Jump to content

Talk:UNIVAC Series 90

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

Virtual Memory

[edit]

"In fact RCA was credited with coining the term Virtual Memory.[citation needed]". This seems not to be true, at least in this context. I found a reference "Virtual Memory in the Stretch Computer" by Cocke and Kolsky from 1959 [1]showing an earlier use of the term. Peter Flass (talk) 12:42, 10 February 2014 (UTC)[reply]

The 1959 paper definitely coins the name early, so you were quite right to remove the claim. However, what the paper seems to be describing (I previewed it quickly via Deepdyve) isn't what we would nowadays call virtual memory: it was more like instruction prefetch. BTW Peter Denning's 2005 article The Locality Principle (preview via Deepdyve here) identifies the Atlas Supervisor as being the first implementation of paged virtual memory, and says that 'the term "virtual memory" appears to have come from [a 1965-66 IBM research project]'. - Pointillist (talk) 14:20, 10 February 2014 (UTC)[reply]
Might that project be the IBM M44/44X or the IBM CP-40? Guy Harris (talk) 20:01, 10 February 2014 (UTC)[reply]
Thanks for pointing to those. I suppose Denning must have been referring to the IBM M44/44X. He was clearly aware of Stretch (he lists it in his article) but had perhaps forgotten that the term virtual memory had already been used by the Stretch designers. - Pointillist (talk) 21:45, 10 February 2014 (UTC)[reply]
Hmm. We're talking IBM here, so "virtual storage" rather than "virtual memory"; see, for example, the title of the Belady paper "A study of replacement algorithms for virtual storage computers," IBM Systems Journal Vol. 5, No. 2 (1966), pp. 78-101. Perhaps RCA was the first to call it "virtual memory" rather than "virtual storage"; if so, perhaps they coined it themselves or perhaps they took "virtual storage" and replaced "storage" with "memory". Guy Harris (talk) 23:12, 10 February 2014 (UTC)[reply]
Could be, but if RCA was first to use the term in this context we need sources. So far there's the 1959 Stretch paper (with "Virtual Memory" in the title) and Denning's article. I'm not sure it's worth the effort taking this any further! - Pointillist (talk) 23:39, 10 February 2014 (UTC)[reply]
I made a quick effort to find the first use, but it's so widely used it's hard to track down. The earliest use Google ngram viewer finds are the FJCC Proceedings from 1965 (not accessible), or a whole slew of references from 1967 including RCA and Engles, R.W. (1967). Concepts and Terminology for Programmers. International Business Machines Corporation. Peter Flass (talk) 00:23, 11 February 2014 (UTC)[reply]
1965 FJCC Proceedings? (Might be behind a paywall.) Guy Harris (talk) 00:53, 11 February 2014 (UTC)[reply]

Nope, no paywall. Introduction and Overview of the Multics System, which says

Some of the highlights of the subsequent papers are: a virtual memory system for each' user involving two-dimensional addressing with segmentation and paging; ...

on page 186 of the proceedings. They don't explicitly define the term, so it might have been used before then. Guy Harris (talk) 01:08, 11 February 2014 (UTC)[reply]

If The Chartered Mechanical Engineer really was published in 1954, it was used well before then; however, the quoted text does not read like something that would have been written in 1954, so I'm not sure I'd trust that reference.
That came from a Google Books search for "virtual" memory in books from January 1948 to December 1965; the other computer-related items it found were a couple of references to the Stretch paper, Information Processing in Japan referring to what I suspect was a Hitachi machine with "two-dimensional addressing" and "virtual memory", and another Multics paper.
So, absent a pre-1965 reference to "virtual memory" from RCA - which I suspect we won't see, given that this document seems to speak of an RCA announcement of the 70/46 in 1967 - I'd say they didn't coin the term. Guy Harris (talk) 01:23, 11 February 2014 (UTC)[reply]

split

[edit]

the series 90 and series 70 were unrelated families of machines. I think this article should be split. Peter Flass (talk) 14:23, 18 October 2019 (UTC)[reply]