Green-Five: An open-source system-level power analysis method for energy efficient RISC-V ISA
Thesis in external company Thesis abroad
keywords ENERGY EFFICIENCY, LOW POWER, RISC-V
Reference persons FABRIZIO RIENTE
External reference persons NASSER Yehya
Research Groups EC à IMT Atlantique/Lab-STICC, Brest, France., VLSI THEORY, DESIGN AND APPLICATIONS (VLSILAB)
Description he problem of high-power consumption in modern embedded systems could be solved by either increasing energy production or reducing the power consumed by emerging embedded applications. To work on reducing the power consumed by our modern embedded systems is an ethical obligation. To efficiently reduce power consumption, software and hardware designers must respect the power budget from the early stage of the system design. To this end, getting early energy-aware embedded systems into the hands of the system designers may strategically push the system to be green and sustainable. The road to energy-aware embedded systems starts with providing power estimation methods and models which help analyze, estimate, and monitor the power consumption of an embedded system.
In this project, I propose combining two levels of information: micro-architecture specifications and software information. In addition, I propose considering implementation-specific information (such as voltage, frequency, and technology node) of the target platform. I believe that this combination of information will make the power estimation more accurate since measurement-based estimation delivers better accuracy than simulation, flexible with the possibility of exploration of many design scenarios, fast (i.e., order of microseconds versus long simulation time), scalable (i.e., independent of application size), and general (i.e., can work for any user application).
Deadline 07/01/2024 PROPONI LA TUA CANDIDATURA