Computer Systems Design for Energy Efficiency

In order to be competitive in the computer science field, it’s imperative to understand the basic building blocks of a modern computer and how they directly impact the speed and efficiency of a program. Whether you work with embedded systems, mobile computer-based systems, or cloud systems, performance and energy efficiency are key drivers of usability and competitiveness of computerized products.

In this course, you will learn how to design modern multicore-based computers, and how the design choices you make affect performance and energy consumption. You will explore design principles governing modern microprocessors, such as pipelining and cache memories, as well as methods for determining the impact of your design on execution time and energy efficiency.

These skills can make a difference for practicing engineers for the purpose of building highly competitive products. Take, for instance, a smartphone with limited battery capacity. By adding value to end users through new or improved functionality, this can lead to significantly shorter operation time between battery charges, thus utilizing the already limited resources of a smartphone far more efficiently.

With this skillset, you can become an expert in computer system performance and energy efficiency - knowledge that is in high demand when designing computerized embedded products. With trends towards IoT (Internet of Things), autonomous systems and mobile computers, such a skillset will be critical in a career in systems engineering.

This course is derived from a Chalmers senior undergraduate course in computer system design.

Week 1: Execution model of computer systems

• The transformation hierarchy
• Instruction set architecture
• Execution stages
• Control, data paths, and a model of a computer
• Execution of a program
• Put it all together

Week 2: Performance and energy models of computers

• Basic performance model of an executing program
• Energy and power consumption in computer systems
• Basic energy model of an executing program
• How performance and energy can be improved
• Reporting performance and energy
• Put it all together

Week 3: Pipelining principles

• The principle of the conveyor belt – pipelining
• Instruction level parallelism and its relation to pipelining
• A pipelined model of a computer
• Pipeline hazards (structural, data and control)
• Simple techniques to avoid hazards
• Put it all together

Week 4: Cache memory hierarchy principles

• Memory system basics
• The memory locality principle
• Basic notion of cache memory and a direct mapped cache
• Set-associative caches
• Multi-level caches
• Put it all together