"Processor architecture??? Is that really to be taken seriously?"

Sure. You can get big speed-ups by exploiting the storage hierarchy, and by writing less-branchy code that does not bamboozle the branch predictor.

These can be studied from the top-level functional perspective. The low-level details can be ignored.