Parallel Computing for Data Science

22 CHAPTER 2. OBSTACLES TO SPEED

that cache hit rates are typically well above 90%. Note carefully, though,

that this can be affected by the way we code. This will be discussed in

future chapters.

A machine will typically have two or more levels of cache. The one in or

next to the CPU is called the L1, or Level 1 cache. Then there may be an

L2 cache, a “cache for the cache.” If the desired item is not found in the L1

cache, the CPU will then search the L2 cache before resorting to accessing

the item in memory.

2.3.2 Virtual Memory

Though it won’t arise much in our context, we should at least briefly dis-

cuss virtual memory. Consider our example above,