| PLEX86 | ||
virtual memory 4468kim kubik virtual memory 4470 common segment" ... having started out as a 1mbyte "shared" segment in every address space. the hardware table look-aside buffers (TLBs... I have to agree with Nomad here, your explanation seems to add confusion. You dwell on thrashing but don't mention the basic mechanism, which is virtualisation of process address spaces in a multiprogrammed system. Paging was the VAX approach and has been most popular since. Given the complexity of today's runtime systems it is hard to conceive of a practical modern computer that does not offer virtual memory in this sense. "Fragmentation" is not "decreased" but conceptually eliminated; each process sees one contiguous address space. And there are various other new efficiencies such as page sharing, access control etc. In the era you are alluding to it would have been SunOS (a BSD)'s paged VM on a 68K family CPU (Sun2 or more likely Sun3). You can overload, or find a pathological-buggy workload for any machine. It's a mundane anecdote and not a statement on memory management. I don't see how the phrase "Extending the address space out to disk" really describes a modern system. It may describe some primitive systems labelled "virtual memory" such as Apple's System 7 implementation. I think even Windows has something like virtual paged memory management, but don't quote me.
|
||||
Alt Folklore Computers from Newsgroups The #1 Usenet Provider on the Internet
|
||||