35364 - Digital Architectures for Signal Processing M

Learning outcomes

Analysis of algorithms for image compression and speech processing from the perspective of the digital system designer. Definitions of the specifications derived from these algorithms. Spec mapping on various computational architectures. Examples of digital signal processing algorithms suitable for parallel architectures such as digital signal processors and application specific system on chip.

Course contents

The course is focused on the specification, design and implementation of digital hardware architecture on chip, with particular attention to dedicated architectures for digital signal processing. It couples theory with a strong practical and laboratory component (>50% of the course).

Theory Module 1 (Conti):

• Introduction to the SystemVerilog hardware description language
• Algorithms for digital signal processing (DSP): numeric formats and representation as data-flow graphs (DFGs)
• Top-down design of DSP datapath microarchitectures through scheduling and mapping of DFGs; optimization techniques (retiming, pipelining, parallelism)
• Integration of DSP datapaths in complex System-on-Chips based on microprocessor cores - either as instruction set extensions or as dedicated accelerators.

Theory Module 2 (Rossi):

• Logic and physical synthesis of digital systems, impact of design choices upon performance, power and energy efficiency.
  

Laboratorio:

• Module 1: Design and simulation (Siemens QuestaSim) of dedicated hardware accelerators for signal processing; integration in an SoC platform (PULPissimo).
• Module 2: Physical implementation of system-on-chip for signal processing: logic synthesis (Synopsys Design Compiler), place and route flows (Cadence Innovus).
  

Readings/Bibliography

Teaching material: all teaching material (slides, links, scientific papers) will be shared by means of Virtuale. Teaching material is distributed in English.

Reference book:

• D. Marković, R. W. Brodersen, "DSP Architecture Design Essentials", Springer 2012
• G. De Micheli, "Synthesis and Optimization of Digital Circuits", McGraw-Hill Education 1994

Teaching methods

Classroom lessons in Italian.
Lab work.

Assessment methods

Final Report on lab work.
Unified oral exam on the full course program (theory mod. 1 + mod. 2 + laboratory).

Teaching tools

The course is very practical, with lab exercises to learn how to use professional grade tool and flows for the development of digital systems:

• Specification aided with high level languages (e.g., Python)
• Architecture design in SystemVerilog HDL
  
• Simulation with Siemens QuestaSim
• Logic synthesis with Synopsys Design Compiler
• Place and Route with Cadence Innovus

Office hours

See the website of Francesco Conti

See the website of Davide Rossi