Definition: The principle of locality says that most programs do not access all code or data uniformly.  (First seen on page 47.)

Since fast memory is expensive, a memory hierarchy is organized into several levels -- each smaller, faster, and more expensive per byte than the level before it.

The importance of the memory hierarchy has increased with advances in performance of processors.

There is a processor-memory performance gap that computer architects must try to close.

The memory hierarchy of the embedded computers is often quite different from that of the desktop and server for four different reasons:

1) Embedded computers are often used in real-time computers, and hence programmers must worry about worst case performance.  This concern is problematic for caches that improve average case performance, but can degrade worst-case performance.

2) Embedded applications are often concerned about power and battery life.  The best way to save power is to have less hardware.  Hence, embedded computers may not choose to use hardware-intensive optimizations in the quest of better memory hierarchy performance, as would most desktop and server computers.

3) Embedded applications are typically only running one application and use a very simple OS, if they use one at all.  Hence, the protection role of the memory hierarchy is often diminished.

4) Finally, the main memory itself may be quite small -- less than 1 MB -- and there is often no storage disk.

Web Accessibility