I the core book, he typically sticks to pseudocode. And with the later sections, most of the implementations he does to check ideas are with cweb.

In that vein, I highly recommend reading some of his programs directly. You can find them here: https://www-cs-faculty.stanford.edu/~knuth/programs.html

Expect code that will challenge some ideas on premature optimization. He codes much closer to the numbers than many are used to.

He now uses MMIX, a modified version of MIX with RISC characteristics.