Servers and desktops; whole different worlds. I don't think any server admin really cares about 3D. Server boards typically have something like an ATI Rage capable of supporting VESA perfectly fine.
Google want to opensource gfx drivers
- Thread starter Davros
- Start date
Aerows
Regular
It is in the best interest of people that *use* Linux to encourage the development of good drivers. There are rogue idiot users running Windows, and bad programmers for every platform under the sun - the vast majority, though, are eagerly awaiting any driver that opens a wormhole to get more frames
. Sooner or later the "oddball in the darkened basement" perception of Open Source and Linux users on the whole needs to change.Believe it or not, it is maturing. That's all
.I'm well aware, but companies don't have much to gain from opening their drivers (they're not going to get a free driver team; they might get the small bugfix here and there, but they're not going to develop themselves). They do, however, have a lot to lose (basically reveal a big part of the architecture, all the shader compiler bits, etc).Believe it or not, it is maturing. That's all
Aerows
Regular
They sure care about what gets *put* on that network though - for instance, bad network drivers can massively influence speed, as malconfigured applications also can. My point is that those folks are often in the meetings where such purchasing decisions are made, and some are making those decisions.
I am not overwhelmingly in favor of Open Source drivers - I was more pointing out the narrowness in thinking that all of the *Nix users are whacko hacks rather than the IT professional, and that there is a very large relevance to development of Linux drivers.
Your idea of standards is a good one. Wholesale dismissal of the *Nix community is not, however. I realize that it has yet to make substantial inroads into the desktop/workstation roles, but more and more people are becoming literate, particularly with the arrival of Ubuntu and more friendly distros. It would be in every hardware vendors best interest to not completely ignore that segment of the market, as you never know if Microsoft might finally push half of the planet over the edge
.I never said anything like that. Besides, I work for TransGaming, so I really care about the succes of Linux. I just don't think open graphics drivers are the answer. And obviously it's not realistic in the first place.
I wouldn't mind if Ubutu positioned itself as the Linux desktop distribution, and sat around the table with NVIDIA and ATI to discuss how they can help each other...
No, actually I think they are pretty smart considering how many people they hoodwink, and how much $Money$ they make, even while selling the song & dance that because their software is "open sourced" that means it's "free software"....
I just read a little ditty by the CEO of the Ubuntu group in which he raved about how he loved "free software." Amazingly, nobody besides me reading the article bothered to comment that it was too bad that he didn't make any himself... I cannot see that "open source" companies operate any differently at all from private companies--a company by any other name smells just the same--just ask Mozilla.I don't object to the so-called "open source" companies--in fact, I think competition is great. What I object to is all the vocal primping and preening they do about "free software" when the truth is that none of them work for free, and would probably rather die first...
I'd venture to say that at present the term "open source" is so laden with half-truths, lies, and distortions, that nobody really knows what it really is or how it really works. It's become a meaningless political banner waved most often by folks who think mixing politics and technology is "fun and cool." Obviously, I think it is neither.
In the case of Google, a for-profit company whose products are the furthest thing in the world removed from "free," the term "open source" takes on particularly ominous tones. Google is just chomping at the bit over the prospect of getting "net neutrality" legislation passed so that it will have a legal back door to proceed against ISPs who may not wish to play the game the Google way, and Google would like nothing better than to somehow wrest graphics control away from the companies who actually manufacture gpus and write the drivers that make them useful--so that Google could have some fun with the graphics people are exposed to--and whether those people desire this from Google or they don't is certainly of absolutely no concern to Google at all. Somebody mentioned that maybe Google wanted to get involved with graphics drivers to make them more tolerant of Google's embedded advertising. That's certainly a possibility, but I can see them taking their "open source" graphics drivers even further by embedding all kinds of stuff into the drivers themselves. Wouldn't that just be Jim Dandy?
Folks better wake up and realize that the "open source" movement (if you can call it that with a straight face) has been seen and recognized for the hollow shell that it is by some very powerful for-profit companies, not the least of which is Google. They know that the term resonates with some people who are, frankly, completely unaware of why it resonates with them (does it just sound good or noble, I wonder? Is that it?)...
As a result, this isn't "your father's open source" concept that we are talking about here, but rather a competitive game among for-profit rivals that is ratcheting up to new levels of PR and greed. It doesn't have to succeed, however, and that's the good news. We can start protecting ourselves from the subtlety and insidiousness of this approach by consigning the term "open source" to the waste bin of technology history, along with other terms just as meaningful, such as "pie in the sky." But I wonder if all of us can do that. Companies like Google, at the moment, are betting that we can't, and they hope to slide in their schemes and dreams of profit and glory under the public radar by using terms exactly like "open source" and "network neutrality."
Decades ago in the US, President Eisenhower warned the nation about a brewing evil that he could plainly see--the merger of the military and industry. Today, I believe, were he still with us, he might very well warn us of another brewing evil that seems very apparent to me, the "political-technological complex."
Politics has no place inside technology, and vice-versa. I hope that we can keep things that way, but I fear it may already be too late. I mean, some people have so thoroughly intertwined the two concepts in their thinking that when they talk about politics they think they are talking about technology, and vice-versa. I find that frightening.
Some people here are right about one thing -- there's nothing I can see to motivate ATi or nV to open source their drivers, and so it's not going to happen anytime soon. However, making comments like anybody skilled enough to work on driver teams aready does so is about as wise as saying that anybody who could contribute to an operating system (or a million other projects) is already paid to do so. Depsite what the pros would like you to think, there are not an infinite number of driver-coding positions available and yet the state of drivers is frequently a mess (or always, in Linux). There are thousands of people with the skills to contribute who either a) aren't quite competitive enough to get into those handfuls of jobs or b) just don't want to devote their entre lives to driver coding. Just as other open source projects have revealed (i believe that the majority of the individual open source projects have failed, but also that the majority of the fields that have open source projects have seen great improvements as a resut of said open source teams), there are countless improvements to be made, whether it's just someone with a brain redesigning the layout of the control panel dialog box, or game-specific hacks for driver performance. To say nothing of all the education these college kids get from studying how these things work.
Again, there's no way this is going to happen so I don't want to argue it too much, But some people sound like stuffy aristocrat morons when they dismiss the capabilities of "the public". These aren't so often neighborhood kids who work on serious open source projects, but professional coders from other fields, or students of the subject field. Wake up. 9_9
Again, there's no way this is going to happen so I don't want to argue it too much, But some people sound like stuffy aristocrat morons when they dismiss the capabilities of "the public". These aren't so often neighborhood kids who work on serious open source projects, but professional coders from other fields, or students of the subject field. Wake up. 9_9
Okay, let me put it this way: There are tons of people with the hacking skills to work on graphics drivers, but only a few have the engineering skills to make good long-term decisions.
Premature optimization is the root of all evil. If you have a bunch of people changing part of the project to their liking they will create a mess. The person with a brain redesigning the layout of the control panel will likely extend the interface with a feature or two which fails on some configurations he thinks are not important (or which he simply doesn't know about). And the game-specific performance hack adds a few thousand lines of zombie code with a tiny dependency so nobody dares to touch it. Nobody really takes responsability, but they'll gladly take the credit when (if) things work out well (for a while).
It takes years to master graphics driver development, and even more years to manage it well.
You're using that as an argument why open graphics drivers would be useful? Seriously? There are a lot more exciting projects for "college kids" to work on, where they would learn a lot more, and where they can actually make a difference. Heck, which college teaches you the ins and outs of OpenGL, assembly programming, hardware design, etc? Let's learn how to walk before we start to run, ok?
I suggest you have a look at the Open Graphics Project. These professionals promised hardware that can beat the early GeForces, years ago...
Their mailing list is quite hilarious at times. They all dream of different features but they haven't even got the basics right. When the project was born I told them to have very clear goals so everybody could work together in a productive way and not create false expectations. Suffice to say they didn't really listen to me and grossly underestimated the tasks. I'll probably sound like a stuffy aristocrat moron again but the number of people working on OGP who know what they're doing is 0.5. By the time they've reached their target of "between 20 and 30 FPS on Quake III at 1280x1024", Mesa will be running it at 40 FPS on an average CPU. If open graphics drivers were really so interesting you'd expect OGP to do much better.
I stand by my point that the most capable engineers already work for an IHV, and those who don't, should. Open drivers won't help us further. Standardization, rigorous conformance tests and a catchy label will. There used to be a time when people asked for an OpenGL capable graphics card, now everybody is asking for a DirectX 10 capable card. All our hopes should be on OpenGL Long Peaks and Mount Evans, and the surrounding kernel components...
I don't think there is an apolitical, neutral ground inside technology.
Silent_Buddha
Legend
Should be amended to be "...inside technology, on forums."
Especially with Nvidia zealots, ATI/AMD zealots, Intel zealots, Apple zealots, Linux zealots, Windows zealots (is there such a thing?
), and the list goes on an on.Meanwhile the majority of the people will try to sift through all of that for useful bits and pieces.
I'm with others on this thread, I just don't see it happening due to a combination of factors. IHV unwillingness to share the underpinnings of their hardware that might lead to a competitive advantage from their competition and the fact that any graphics driver unless tightly controlled and targeted at a specific Linux release is going to be a big headache for said IHVs in the amount of service calls they'll receive as their hardware "doesn't work" on X flavor of Linux.
Regards,
SB
I agree with you as far as this. The ones who can probably do and the ones who should probably should, but I still think it likely that the budgets for driver personnel are probably short a few steps of hiring every person who could possibly be useful to driver development. You're probably right about driver development being too complex at its deepest levels and you surely know more about it than I do. However, I'm sure that an open source project could come up with useful stuff. There are legitimate 3rd party improvements to drivers already. You might not find them useful to you but I have needed them several times in the past. In general I'd say that the general public is far more considerate regarding not closing off options useful to some, and not breaking so much to fix problems, than the IHVs are. However, I'm sure it would not be some panacaea, nor some wildly successful project in the vein of Firefox. Also, I'm sure the IHVs aren't interested, so, I'll quit now.
Captain Chickenpants
Regular
On a related note, we have trouble finding people with the right skill set.
Especially new graduates. It seems a lot of Universities are switching to teaching Java or other higher level/ more abstract languages.
When it comes down to drivers you are generally looking at pure C sometimes with some assembler thrown in.
CC
Especially new graduates. It seems a lot of Universities are switching to teaching Java or other higher level/ more abstract languages.
When it comes down to drivers you are generally looking at pure C sometimes with some assembler thrown in.
CC
This is actually an interesting topic. A few years ago here at the UofM there was something of a civil war going on in the CS department over a similar topic. The problem is that once you factor in all the general requirements, there isn't enough time left in a 4 year degree to teach people how to be good software engineers and how to be good computer scientists. The smart ones got enough of the basics of both to go on in whatever area they decided to pursue. On the other hand, there were a lot of people graduating that didn't really understand either.
Nite_Hawk
People who really know how computers work (ie: they experienced the history) are dinosaurs nowadays: 99.99% of them are in management or other jobs that pay more than programming. The new ones only know about the outside and fluff.
Politics. There's more job opportunity for Java programming, so that's what they (have to) teach.
In my opinion schools should start with assembly, then C, then higher languages. Only then will the students fully appreciate the abstractions, but still know what they're doing at the lower level.
Last edited by a moderator:
You are way past crazy. Dropout ratio in CS is already around 50% for first-year students anyway at most US universities, and that's with Java. Switch to asm first, and you're going to hit 99%.
When I first started programming, you needed to look inside the box, find the chip that performed the function, go to the library and find a book that covered that chip in detail, experiment with writing to and reading from it and write your own driver to be able to do something. Or see if there was a BIOS or OS function that did the same.
If you wrote that program while sticking to the conventions, it worked from DOS 2.0 up to Windows ME without any modifications. And the programs I wrote for the win32 API or NT4 still work today.
If I tell other programmers or managers nowadays that "I'll simply write an OS for that machine", they freak out. Because they think writing an OS is the hardest thing they can imagine. "Windows Vista consists of billions of lines of code, and took 10 years to write, by the best and biggest IT company ever!!!" While writing a simple OS is actually very simple.
If you wrote that program while sticking to the conventions, it worked from DOS 2.0 up to Windows ME without any modifications. And the programs I wrote for the win32 API or NT4 still work today.
If I tell other programmers or managers nowadays that "I'll simply write an OS for that machine", they freak out. Because they think writing an OS is the hardest thing they can imagine. "Windows Vista consists of billions of lines of code, and took 10 years to write, by the best and biggest IT company ever!!!" While writing a simple OS is actually very simple.
So, who is going to do the low-level work?
Similar threads
