Politecnico di Torino | Servizi per la didattica

KEYWORD

DAUIN - GR-06 - ELECTRONIC DESIGN AUTOMATION - EDA

RISC-V for automotive

keywords AUTOMOTIVE, ENERGY, RISC-V, SIMULATION

Reference persons ALESSIO BURRELLO, DANIELE JAHIER PAGLIARI, MASSIMO PONCINO, SARA VINCO

Research Groups DAUIN - GR-06 - ELECTRONIC DESIGN AUTOMATION - EDA

Thesis type RESEARCH

Description Over the last 50 years, we’ve seen incredible changes in the automotive industry. Who would have thought that our cars could run without gas? That they could park themselves? That we could talk to them (and they would reply)? The transformation we will see over the next decade will bring even more incredible changes as the industry moves towards automobiles that are more intelligent, capable, safe and sustainable.
To create our increasingly “software defined” automobiles that have sophisticated capabilities and a growing number of services, automotive designers today must build powerful, sophisticated computing platforms, and support the design with a monitoring of the interaction between electronic/software aspects with physical aspects like power consumption and management, thermal modeling, reliability measures, etc.
This evolution is accelerating the adoption of RISC-V in automotive. With RISC-V, designers get a flexible, modern architecture supported by an increasingly broad ecosystem. Designers can leverage RISC-V in creating Systems-on-Chips (SoCs) for automotive applications that not only meet performance/cost/power requirements, but also have a high degree of code portability.
It is however necessary to extend RISC-V frameworks with an enhanced support for extra-functional aspects, like mechanical aspects, power distribution, thermal modeling, etc. To achieve this goal, it is possible to leverage C-based RISC-V simulators, like GVSoC, and to integrate them with simulators that cover not only hardware and software aspects, but also physical modeling and, more in general, extra-functional modeling. In this scenario, SystemC-AMS is a very powerful framework: it natively supports heterogeneous systems and continuous time/mixed signal modeling, still being C-based.
The thesis involves the development of an automotive simulator that includes functional aspects, energy-related aspects, like consumption and energy production/management, and mechanical aspects:
• the functional aspects rely on a RISC-V simulator implemented in C (GVSoC), which can run the drone control software;
• the mechanical model includes various physical dynamics;
• the power model includes consumption estimation and modeling of the energy distribution system.
The simulation infrastructure is C-based, as the RISC-V simulator is implemented in C and both extra-functional models will be modeled in SystemC-AMS. The integration between GVSoC and SystemC-AMS is already partially implemented.