The end

It is time to wrap up this journey now. After going through the code, I summarize the steps of implementing a home-brew performance monitor tool:

a) Get the performance raw data: /proc is a treasure for the whole system while /proc/[pid] for a specified process. If you want to observer proprietary devices (e.g., NVIDIA GPU), please refer vendor APIs;

b) Understand and parse data;

c) Display them. You can choose primitive ncurses or more fancy modern GUI libraries.

If you are learning a new programming language, reinventing a monitor wheel may be a good exercise, isn't it?

P.S., if this small project gives you little help, please consider give it a star in github. :-)