Parallel Computing for Data Science

Chapter 2

“Why Is My Program So

Slow?”: Obstacles to

Speed

Here is an all-too-common scenario: An analyst acquires a brand new mul-

ticore machine, capable of wondrous things. With great excitement, he

codes up his favorite large problem on the new machine—only to find that

the parallel version runs more slowly than the serial one. What a disap-

pointment!

Though you are no doubt eager to get to some more code, a firm grounding

in the infrastructural issues will prove to be quite valuable indeed, hence

the need for this chapter. These issues will arise repeatedly in the rest of

the book. If you wish, you could skip ahead to the other chapters now, and

come ...