Objective

The course deals with the high-level structure of modern digital systems, with a particular focus on low-power systems-on-chip and on massively parallel computing platforms. Lectures combine the theoretical description of architectural concepts and techniques with the exploration of real-world implementations. Case studies are investigated detailing the characteristics of the ARM family of systems-on-chip and of the OpenCL framework. The course showcases how application domains shape the hardware features of computing ICs, which in turn demand dedicated execution and programming models.

Contents

• Introduction to computer architectures
  • Basic concepts, instruction level parallelism, hazards
  • Memory hierarchies: caches and scratchpads
• Embedded computing platforms: the ARM architecture
  • The Cortex-M processor core
  • The ARM and Thumb instruction sets
  • Programming ARM microcontrollers: C and assembler
  • ARM-based systems: interfacing peripherals on the AMBA bus
  • Managing interrupts and DMA transfers
• Massively parallel GPUs for general-purpose computing
  • Overview of GPUs organization: computing and memory resources
  • Parallelizing computational kernels: OpenCL fundamentals
  • Data placement strategies for performance optimization
  • Inter-thread synchronization

References

J. L. Hennessy and D. A. Patterson. "Computer architecture: a quantitative approach, Fifth Edition". Elsevier, 2012.

M. Scarpino, “OpenCL in Action”, Manning, 2012.