Paging and the TLB

The Atlas computer, built by Tom Kilburn's team at the University of Manchester with Ferranti and inaugurated in December 1962, had a small fast core memory and a much larger magnetic drum. Programmers on other machines spent much of their effort copying data between the two by hand. Atlas did it for them.

It split memory into pages of 512 words and used hardware page address registers to map each page a program referred to onto wherever it really lived. If the page was on the drum, the supervisor brought it in. Its designers called it a one-level store. We call it virtual memory, and paging is how every general-purpose CPU still implements it.