Introduction

In this segment of the course we consider loops involving arrays. We will see

How execution order of a loop is constrained,

How a compiler can extract dependence information, and

How this can be used to optimise a program.

Transforming the execution order of a loop enables various architectural features to be used efficiently, in particular:

• Pipelined processors.
• CPUs which can process more than one instruction at a time (superscalar and VLIW).
• Systems which rely heavily on caches.
• CPUs with special instructions for vectors.
• Multiprocessors.