Talk:General-purpose computing on graphics processing units (hardware)
Latest comment: 4 days ago by Sirfurboy in topic Proposed merge of General-purpose computing on graphics processing units (software) with General-purpose computing on graphics processing units (hardware)
Proposed merge of General-purpose computing on graphics processing units (software) with General-purpose computing on graphics processing units (hardware)
edit- The following discussion is closed. Please do not modify it. Subsequent comments should be made in a new section. A summary of the conclusions reached follows.
- The result of this discussion was merge Sirfurboy🏄 (talk) 07:41, 22 August 2025 (UTC)
This sub of an article would be better suited in the general article that was moved to (software). TiggerJay (talk) 03:52, 14 August 2025 (UTC)
- thank you for starting the proper procedure. the answer is flat "no". everyone who knows about hardware has been referring to GPGPUs as a hardware concept. as a software engineer of 25 years who has been using computers continuously since 1978, who followed the evolution of DirectX and gaming by Microsoft in the early 90s, I had no idea that GPGPU was quotes an exclusive software term quotes. I've just found that even on the Supercomputing page and others - dozens if not hundreds of pages - all refer to GPGPU as hardware. there's *177* pages to go through https://en.m.wikipedia.org/w/index.php?title=Special:WhatLinksHere/GPGPU&limit=500 Lkcl (talk) 05:12, 14 August 2025 (UTC)
- found a way to identify uses of GPGPU without transclusion: *one* of the aliases shows a considerable portion of articles referring to GPGPU hardware https://en.m.wikipedia.org/w/index.php?search=insource%3A%2F%5C%5B%5C%5BGeneral-purpose+computing+on+graphics+processing+units%5C%7C%2F&title=Special%3ASearch&profile=advanced&fulltext=1&ns0=1 Lkcl (talk) 12:53, 14 August 2025 (UTC)
- cross-reference mistake on Nintendo switch page which has used GPGPU in the hardware sense. https://en.wikipedia.org/wiki/Talk:General-purpose_computing_on_graphics_processing_units_(software)#c-Lkcl-20250814130200-Lkcl-20250814125000 Lkcl (talk) 13:04, 14 August 2025 (UTC)
- requested hekp to identify how to get help... Lkcl (talk) 13:31, 14 August 2025 (UTC)
- another page using GPGPU as a hardware term - Supercomputer_architecture. this is an epidemic issue - will do an insource search to highlight that specific conflation. the other one had 34 uses, each of which needs review Lkcl (talk) 16:01, 14 August 2025 (UTC)
- 16 uses, a quick check shows a mix of GPGPU software, GPGPU hardware, and plain GPU https://en.m.wikipedia.org/w/index.php?search=insource%3A%2F%5C%5B%5C%5BGPGPU%5C%7C%2F&title=Special%3ASearch&profile=advanced&fulltext=1&ns0=1 Lkcl (talk) 16:03, 14 August 2025 (UTC)
- georgia tech professor presentation on GPGPU hardware using Larrabee as an example, explicitly states "most GPGPUs have fixed functions, Larrabee does all but texture sampling in SW" slide 13 https://bpb-us-e1.wpmucdn.com/sites.gatech.edu/dist/e/466/files/2008/12/Larrabee_ECE4893.pdf Lkcl (talk) 13:19, 14 August 2025 (UTC)
- Direct_memory_access talks about GPGPU hardware Lkcl (talk) 05:15, 14 August 2025 (UTC)
- Vector processing talks of GPGPU *hardware* and has not the slightest reference to the practice of utilizing such hardware for running General-purpose Operating Systems... because they already do precisely that (in most instances, today). Lkcl (talk) 05:20, 14 August 2025 (UTC)
- Your statements simply reinforce the need to have an article about GPGPU, and within there, a discussion about the concepts of GP-GPU, and how it is handled both at the hardware and software layers. It is somewhat mystifying why an article about hardware mostly references github (ie software) repos. That article is on the verge of being nominated to be deleted anyways regarding the quality of the article itself, but I think it would be best served if it was simply merged into a combined article with software. If the software page is so egregious as you state above, it would make sense to bring clarity to the confusion instead of further fracturing the article space. TiggerJay (talk) 05:26, 14 August 2025 (UTC)
- Support merge - The discussion above is a bit of a mess, but the core issue is the a reader with an information need to know about GPGPUs is poorly served by two articles that artificially divide the subject into both hardware and software. The subject is best covered in a single page. Sirfurboy🏄 (talk) 19:14, 14 August 2025 (UTC)
- ah. they don't. if you are interested to know why, please ask questions. making categorical statements is, unfortunately i ave to point out diplomatically, an *implied* combatative position that is detrimental to us both. i choose from experience not to engage in that fashion. Lkcl (talk) 06:06, 15 August 2025 (UTC)
- to other reviewers: please disregard Siefoyboy's comments, there are no questions and no engagement. by contrast i welcome and am enjoying the interaction with HelpfulRacoon as it allows for clarity on this complex (messy) topic. Sirfoyboy, what you describe as "mess" is important research ongoing into the conflation of the three (four) misuses. it identifies HUNDREDS of pages that have misused and conflated three (four) separate and distinct concepts. if you consider pointing that out to be "insulting" i can only apologise: but, nicely, it was you that made closed-minded categorical judgements, instead of asking questions. so sorry to have to highlight that publicly, i really din't like doing it, but cannot leave it unsaid either, precisely because it is so public. i don't like this aspect ofvwikipedia, it is quite painful and poignant. Lkcl (talk) 07:31, 15 August 2025 (UTC)
- It is incredibly inappropriate to call someone "closed-minded" and ask that their comment be disregarded because they didn't respond to you within 90 minutes of your previous reply. There is nothing wrong with Sirfurboy's comment here, and your bludgeoning of this talk page is concerning. REAL_MOUSE_IRL talk 08:03, 15 August 2025 (UTC)
- sadly it's just a fact. i didn't - don't - enjoy pointing it out, particularly as it risks precisely the kind of misunderstanding (and implied accusation) that you highlight. the use of the loaded and disparaging phrase "bludgeoning" is indicative that you have also disregarded protocol, i find the term offensive and belittling, where i am doing my best to be true to my nature and open and honest as i have always been (note i did not say "nice", there). welcome to the dark end of Asperger's... so sorry to have to highlight that. Lkcl (talk) 08:09, 15 August 2025 (UTC)
- Your condition does not excuse being rude to other editors. REAL_MOUSE_IRL talk 08:13, 15 August 2025 (UTC)
- stop. please. you have misunderstood. as people keep doing. just... stop. okay? let's focus on the task, which i note you kindly already helped with by observing that Ethereum page refers to GPGPU software, where Litecoin i noted refers to GPGPU hardware. it's a massive mess, that needs a huge concerted effort to clean up. Lkcl (talk) 08:22, 15 August 2025 (UTC)
- Your condition does not excuse being rude to other editors. REAL_MOUSE_IRL talk 08:13, 15 August 2025 (UTC)
- sadly it's just a fact. i didn't - don't - enjoy pointing it out, particularly as it risks precisely the kind of misunderstanding (and implied accusation) that you highlight. the use of the loaded and disparaging phrase "bludgeoning" is indicative that you have also disregarded protocol, i find the term offensive and belittling, where i am doing my best to be true to my nature and open and honest as i have always been (note i did not say "nice", there). welcome to the dark end of Asperger's... so sorry to have to highlight that. Lkcl (talk) 08:09, 15 August 2025 (UTC)
- It is incredibly inappropriate to call someone "closed-minded" and ask that their comment be disregarded because they didn't respond to you within 90 minutes of your previous reply. There is nothing wrong with Sirfurboy's comment here, and your bludgeoning of this talk page is concerning. REAL_MOUSE_IRL talk 08:03, 15 August 2025 (UTC)
- Support merge: I see where the split is coming from, physical devices designed to support general-purpose programming are a slightly different topic from the programming itself. It is currently unclear whether separate articles would be useful because the pre-existing GPGPU article is very poorly written. A lot more work has to be done to figure out how to present the hardware and software layers. This new article doesn't really help, in large part because it mainly relies on primary sources, so for now it makes sense to merge. Helpful Raccoon (talk) 06:05, 15 August 2025 (UTC)
- this is a good summary of the situation, Racoon. i've identified *three* separate decade+ long uses of the term GPGPU by doing extensive queries, and two different communities used the exact same term. the article is a mess. re "primary sources", this is not correct. i haven't finished doing the collation of sources, i mean come on it's a new article, barely out of "stub", i feel like i'm being pressurised instead of assisted here. i got some of those sources *from the RISC-V* page (the innovation section), and i know that there were presentations given at RISC-V conferences about Esperanto (because ibwas there, and Roger Espacer is on the RVV technical committee). MIAOW and VortexGPU are most categorically not "primary" sources, they are peer-reviewed work that is cross-ref'd and cited! Larrabee is highly significantly notable in that it is the work that created [[AVX512]! and that is cited *many many times* - hence the Georgia Tech reference which is *not in any way* a primary source. the problem is this: the actual number of GPGPUs in the world is very very small. Intel, AMD, NVIDIA and Qualcomm (Adreno), ImgTec (PowerVR), MALI (sold by mediatek to ARM) - those are the well-known ones. to even *know of the existence* of the others, which i only know about as an Asperger's expert in this field, you would have to do *literally months* of systematic research. i have listed a HUNDRED PERCENT more GPGPUs on this page than most people in computing even know exist! :) so please, *ask questions*, ok? really appreciated. Lkcl (talk) 06:19, 15 August 2025 (UTC)
- https://tomforsyth1000.github.io/larrabee/larrabee.html see quote on how Larabbee became AVX512: it is not just notable it is important computing history! H Lkcl (talk) 06:37, 15 August 2025 (UTC)
- slide 13. not a primary source. a PROFRSSOR. at an INDEPENDENT AND RESPECTED. university. citing Tom Forsyth's work. https://bpb-us-e1.wpmucdn.com/sites.gatech.edu/dist/e/466/files/2008/12/Larrabee_ECE4893.pdf Lkcl (talk) 06:47, 15 August 2025 (UTC)
- ah! roger espaca not espace. huge number of talks / references / articles / peer-reviewed academic research including the barcelona one i attended. https://google.com/search?q=rvv+roger+espaca+esperanto+RISC-v+vector - pretty obvious that Esperanto is not "primary source" but you have to know what to look for. if you don't know that Roger was on the RVV committee as a primary contributor, there's almost no way to find the secondary sources. Lkcl (talk) 08:16, 15 August 2025 (UTC)
- https://tomforsyth1000.github.io/larrabee/larrabee.html see quote on how Larabbee became AVX512: it is not just notable it is important computing history! H Lkcl (talk) 06:37, 15 August 2025 (UTC)
- the academic papers i have read cite a ton of research into GPGPUs. i just haven't had time to investigate this HIGHLY SPECIALIST area. VortexGPU cites a list. MIAOW GPU cites a list. Nyuzi cites a list. all of these are INDEPENDENT credible Academic Research institutions working on a very rare and very costly topic of computer science. none of which the contributors to the conflated software API page know anything about, because virtually nobody else in the world does either! ok? start to see the extent of the problem, here, which Tiggerjay went massively inappropriately overboard on, knowing nothing about the topic either! Lkcl (talk) 06:52, 15 August 2025 (UTC)
- It would take a major overhaul to cleanly divide the GPGPU page into hardware and software, especially because these layers are so interconnected. Right now there is a discrepancy between the current structure of the the GPGPU pages and how you believe they should be structured. To avoid being pressured, you could try writing a more developed draft. Helpful Raccoon (talk) 06:57, 15 August 2025 (UTC)
- actually imo the GPGPU (software) page is *already* focussed exclusively on software. it has to *mention* hardware, and i made some correctins on that already, but overall aside from being "original research" and being a bit disjointed and having not enough references, it's actually quite good. the fact that its authors genuinely believed GPGPU to be a software-only term "saved" it, there. (i spent about 5 hours reading and correcting it already). the Hardware page, i know my stuff, from 45 years continuous working with computing: it's exclusively hardware. and notable hardware at that (claims to the contrary indicate inherent lack of knowledge on the subject by reviewers: every good programmer knows that AVX512 is in every modern Intel and AMD processor). no, the real problem is, as i said: the fact that *three* separate communities used *three* (actually, four due to Nintendo switch) separate meanings for the exact same term. that has *nothing* to do with the two pages, and i already added the third meaning to the DAB page. bottom line: the destructive effects of merging ripple out into the world by destroying the need for the DAB page, ving the catastrophic false impression that one *or more* of the uses of the phrase "GPGPU" do notvexist. reducto ad absurdum, it *almost* doesn't matter that the hardware page is a stub. Lkcl (talk) 07:12, 15 August 2025 (UTC)
- Looking at this further, I think half the problem lies with the article titles. "GPGPU" can refer to software or the physical device, but "General-purpose computing on graphics processing units" refers more to software. Helpful Raccoon (talk) 07:33, 15 August 2025 (UTC)
- eeexactlyyyyy, because they are the *same acronym*. now you are getting it. GPGPUs (hardware) runs GPGPU (software) such as OpenCL, DirectX, and in the case of Larrabee which became AVX512, full operating systems. Tom Forsyth's talk is actually very funny, he got a lot of laughs on describing how a Graphics card pretended to be a Graphics card by running a full Linux OS. https://vimeo.com/450406346 MIAOW GPU on the other hand barely gets away with being a co-processor. (not in anyvway intended to disparage the incredibly valuable insights that academic research created). bottom line: this very subtle conflation is precisely why the topic has become such a mess. that, and a very large community (relatively speaking) dominated the term without even knowing that the very small (hardware) community existed, and without checking that the common usage of the term "GPGPU" is massively-conflated with "GPU" as a hardware term by over a hundred articles on wikipedia. Lkcl (talk) 07:44, 15 August 2025 (UTC)
- even the HPC and Supercomputing pages - i have to note before i edited them - were referring to GPGPU as a hardware term. before i had to divert my full attention into "defense" mode (something i reaaaaly do nor like doing) i began correcting muktiple templates as a source of massive confusion. will list them here. gimme a sec... Lkcl (talk) 07:47, 15 August 2025 (UTC)
- hmm it's just easier if i refer you to where i am maintining the list User:Lkcl#Categories - the last 4 all needed a *lot* of careful digging, due to the categorisation and sub-categorisation. one of them (parallel computing template) needed *both* adding (!) because of course you get parallel computing *software* and you get parallel computing *hardware*... aka *GPGPUs*! it's just not something that the average person would spot, let alone an expert in computing. it requires *really heavy-duty* ___domain expertise in multiple fields of computer science, which i can say i have the very rare privilege of having. Lkcl (talk) 07:56, 15 August 2025 (UTC)
- and we (wikipedians) can't change the terms to different ones, because there aren't any (not that i can think of), it's just a reality / fact, like the page on Consciousness which has a similar conflationary mess: 40 different uses for the same word! but at least, there, it is a popular enough topic to have had someone actually do a peer-reviewed academic researched paper, and i heard recently that another prominent scientist found an astounding 120 uses of the word. here, we have three. four. and one community is tiny (and not going to spend this kind of effort on defending wikipedia misinformation/miscategorisation) Lkcl (talk) 08:03, 15 August 2025 (UTC)
- Looking at this further, I think half the problem lies with the article titles. "GPGPU" can refer to software or the physical device, but "General-purpose computing on graphics processing units" refers more to software. Helpful Raccoon (talk) 07:33, 15 August 2025 (UTC)
- the amount of work needed on the software page is too great. the hardware page i am happy with, due to the notability of at least AVX512, and do not need the extra effort. if forced *as a volunteer* to perform additional work, particularly in a hostile environment of easily-demonstrably technically-false assertions (easily done due to the *world-wide* lack of knowledge), i will simply walk away from the page and wikipedia suffers and wilfully fails in its mission by knowingly publishing false and misleading information. i don't mind educating people: i *do* mind them categorically asserting false statements without a willingness to engage and ask questions. Lkcl (talk) 07:17, 15 August 2025 (UTC)
- actually imo the GPGPU (software) page is *already* focussed exclusively on software. it has to *mention* hardware, and i made some correctins on that already, but overall aside from being "original research" and being a bit disjointed and having not enough references, it's actually quite good. the fact that its authors genuinely believed GPGPU to be a software-only term "saved" it, there. (i spent about 5 hours reading and correcting it already). the Hardware page, i know my stuff, from 45 years continuous working with computing: it's exclusively hardware. and notable hardware at that (claims to the contrary indicate inherent lack of knowledge on the subject by reviewers: every good programmer knows that AVX512 is in every modern Intel and AMD processor). no, the real problem is, as i said: the fact that *three* separate communities used *three* (actually, four due to Nintendo switch) separate meanings for the exact same term. that has *nothing* to do with the two pages, and i already added the third meaning to the DAB page. bottom line: the destructive effects of merging ripple out into the world by destroying the need for the DAB page, ving the catastrophic false impression that one *or more* of the uses of the phrase "GPGPU" do notvexist. reducto ad absurdum, it *almost* doesn't matter that the hardware page is a stub. Lkcl (talk) 07:12, 15 August 2025 (UTC)
- It would take a major overhaul to cleanly divide the GPGPU page into hardware and software, especially because these layers are so interconnected. Right now there is a discrepancy between the current structure of the the GPGPU pages and how you believe they should be structured. To avoid being pressured, you could try writing a more developed draft. Helpful Raccoon (talk) 06:57, 15 August 2025 (UTC)
- another page that misuses the term GPGPU to refer to hardware (correctly, imo) where software is definitely not intended, from the context of saying "in GPGPUs". there are dozens more, it is a massive effort to correct them all https://en.m.wikipedia.org/wiki/Litecoin Lkcl (talk) 07:35, 15 August 2025 (UTC)
- "done primarily with GPGPUs" (hardware term...) https://en.m.wikipedia.org/wiki/2020%E2%80%932023_global_chip_shortage Lkcl (talk) 08:04, 15 August 2025 (UTC)
- Etherium mining was done on gaming graphics cards, not specialized GPGPUs. This is confirmed in the source for that paragraph. Therefore it is the "software term" of general purpose computing on GPUs. REAL_MOUSE_IRL talk 08:15, 15 August 2025 (UTC)
- thank you, that is very helpful in the massive disambiguation process needed. i hadn't got to Ethereum yet. and gaming graphics cards *are* considered GPGPU (hardware) - that's the point!! :) see how easy it is to make that mistake? it all comes down to a very subtle conflation between two identical acronyms https://en.wikipedia.org/wiki/Talk:General-purpose_computing_on_graphics_processing_units_(hardware)#c-Helpful_Raccoon-20250815073300-Lkcl-20250815071200 Lkcl (talk) 08:19, 15 August 2025 (UTC)
- If a RTX 3080 is a GPGPU then every modern GPU is a GPGPU, and there can be no software GPGPU if every GPU is a hardware GPGPU. Your framework for understanding this is fundamentally unsound. REAL_MOUSE_IRL talk 08:24, 15 August 2025 (UTC)
- thank you, that is very helpful in the massive disambiguation process needed. i hadn't got to Ethereum yet. and gaming graphics cards *are* considered GPGPU (hardware) - that's the point!! :) see how easy it is to make that mistake? it all comes down to a very subtle conflation between two identical acronyms https://en.wikipedia.org/wiki/Talk:General-purpose_computing_on_graphics_processing_units_(hardware)#c-Helpful_Raccoon-20250815073300-Lkcl-20250815071200 Lkcl (talk) 08:19, 15 August 2025 (UTC)
- Etherium mining was done on gaming graphics cards, not specialized GPGPUs. This is confirmed in the source for that paragraph. Therefore it is the "software term" of general purpose computing on GPUs. REAL_MOUSE_IRL talk 08:15, 15 August 2025 (UTC)
- "done primarily with GPGPUs" (hardware term...) https://en.m.wikipedia.org/wiki/2020%E2%80%932023_global_chip_shortage Lkcl (talk) 08:04, 15 August 2025 (UTC)
- this is a good summary of the situation, Racoon. i've identified *three* separate decade+ long uses of the term GPGPU by doing extensive queries, and two different communities used the exact same term. the article is a mess. re "primary sources", this is not correct. i haven't finished doing the collation of sources, i mean come on it's a new article, barely out of "stub", i feel like i'm being pressurised instead of assisted here. i got some of those sources *from the RISC-V* page (the innovation section), and i know that there were presentations given at RISC-V conferences about Esperanto (because ibwas there, and Roger Espacer is on the RVV technical committee). MIAOW and VortexGPU are most categorically not "primary" sources, they are peer-reviewed work that is cross-ref'd and cited! Larrabee is highly significantly notable in that it is the work that created [[AVX512]! and that is cited *many many times* - hence the Georgia Tech reference which is *not in any way* a primary source. the problem is this: the actual number of GPGPUs in the world is very very small. Intel, AMD, NVIDIA and Qualcomm (Adreno), ImgTec (PowerVR), MALI (sold by mediatek to ARM) - those are the well-known ones. to even *know of the existence* of the others, which i only know about as an Asperger's expert in this field, you would have to do *literally months* of systematic research. i have listed a HUNDRED PERCENT more GPGPUs on this page than most people in computing even know exist! :) so please, *ask questions*, ok? really appreciated. Lkcl (talk) 06:19, 15 August 2025 (UTC)
Support Merge and moves back to status quo ante bellum. REAL_MOUSE_IRL talk 08:26, 15 August 2025 (UTC)
- Not going to deny that the original article could do with some work, but this is a real mess now. The hardware/software split as currently being implemented is nonsensical.
- At Litecoin
By 2011, Bitcoin mining was largely performed by GPUs.
refers to consumer gaming GPUs, much like the Ethereum example above. Before this split that linked to a page on using GPUs for non-graphics workloads, but Lkcl considers this "hardware", so it should link to this page, which lists a handful of specialized GPGPUs - some of which have not even left the drawing board - that bear no resemblence to those consumer GPUs. - Above Lkcl states
Vector processing talks of GPGPU *hardware* and has not the slightest reference to the practice of utilizing such hardware for running General-purpose Operating Systems... because they already do precisely that (in most instances, today).
, and alsogaming graphics cards *are* considered GPGPU (hardware)
. Even by a narrow definition of "GPGPU" the first statement would be false, so it is impossible to square the circle here. Made even worse by the shoehorning of AVX-512 (a CPU instruction set) based on a tenuous link to a scrapped GPGPU with a primary source. - The topics of:
- Running general purpose software on GPUs, taking advantage of their massively parallel nature, especially machine learning, cryptocurrency mining etc.
- The evolution of GPUs to include some general purpose stream processing ("compute units") as a consequence of #1
- GPUs for the purpose of general purpose computing, largely as co-processors (Nvidia Tesla, AMD FirePro, datacenter GPUs, etc.), distinct from non-GPU parallel processors
- are interlinked do not fit well with this split, because the cart was put before the horse in splitting pre-expansion. The changes to surrounding articles on parallel computing have been a complete rush-job too. REAL_MOUSE_IRL talk 11:06, 15 August 2025 (UTC)
Oppose MergeMost references are self-published and if anything only 2 sources that are reliable sources. Better to take to AfD than merge with the other article unreliably sourced material. Tenshi! (Talk page) 19:37, 15 August 2025 (UTC)- Tenshi, if your view is that this split out page should not exist, opposing merge may be counter productive. If the merge fails and we take it to AfD, there is no saying if the consensus at AfD will be to delete, keep, or, indeed, merge. We could end up with a page we don't think should exist, but we are stuck with. Yes, there are big issues with the information in this page. A merge can be quite limited. We don't have to merge it all. But the principle we need to decide was whether or not this spin off should have happened. I'd support a minimal merge. Sirfurboy🏄 (talk) 20:40, 15 August 2025 (UTC)
- What do you see that could be merged here? Tenshi! (Talk page) 20:44, 15 August 2025 (UTC)
- A minimal merge can be achieved by simply redirecting the page without copying any content. If any editor finds mergeable material in the edit history, they may merge it. AFD merges often happen this way. Sirfurboy🏄 (talk) 22:19, 15 August 2025 (UTC)
- It seems we have agreement for what the end-result should be for these pages, and a disagreement in how exactly to get there. I am in favour of a merge in spirit (adding content about hardware GPGPUs to the other article), just not any of this content. REAL_MOUSE_IRL talk 10:04, 16 August 2025 (UTC)
- Yep, I agree. Sirfurboy🏄 (talk) 11:48, 16 August 2025 (UTC)
- Support minimal merge per above. Tenshi! (Talk page) 14:35, 16 August 2025 (UTC)
- Yep, I agree. Sirfurboy🏄 (talk) 11:48, 16 August 2025 (UTC)
- It seems we have agreement for what the end-result should be for these pages, and a disagreement in how exactly to get there. I am in favour of a merge in spirit (adding content about hardware GPGPUs to the other article), just not any of this content. REAL_MOUSE_IRL talk 10:04, 16 August 2025 (UTC)
- A minimal merge can be achieved by simply redirecting the page without copying any content. If any editor finds mergeable material in the edit history, they may merge it. AFD merges often happen this way. Sirfurboy🏄 (talk) 22:19, 15 August 2025 (UTC)
- What do you see that could be merged here? Tenshi! (Talk page) 20:44, 15 August 2025 (UTC)
- Tenshi, if your view is that this split out page should not exist, opposing merge may be counter productive. If the merge fails and we take it to AfD, there is no saying if the consensus at AfD will be to delete, keep, or, indeed, merge. We could end up with a page we don't think should exist, but we are stuck with. Yes, there are big issues with the information in this page. A merge can be quite limited. We don't have to merge it all. But the principle we need to decide was whether or not this spin off should have happened. I'd support a minimal merge. Sirfurboy🏄 (talk) 20:40, 15 August 2025 (UTC)
- Support. "General Purpose computing on GPUs" is a common phrase that is shortened to GPGPU (for example in Stallings, Computer organization and architecture, 2016). It is not a type of hardware or a type of software, it is an activity in which humans use GPUs for something that they were not originally designed for. It is notable and should have a Wikipedia article, not just a disambiguation page. The current "(software)" article says "GPGPU is fundamentally a software concept, not a hardware concept; it is a type of algorithm, not a piece of equipment". That was added in 2015, and in my opinion it is misleading and should be removed. The merged article should say there is software and new features of GPU hardware that assist general-purpose use. Note that "GPGPU" is used by some people to mean "General Purpose GPU" (for example in Chien, Computer architecture for scientists, 2022) and that phrase could be used for the newer GPUs. There may be little of the "(hardware)" article that should be merged; in particular, only hardware that is described as a GPU should be included. [I originally wrote some this in the discussion of the move proposal which I also support.] JonH (talk) 08:15, 18 August 2025 (UTC)
The discussion above is closed. Please do not modify it. Subsequent comments should be made on the appropriate discussion page. No further edits should be made to this discussion.