| KEYWORD |
TESTGROUP - TESTGROUP
Tesi al Politecnico
Mapping an embedded Video application on FPGA using Berkeley's Rocket Chip RISC V platform
Tesi all'estero
Parole chiave EMBEDDED SYSTEMS, VIDEO PROCESSING
Riferimenti ALESSANDRO SAVINO
Riferimenti esterni Prof. Mounir BENABDENBI (TIMA Laboratory, Grenoble)
Prof. Michele PORTOLAN (TIMA Laboratory, Grenoble)
Gruppi di ricerca TESTGROUP - TESTGROUP
Tipo tesi RESEARCH / EXPERIMENTAL
Descrizione The RISC-V architecture is an open source instruction set architecture (ISA), developed at University of California Berkley (UCB) to serve as an open standard for both academia and industry. RISC V based processors are gaining more and more popularity in the semiconductor industry (Nvidia, Greenwaves, ...) and make them good challengers to ARM microprocessors. UCB is also proposing Rocket Chip, a system on chip generator embedding RISC V processor cores. The Rocket Chip framework allows generating tunable hardware and software binaries.
In this context, the proposed internship goal is to select a modern video application and map it on a Xilinx FPGA using the Rocket Chip toolset.
This project is a part of an educational project aiming at teaching the tight relationship between hardware and software using RISC V processors.
A preliminary work has been done to understand and use the Rocket Chip platform (software and hardware components). First labs have been defined but they need to be improved to target a realistic application running on a modern FPGA.
Depending on the work progress, the platform may also be tuned to allow multi-tasking applications running on a multi-core SoC: neuromorphic and/or machine learning applications
Internship tasks:
• Become familiar with the RISCV processor and Rocket Chip ecosystem
• Improve the existing labs in order to run small educational programs on different
hardware versions
• Map the system design, including the processor, a memory, a bus interconnect, ...
on a high end Xilinx FPGA
• Validate the architecture running on the FPGA, first with simple applications and
then with a realistic video application, probably a convolutional neural network
(CNN)
• Write documentation and labs/tutorials for students
• Study how to design and run a multi-tasking application on a multi-core SoC
Conoscenze richieste Very good knowledge of VHDL/Verilog, CPU based system architecture, embedded system knowledge (HW and SW), FPGA programming.
Bonus skills: Experience with RISC V architecture, System Verilog and Chisel language is greatly appreciated.
Note Possibility to spend 5 months at the University of Grenoble (TIMA Laboratory)
Scadenza validita proposta 29/01/2020
PROPONI LA TUA CANDIDATURA