Talk:Gallium3D

Latest comment: 7 years ago by Klbrain in topic Merger proposal

useful information

edit

I have added this link, with useful information about Gallium 3D, but a bot has removed the link. Gallium 3D introduction

Is any problem if the link is added? --190.154.78.75 (talk) 20:31, 1 May 2008 (UTC)Reply


As described here Gallium 3D and in other web pages, Gallium 3D will no be a replacement to Mesa 3D, or will it be a replacement? --190.10.170.17 (talk) 01:47, 9 July 2008 (UTC)Reply

Notability?

edit

Isn't the link to the presentation at the 3rd party KDE's aKademy enough to establish notability? Diego (talk) 08:08, 12 August 2008 (UTC)Reply

And there's also a talk[1] at FOSDEM. Diego (talk) 08:10, 12 August 2008 (UTC)Reply
They should be enough, yes. Gallium3D is notable enough. --136.142.214.19 (talk) 13:58, 12 August 2008 (UTC)Reply
Ok - I've reviewed the sources available and they don't qualify as significant coverage (i.e. they don't "address the subject directly in detail"). I can see the potential for this library though, and the signs that it's getting traction between Open Source developers, so it might not be wise to challenge it for deletion, for a while.Diego (talk) 13:54, 14 August 2008 (UTC)Reply
I agree, keep for the moment. Adamantios (talk) 19:33, 14 August 2008 (UTC)Reply
It seems that wikipedia is getting really whiney about notability. Even if the project fails, there is enough code available already to make this more than notable enough for an online encyclopedia. Even if the effort fails, I'd love to see it kept as part of the historical record. There are much less popular packages that are (and should be) mentioned by wikipedia. --208.104.220.127 (talk) 05:24, 17 August 2008 (UTC)Reply
Exactly! Adamantios (talk) 10:27, 17 August 2008 (UTC)Reply
So how are you going to guarantee verifiability if there are no reliable sources on the subject? That's why Wikipedia requires notability, not because we're mean. -- intgr [talk] 21:04, 17 August 2008 (UTC)Reply
You might be confounding notability and verifiability here. For the project's description to be verifiable, it only needs a link to the source code, since any assertion about what Gallium3D does can only ultimately be verified by reading the program. For other assertions about the project, the evidence is that there are third party places where the project is being discussed.
For it to be notable, OTOH, you really do need reliable sources. But if you read the notability policy guideline with care, it explicitly warns against deleting an article based on the way it currently proves its notability: "it is important to not just consider whether notability is established by the article, but whether it readily could be.[...]the discussion should focus not only on whether notability is established in the article, but on what the probability is that notability could be established".
On this premise, my reflection above was just a warning to be cautious and not too ready to deletion. It may be too soon to decide if there exist enough sources for this decision to be called now, given that the outlook seems to be changing rapidly. And a Google count is not enough to assert or deny notability - this project could be being discussed in private expert forums, and Google have not noticed about them yet. We should wait to see what do experts say about Gallium3D now that it's beginning to fall under their radar. Diego (talk) 19:31, 18 August 2008 (UTC)Reply
I'm just clarifying why Wikipedia needs a notability criteria: articles should display evidence that they could be made verifiable. Since there aren't sufficient references as of now, it wouldn't really belong to an encyclopedia. But because this is fairly certain to become notable soon, nobody is proposing it for deletion, and I doubt it would pass the deletion discussion anyway. I have seen lots of cases like this where notability-tagged articles stick around for months, before someone comes along and fixes them up. I fail to see a problem with this.
And no, source code is not a legitimate reference on Wikipedia. -- intgr [talk] 21:17, 18 August 2008 (UTC)Reply

This article sucks

edit

Gallium3D is not a replacement for Mesa or DRI. It is just a new internal API between Mesa and 3D DRI drivers. It is sort of an abstraction of a contemporary graphics accelerator (which is the proper way to design device drivers). This is in contrast to the old DRI design, where the 3D drivers reimplemented parts of the OpenGL pipeline again and again.

Gallium3D will make it easier for developers to write the drivers. But for end users it will be quite invisible. While it generates a lot of talk, it may be notable enough to get a section in the DRI article, but hardly notable enough to have an article of its own. Jiri Svoboda (talk) 14:27, 10 April 2009 (UTC)Reply

Intel seem to disagree with you. They rather write their driver the old way, and do not loose 5-10% of computing power. For people who would like to use Linux as a gaming platform, 5-10% loss in performance due to (avoidable?) layers of abstraction (i.e. John D. Carmack's "layers of crap") is quite much. The claim, that Gallium3D could cost maybe 5-10% of performance is from some VMware employee, in some Video I once watched on YouTube. User:ScotXWt@lk 06:59, 19 June 2014 (UTC)Reply
Please

Current status

edit

The section Current status should be reworked to a summary what is implemented in the current release and what is to come next. Mranderson2008 (talk) 14:38, 20 July 2009 (UTC)Reply

platform applicability

edit

the article as it stands takes for granted or relies on implicit knowledge wrt what platform this is relevant on. We have a mention of Linux in the article text and a "cross-platform" link in the sidebar. Is this targetting Linux? More than just Linux? Is it for embedded, desktop, etc.? -- 188.222.50.68 (talk) 17:50, 13 December 2009 (UTC)Reply

It is used for 3D video drivers, I imagine it can be useful on any platform. It's already said that it could support OpenGL and Microsoft's DirectX in a single framework.
Being open source software, it's unlikely to be used by proprietary operating system drivers, but nothing in Gallium3D precludes porting to Windows. Gallium3D has already been ported to FreeBSD. -- intgr [talk] 20:12, 14 December 2009 (UTC)Reply
VMware does virtualization. So, it is relevant for any platform. Developers of Free and open-source graphics device drivers/Mesa 3D make use of it, since it lessens their burden to develop device drivers. But e.g. Intel, still programs their Linux driver the old way, most probably because the Gallium3D drivers introduce a performance penalty (and maybe also, because improvements in Mesa3D directly benefit the device drivers for the competition). User:ScotXWt@lk 07:05, 19 June 2014 (UTC)Reply

OpenCL

edit

Does a developed driver with Gallium3D also provide OpenCL? —Preceding unsigned comment added by 41.210.9.210 (talk) 10:54, 18 November 2010 (UTC)Reply

Yes, by relying on the "Clover" (CL over Gallium) state tracker. Czukrae (talk) 02:04, 21 October 2015 (UTC)Reply

Merger proposal

edit

I'd like to propose to merge Gallium3D into Mesa (computer graphics). Really Gallium3D, as part of Mesa, never should have had a separate article (except as a highly technical expansion, but that's not the case). Gallium3D can hardly be explained outside the scope of Mesa, and any description of Mesa is incomplete without Gallium3D. Furthermore, Gallium3D is currently very brief and incomplete, and it's easy to migrate its content to its own section in Mesa (computer graphics). User:ScotXWt@lk 14:25, 16 March 2015 (UTC)Reply

Yes, it doesn't seem like there is enough content to justify forking. Sizeofint (talk) 02:40, 21 October 2015 (UTC)Reply
This seems fine by me. Gallium is said to have been designed with genericity and platform-independence in mind, but is there any other implementer apart from the drivers in the Mesa project? Czukrae (talk) 02:53, 22 October 2015 (UTC)Reply
I'm still fine with this, but I just found that there are, in fact, publicly-known proprietary users of Gallium3D which are not part of the Mesa project codebase; see the conversation in the section below. Czukrae (talk) 12:02, 23 October 2016 (UTC)Reply
I think it probably can be merged in because all of the active developed portions of Gallium 3D are part of the Mesa3D Codebase branch. As far as I know, there are just a few legacy drivers which are proprietary. These are mostly for archaic systems that are still in use in a handful of specialty fields and difficult to upgrade to modern PC hardware without major effort on retrofit parts. — Preceding unsigned comment added by RuediiX (talkcontribs) 04:53, 2 August 2017 (UTC)Reply
Agreed and   Done Klbrain (talk) 02:27, 12 September 2017 (UTC)Reply

Differences from classic graphics drivers

edit

The following statement seems incorrect to me:

"All vendor device drivers, due to their proprietary and closed-source
nature, are written that way, meaning that, e.g. the AMD Catalyst implements
both OpenGL and Direct3D, and the vendor drivers for the GeForce have their
implementations."

It seems to me that the fact of them being proprietary or closed or binary blobs does not preclude them from hooking into the Gallium3D architecture. Or does it? Furthermore, notice that the license information in this page says "MIT". This is a weak license, so it's entirely legal for any of those proprietary blobs to be embedding Gallium or a Gallium-derived architecture into them.

Care to explain where you pulled this from ScotXW? [1]

[1] https://en.wiki.x.io/w/index.php?title=Gallium3D&diff=next&oldid=573141839
The statement is correct. You'd have to better understand the nature of Gallium3D to see that. Feel free to edit the entire article. Also, please note my proposition to merge this into the Mesa 3D article. User:ScotXWt@lk 14:27, 21 October 2015 (UTC)Reply
Hey, it may very well be. Regardless, you're not really giving much of an answer. I wanted some explanation or perhaps a reference, you are just giving me "it's correct" and "you don't know enough".Czukrae (talk) 02:40, 22 October 2015 (UTC)Reply
Hey ScotXW, look what I just found:
"In November 2009, the Linux Foundation released the details of a new, rewritten Linux driver that would support this chipset and Intel's other upcoming chipsets. The Direct Rendering Manager and X.org parts would be free software, but the 3D component (using Gallium3D) will still be proprietary.[52]"
Quoted from Intel_GMA#Linux, edited by BeebLee. The reference points here.
I think this lends strength to the point I was trying to make (almost exactly a year ago, funny). What do you think? Czukrae (talk) 11:57, 23 October 2016 (UTC)Reply

Maybe something like this would be more accurate:

"Many proprietary drivers are likely written that way, meaning that, e.g.
AMD's and Nvidia's proprietary drivers have each their own implementations of
the same functionality."

The "likely" being on purpose. Czukrae (talk) 02:55, 21 October 2015 (UTC)Reply

May want to ping @ScotXW: Sizeofint (talk) 04:53, 21 October 2015 (UTC)Reply