Review V2: Principles of virtual memory

A straightforward implementation of virtual memory places the page table in memory. Unfortunately, this means that each attempt to access a memory location actually requires two memory accesses — one to look up the page table entry in virtual memory, and one to access the requested memory within the page frame given by that entry. This effectively halves the time to access each page, when compared to a system that does not use virtual memory.

Designers reduce this problem dramatically by caching a small number of frequently-used page table entries on the CPU chip, in a portion of the chip called the translation lookaside buffer, or TLB. Since accessing data stored on the CPU chip is much faster than accessing data in memory, the TLB can dramatically reduce the time for looking up page table entries, restoring a access speed more comparable to a system without virtual memory.

A page directory is a table of pointers to page tables for regions of virtual memory. A CPU uses the page directory instead of a single page table, because a single page table can become very large and it must be in memory in order to allow for address translation. With a page directory, the partial page tables can be paged in virtual memory. (The page directory must occupy RAM for address translation, but the page directory would be much smaller than a single exhaustive page table.) Moreover, page tables for unused regions of memory need not occupy even virtual memory.

• It provides a larger range of memory address beyond what actual memory the computer holds.
• It allows each process to have its own address space.
• It enables two processes to share the same memory easily.
• It enables memory to be moved between addresses rapidly, since altering the paging table moves the page without any actual memory copying going on.