraction method of INL-CNRS with the fast simulator of ASYGN. .
In particular, the partners developed a method to build predictive models of electronic subsystems based on Pareto fronts. It takes advantage of the use of a) Tactyle (ASYGN's proprietary tool) to speed up model simulation, and b) Ant Colony Optimization (ACO) algorithms to implement multi-objective optimization.

WP4. Virtual demonstrator
ASYGN provided the consortium with details (documentation, models, mailing lists and design data) of its existing multi-sensor platform (AS3125 SDK). The electronic subsystems of this platform have been described previously in D411. Deliverable D412 therefore describes the multi-physical subsystems as well as the complete architecture of the platform, which will be used to test the simulation and design tools developed within the framework of the project.

The consortium will work in the rest of the project and in particular during the period M30-M42 on the finalization of the MS2 milestone, dedicated to the generation of predictive models for the subsystems in the architecture of the complete multi-sensor system, and highlighting work of multi-physical components and / or whose nature is described as much by discrete variables as by continuous variables. In particular, the AS3125 SDK platform made available by ASYGN will be the subject of an optimization study carried out by INL-CNRS and simultaneously exploiting discrete and continuous variables. The Midaco-Tactyle coupling set up for D311 will be a crucial element for this work.

- Adil Brik, Lioua Labrak, Laurent Carrel, Ramy Iskander and Ian O'Connor, «High speed extraction of integrated circuits predictive models using n-performance Pareto fronts,« IEEE Conference on Design of Circuits and Integrated Systems, Lyon, 14-16 November 2018
- Adil Brik, «Hierarchical design methodology for analog and heterogeneous systems,« Colloque du GdR SOC2, Montpellier, 19-21 juin 2019
- Adil Brik, Lioua Labrak, Laurent Carrel, Ian O'Connor and Ramy Iskander, «Fast extraction of predictive models for integrated circuits using n-performance Pareto fronts,« IFIP/IEEE International Conference on Very Large Scale Integration (VLSI-SoC), Cusco, Peru, 6-9 October 2019
- Adil Brik, Lioua Labrak, Daniel Saias and Ian O'Connor, «Fast hierarchical system synthesis based on predictive models,« IEEE International NEWCAS Conference, Montreal, 2-4 November 2020

The ambition of this project is to develop novel design methods for More-than-Moore ICs (i.e. micro-nanoelectronics-centered, and including heterogeneous components such as analog/RF circuits and multi-physics sensors and actuators). Such ICs are the root technologies for nanoscale sensor nodes and the next big waves of Internet of Things (IoT) in many sectors (healthcare, transport, security, etc.), but present huge design challenges and increasingly slow time-to-market cycles, as well as severe sub-optimality. New design approaches are mandatory to enhance current "state of the art" design environments (such as schematic and SPICE-like simulators), which scale badly and which will hamper the full deployment of MtM technology for sensor hub integration.
To solve this, we propose a novel analog and heterogeneous system design flow based on industrially focused research on design space extraction, design automation, and behavioral heterogeneous system verification. The principal objectives will be:
Objective 1: to build a comprehensive hierarchical modeling approach and design flow over several levels that enforces reuse, allows architectural exploration and system-level verification, supported by an industry-accepted modeling language.
Objective 2: to extend a graph-based design approach to formalize operators in a more generic way to address multi-physics components, and use them for the fast generation of predictive models for heterogeneous components.
Objective 3: to extend the interpolation of Pareto-front solutions in model generation to predict both performance indicators and sizing parameters for multi-physics components and continuous-discrete systems over many levels of abstraction.
Objective 4: to build a hierarchy of data-assisted performance models for a sensor hub system and to use the models to demonstrate the power of the approach both in terms of design efficiency and in terms of final performance improvement over manual design.
The success of the project will be measured by the speed of design of the sensor hub demonstrator, and the capability of exploring the impact of system-level tradeoffs on device-level constraints. It is stressed that the project will not seek to develop a novel sensor hub architecture design, although this is a possible outcome. The objective of the project is to focus on improving the design process itself to enable faster product time to market and optimized product performance operating at the edge of the performance envelope.
The main outcome of the project will be a new and systematic design method for heterogeneous components, encapsulated in a set of tools and models. We expect the project to lead to the following improvements:
• Five-fold reduction in design time (measured with respect to an existing design)
• Three-fold improvement in design optimality (measured on a single performance metric to be optimized, all other metrics held equal)
• Library of models and data that can be reused in future projects, leading to still easier design of other, optimized, applications.

Project coordination

Ian O'Connor (Institut des Nanotechnologies de Lyon)

The author of this summary is the project coordinator, who is responsible for the content of this summary. The ANR declines any responsibility as for its contents.

Partner

ASYGN
INTENTO DESIGN
INL - CNRS Institut des Nanotechnologies de Lyon

Help of the ANR 508,030 euros
Beginning and duration of the scientific project: September 2016 - 42 Months

Useful links

Explorez notre base de projets financés

 

 

ANR makes available its datasets on funded projects, click here to find more.

Sign up for the latest news:
Subscribe to our newsletter