Jump to content

Talk:IBM System/360

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


Compatibility Operating System?

[edit]

While the original emulator programs prior to the 360/85 ran standalone, IBM later offered Compatibility Operating System (COS) options for running IBM 1400 series under DOS/360[1] on a 360/30 or 360/40 and 1410/7010 under OS/360[2] on a 360/50. -- Shmuel (Seymour J.) Metz Username:Chatul (talk) 13:59, 31 May 2024 (UTC)[reply]

References

  1. ^ IBM System/360 Disk Operating System 1401/1440/1460 Emulator Programs Compatibility Support/30 Compatibility Support/40 Program Number for CS/30: 360N-EU-484 Program Number for CS/40: 360N-EU-4BS (PDF) (Fifth ed.), IBM, September 1970, GC27-6940-4
  2. ^ Compatibility Operating System for Emulation of 1410/7010 Programs under OS/360 on System/360 Model 50 [COS 50] [360D-11.1.025] Program Logic Manual (First ed.), IBM, September 1970, GY20-0597-0

Multiple issues in permalink/1281322478

[edit]

Edit permalink/1281322478 changed Producing a single machine in § "Family" concept to Producing a single central processing unit (CPU). However, every model of the S/360 include more than the CPU, either in a single frame or in multiple frames, e.g., core[a] and I/O channels.

The edit also added This meant that the CPU design could implement only those instructions needed for its target market in hardware, and implement the rest in the microcode. to the beginning of the next paragraph, although the old text was already incorrect. There was no S/360 model in which some instructions were implemented in hardware and some in microcode. On the smaller[b] models and 360/85 all instructions were implemented in microcode and on the larger models other than the 360/85[c] all instructions were implemented in hardware.

Channel on everything through 360/50 were implemented in microcode and channels on the 360/65 through 360/195 were implemented in hardware residing in separate[d] frames. -- Shmuel (Seymour J.) Metz Username:Chatul (talk) 20:49, 19 March 2025 (UTC)[reply]

"Implemented in hardware" is a bit of a fuzzy concept. As far as I know the microcoded S/360 machines all have a fetch-decode-execute loop in microcode, but, on some if not all machines, there might be hardware data paths that reduce the amount of work that the microcode has to do to decode or execute instructions. For example, a lower-cost machine might not have data paths that assist floating-point arithmetic and a higher-cost machine might. (For example, the 360/65 - and, presumably, the 360/67 - has a parallel adder for binary integer arithmetic and floating-point mantissa arithmetic and a serial adder for floating-point "charistic" (characteristic?) arithmetic (both are used for variable field-length operations) - see the IBM FE data flow diagram for the 360/65.
(The channel thing is a bit interesting - I guess, when they were designing S/360, channels independent of the CPU were a high-end-computer thing, but they presumably wanted I/O to work the same on all models, so, on the lower-end machines, they implemented non-independent channels that steal CPU cycles.)
(Oh, and you probably meant "Except 360/95, which had thin film storage." in the first footnote. :-)) Guy Harris (talk) 21:57, 19 March 2025 (UTC)[reply]
Yes, there was considerable variation in how much hardware was available to the micrcode and in the width of internal registers. The F/D/E loop disappeared on the 360/85, which had a separate I-unit.
Yes, external channel boxes were the norm on the 7000[e] series, and the S/360 PoOps manual definitely requires that channels look the same regardless of how they are implemented.
Thanks for catching the typo. -- Shmuel (Seymour J.) Metz Username:Chatul (talk) 13:42, 21 March 2025 (UTC)[reply]
Should footnotes include the abortive models 60, 62, 54, 55, 70, 90 and 92? -- Shmuel (Seymour J.) Metz Username:Chatul (talk) 14:14, 21 March 2025 (UTC)[reply]
In my opinion, model numbers that were announced but never shipped should be mentioned but when referring to actual computers they can be omitted. Thus, models 60 and 62 were replaced by model 65, models 64 and 66 were replaced by model 67, model 70 was replaced by model 75, and model 90 was replaced by model 92 which was replaced by model 91. John Sauter (talk) 21:21, 22 March 2025 (UTC)[reply]

Notes

  1. ^ Except 360/95, which had thin film storage.
  2. ^ Up to the 360/67.
  3. ^ 75, 91, 95 and 195
  4. ^ 2860 Selector Channel, 2870 Multiplexer Channel or 2880 Block Multiplexer Channel unit.
  5. ^ I don't know whether ther 1410 and 7010 had outboard channels.

Hatnotes and see also?

[edit]
Should there be hatnotes mentioning ? See also entries? The infobox does show them. -- Shmuel (Seymour J.) Metz Username:Chatul (talk) 14:03, 30 April 2025 (UTC)[reply]

Magnetic cartridges?

[edit]

Should media in the infobox list the tape cassettes written by the MT/ST and read by the 2495? -- Shmuel (Seymour J.) Metz Username:Chatul (talk) 13:47, 1 August 2025 (UTC)[reply]

Sure, kind of a niche product, but why not. (I just worked on IBMTCRIN - interestingly, the source code for this is available). Peter Flass (talk) 14:37, 1 August 2025 (UTC)[reply]
I would suggest no, in fact note 9 and 7 track tape are listed but none of the other formats, so I think maybe the better idea is to minimize as I just did Tom94022 (talk) 01:01, 2 August 2025 (UTC)[reply]
This article is System/360 only, so no 3480 or 3490. I don't know what other types of tape cartridges it could use. Peter Flass (talk) 03:11, 2 August 2025 (UTC)[reply]

Processor families in see also

[edit]

@Peter Flass: A recent edit, permalink/1303696378, added IBM 9370 to § See also. That processor is one of many implementations of IBM System/370 architecture, which is already in the list and includes

Some of these fall in multiple architectures in the list, and it is not clear that the 9370 has a better claim that any of them. -- Shmuel (Seymour J.) Metz Username:Chatul (talk) 20:47, 1 August 2025 (UTC)[reply]

I see at least two flavors of "follow-on" to S/360. There are product lines, in the sense of "IBM gave this set of S/360 descendants a name", differentiating the "IBM System/370 Model 1xx" from the "System/370-compatible" 303x, 308x, 4300, 9370, and so on, and doing similar things for 370-XA and see successors (IBM System/390 had various subgroups as shown in IBM System/390 § S/390 computers. There are architectures, in the sense of "everything in this set of machines had the same architecture" (S/370 including the compatibles, S/370-XA, ESA/370, ESA/390, z/Architecture). Some product lines may have products with different architectures, and some architectures may be used by more than one product line.
And, for some architectures, there's an overall page and an architecture details page, e.g. IBM System/360 and IBM System/360 architecture, IBM System/370 and IBM System/370 architecture, IBM System/390 and IBM Enterprise Systems Architecture § ESA/390 architecture, [IBM Z]] and z/Architecture.
IBM 9672, like ES/9000 is part of IBM System/390. So is IBM Multiprise, but it has its own page for some reason.
So it's a bit of a mess. Guy Harris (talk) 21:53, 1 August 2025 (UTC)[reply]

Centralize S/360 through z register layouts?

[edit]

It may be desirable to put all of the floating point and register layouts for S/360, 360/67, S/370, S/370-XA, ESA/370, ESA/390 and z/Architecture in one place and include them in, e.g., IBM System/360 architecture, as needed with, e.g., {{excerpt}}. That would eliminate the need to make changes and corrections on multiple pages. -- Shmuel (Seymour J.) Metz Username:Chatul (talk) 11:38, 21 September 2025 (UTC)[reply]