Should we clean up/orthogonalize the microprocessor/architecture/CPU tags?

Question

Our tags for various CPU families and architectures are a bit of a mess right now because we're using a lot of different conventions for them:

• Just the number: 6502, 65816, 6809, 8080, z80.
• Manufacturer name and number, separated with a hyphen: motorola-68000, motorola-680x0, rca-1802.
• Some sort of abbreviation: m68k.

Is it worth trying to standardize on something for this and clean up our current tags (presumably adding tag synonyms where necessary)? If you think we shouldn't clean up the tags, please do post that or upvote a posting saying that so it's not confused with no vote meaning "I don't care what anybody does with the tags."

Other answers ideally should be specific proposals for cleaning up the tags. As well as suggesting a tag format, it would be good to discuss how you propose to deal with other specific issues.

A particular one is CPU families. 68000 vs. 680x0 is pretty easy, but what about the family that includes the 8080, 8085 and Z80, but not the 8086 or 8088? Even more fun is Motorola and Hitachi CPUs with the 68nn designations; that's complex enough that I've summarized it at the end of this post.

By the way, I used "just the number" format for the new 6800 tag I just created not becuase I have any particular preference for it, but because that's what's most of the other 1970s 8-bit CPUs are using. I have no objection to replacing it with a different format, so long as it's clear it's about 6800-family (and not 6809-family) CPUs.

Related Questions:

• sixty eight thousand 68000 tags

Motorola and Hitachi 68xx CPUs:

These consist of two different major families, the 6800 and 6809, with entirely different opcodes. Within each family there are:

• "Variants," which do not change the CPU architecture.
• "Extensions" that change the CPU architecture compatibly.
• "Modifications" that change the CPU architecture incompatibly (some old opcodes no longer work the same way) but not so greatly it becomes a new major family (arguably).

Here's a (very) abbreviated tree to give the general idea (also see 6800 for more):

- 6800: Major family
  - 6802: variant; 6808 similar variant
  - 6801: extension of 6800; 6803 similar
    - 6811: extension of 6801
      - 68711, 68811 variants (field-programmable ROM)
  - 6805: modification of 6800 (not fully compatible)
- M14605: CMOS variant
    - 68705: variant (field-programmable ROM)
    - 6808: extension that becomes compatible w/6800 again (!)

- 6809: Major family
  - Hitachi 6309 extends 6809

Answer 1

Cleaning up the tags is a good thing but not necessarily the same as deleting the manufacturer!

I would be inclined to keep the manufacturer. There is no great benefit in short/cryptic tags, all-numeric tags don't work for other things (e.g. amiga-500 or amiga-1200) and there is some potential for confusion - e.g. there are several ICs with different functions numbered "1802" but only one RCA-1802 CPU chip. We might want a tag for such a "number-clash" IC in future for some reason. And we don't really want visitors who were looking for a LT1802 op amp, a PCM1802 A/D converter, or an NTE1802 in-car-entertainment amplifier chip.

IMO abbreviations like m68k tend to subtract value rather than add it. People are not likely to google for "m68k", for example.

Answer 2

I'll admit up front that this isn't a full answer, but it is something I still feel should be kept in mind.

Architecture is probably more important than the manufacturer of the physical chip variant.

Take the early x86 CPUs, which could be second-sourced and for which licensed clones existed. Consider the 8086, for which Wikipedia lists eight different Western manufacturers, besides Intel. Unless I am interested in a specific peculiarity of, say, NEC's variant, it's not reasonable to differentiate between Intel's 8086 and NEC's μPD8086D. If I am interested in the software side, those are both, for all intents and purposes, 8086 CPUs.

Answer 3

Our previous consensus is Matt Lacey's answer to Which is better: [intel-8080] or [8080]? Do we need both?:

You've got a good point about numbers, but I think in this realm people will actually be more familiar with 386 than they would with 80386, and in that scenario I'd be hesitant to use intel-386 unless it specifically referred to an Intel chip since in those days competing manufacturers could also make a chip called a 386.

While personally I associate 6502 with mos-6502 I feel that those searching for information online are more likely to simply type '6502' into their search engine of choice.

In short, I feel like we should lean towards the the 'common' names for tags such as 68k, 486 and 8080 but use synonyms where appropriate, i.e. if a question is more specific to a vendor or particular model. A case of that would be using 68030 instead of 68k if you were interested in the details of the 68030's instruction cache.

Upvotes on this answer should be seen as support for this proposal.

Answer 4

In general I do agree to a CVn's answer. The numbers are what we're used to. Including manufacturers is more a fanboy game than useful information. Especially since it's always possible to add the manufacturer as another tag if it's really manufacturer related (that's in fact a general argument against any combinations within one tag).

I do not think that 80386 is hard to understand or a problem, as exact naming never is.

What intrigues me more here is Curt bringing up a tree structure, as these are awesome tools for data organisation - but unlike many people think, they do not have to be used hierarchical and even less do they need to be fully defined. Using a part defined flatened tree may be exactly the way we could structure the data.

So for example CPU trees could look like this:

68k
 - 68000
 - 68020 
x86
 - 8086
 - 80386
  - 80386-SX
 - Core-2
68xx
 - 6800
  - 6801
   - 6811
 - 6809

and so on.

Not all variants are present and not all need to be present, as Tags should only be given when there is a considerable number of questions to form a useful group. Spending tags for every variant will be useless and require the addition of 'higher up' tags - which logical means that the lower level ones not need to be there anyway.

Question should be tagged at a reasonable detailed level, but only if it's about details unique to that level. Otherwise the next higher is to be used. So a generic question of like about a word move on a 68.000 is rather to be tagged 68k than 68000. Or x86 when it's about the PC. Usually this will result to collect generic questions under the generic tag, making it easy to find related questions. Doing otherwise would spread them out on many tags and make it hard to find related information.

Using such a tree (maybe maintained in some meta post?) to organize the tags seams quite helpful while keeping the tags themself as simple as possible.