Talk:CDC 7600
Template:WikiProject banner shell > The 7600 was main processor code compatable to the > 6600, but the 7600 PPs were not compatable to 6600 PPUs.
I have to deal with what I believe to be a fallacy in two different places.
I have no indications that the 7600 was in any way compatible to the 6600, not OS, not compiler, and not instruction set. And this was SOP. My experience was on a 6400 running I think SCOPE (GRC) for a firm (Information Magnetics) which had serious computing problems evaluating getting 7600 time (we never did in the end, using UCLA's 360/91).
So I checked with an ex-officemate who used S/N 1 7600 and the run down he gave me was the lack of availability of parts (memory) which resulted in the 2 stage memory which would vastly make a difference in the instruction set irrespective of the PPUs.
So I need references to convince all of us (include compilers [OS],
specific instructions, etc.).
--enm 8/18/2004
The 7600 CPU was 'almost' entirely instruction set compatible with the 6000 series. There were only a few new instructions - like being able to address Extended Memory on a word-by-word basis - rather than block copies. There were a few monitor-mode differences as well. (This is worthy of more information, because Mr. Cray redesigned the I/O in some interesting ways. Sorry, I don't have time today...) But COMPASS (the assembler), the Fortran compilers, and many other pieces of 'normal' CDC software - even COBOL - ran on the 7600 with minor changes. These changes were mostly related to the OS differences - not the CPU instructions themselves.
If you want specific references, please look at any CDC COMPASS or Fortran manual from about 1970 forwards. In fact, the later Cyber-176 system was basically a 7600-style CPU and peripheral processors, combined with a set of 6000-compatible peripheral processors. It could run the normal NOS operating system, in addition to the older 7600-specific versions of SCOPE.
--Wws 6/3/2006
Agreed with Wws. The claims in the main page are completely misleading; no other reference makes the claim that there was no object code compatibility. I propose to weaken the article in line with this discussion and other articles (notably the conflicting information in [CDC_6600] unless somebody posts some good reasons before a get a round tuit. Groogle 08:35, 28 April 2007 (UTC)
ECS word size
ECS word size was 8*60=480 (logical) or 8*61=488 (physical) with one parity bit per each 60-bit CM word.
I can't find 7600 manual online, but there is 6600 manual referenced on the 6600 page. In the ECS section (page D-1) there is a list item:
Memory organized in logically independent banks of 488-bit words (eight 60-bit words plus parity bit for each) with corresponding multiphasing of banks
6*60=360 bit words would require multiply/divide by 6 in the CM vs. ECS address conversion logic.
70.187.140.60 (talk) 03:07, 11 December 2015 (UTC)
"No fixed point arithmentic
Not entirely accurate. Consult [1] and you'll see that, as with the 6600, there was no fixed-point multiply/divide, but there was a separate 60 bit integer "long add" functional unit. Of course, the increment units also did 18-bit add/subtract operations.
Since the multiply units could do unnormalized double-precision operations, the lack of a fixed-point multiply was not a serious issue, since one could still recover a 48 bit integer product with a single floating point multiply operation.
174.22.17.211 (talk) 19:55, 11 May 2016 (UTC)CPG
- I changed the text.--agr (talk) 20:18, 11 May 2016 (UTC)
No integrated circuits were used!
instead high density modules were used, see https://people.cs.clemson.edu/~mark/cdc7600.html
Stephanefr (talk) 04:14, 21 February 2017 (UTC)
External links modified
Hello fellow Wikipedians,
I have just modified one external link on CDC 7600. Please take a moment to review my edit. If you have any questions, or need the bot to ignore the links, or the page altogether, please visit this simple FaQ for additional information. I made the following changes:
- Added archive https://web.archive.org/web/20110720064938/http://www.cisl.ucar.edu/computers/gallery/cdc/7600.jsp to http://www.cisl.ucar.edu/computers/gallery/cdc/7600.jsp
When you have finished reviewing my changes, you may follow the instructions on the template below to fix any issues with the URLs.
Cheers.—InternetArchiveBot (Report bug) 12:31, 28 July 2017 (UTC)
Acronym LLNL I think needs a definition. Lawrence Livermore National Laboratory. 24.197.209.81 (talk) 14:57, 23 February 2024 (UTC)