Chapter 4.
Memory

In the previous chapter, we saw that parallelization can dramatically speed up computation. But even with perfect parallelization, the cores can sit idle, waiting for data to arrive from memory. Before we can understand how to fix that, we need to understand memory itself.