Advanced Computer Engineering

Overview

Streaming workload modelling languages
* Algorithm optimisation schemes
* Handling of time in algorithm design
* Number systems and operations
* High Level Synthesis (HLS) technology for Field Programmable Gate Array (FPGA)
* Application partitioning for parallel processing platforms
* System optimisation

Learning Objectives

Understand the design principles for design of heterogeneous hardware/software embedded digital signal processing (DSP) systems, in five specific areas: computing architectures, application modelling, parallel partitioning, scheduling and code generation, and implementation optimisation.

* In the area of computer architectures, students will be able to:
* The influence of number system on accuracy and cost of different number systems
* Handling time and dependency in custom architecture design
* Discriminate how behavioural expressions of a function translate to circuit architectures using High Level Synthesis (HLS) technology.

* In the area of application modelling and code generation, students will be able to:
* Analyse and compare dataflow languages for a given application
* Derive firing rules for dataflow actors
* Apply mathematical consistency checks to static dataflow models
* Appraise the implementation concerns of parallel processing algorithms
* Investigate constructive hierarchical and multi-stage partitioning algorithms
* Relate constructive and iterative partitioning algorithms
* Relate partitioning algorithms to achieve specific implementation goals
* Analyse dataflow models for deadlock
* Analyse the code and data memory requirements, throughput and efficiency of the resulting embedded schedules

* In the area of optimisation of custom systems, students will be able to:
* Outline the behaviour of system optimisation approaches.
* Contrast graph transformation techniques for optimisation of embedded dataflow schedules
* Transform embedded schedules for optimisation with respect to data memory, throughput and efficiency
* Relate advanced dataflow models for further optimisation with respect to a given criteria
* Illustrate retiming, folding and unfolding, hardware sharing for dedicated hardware optimisation

Skills

Assimilation of technical material
Critical thought in the design of resource-constrained computer designed problems
Application to practical data processing design examples

Assessment

None