Estimating CPU cache size with the memory mountain | One Step! Code

The Memory Mountain is a graphical representation of the read bandwidth (the rate data is read by a program in MB/s) in terms of the total data size processed and its access pattern. It provides great insight into the way CPU cache helps bridge the gap between the high-speed processor and the relatively slow access times of the main memory. It was introduced by Stricker and Gross, and I first read about it in Computer Systems: A Programmer’s Perspective (CS:APP).

In this post, we’ll generate a memory mountain with the diagnostic program provided by CS:APP and plot it using Python’s matplotlib. Then we’ll analyze it to gain some insight about the memory hierarchy. In particular, we’ll estimate the CPU cache sizes.

As it was mentioned, the program used to measure the read bandwidth is the one provided by CS:APP. You can get it from their website (file mountain.tar in the Chapter 6 section).

Leave a Comment