Slashdot Mirror
SGI and Mesa on Linux/OpenGL Base
An anonymous reader sent us an interesting linke over to SGI on their new
Linux/OpenGL Base. Its a project to standardize OpenGL on Linux, but
Mesa Compatibility is actually part of the plan. Nice strong Linux push for one of SGIs most important technologies. Very cool. Also, but not linux related, is their new Reality Center. Dibs on one of those "Walls".
No worry about FreeBSD. The Mesa/Glide combination has been in the ports collection for some time now, and we have the glx from openprojects.net working too, yielding accelerated 3d for Matrox G200/G400 and nvidia RIVA128/128ZX/TNT/TNT2 etc.
Grab it here:
http://www.freebsd.org/~3d/distfiles/glx
The Direct Rendering Infrastructure for the upcoming XFree86 4 release will get ported to FreeBSD too.
And theres is more. Please subscribe to
freebsd-multimedia@freebsd.org
to participate in the discussion of this exciting subject or send me an e-mail to 3d@freebsd.org if you have any questions left.
Add to that support for the Matrox G400, NVidia TNT2, ,,
See my comment above, on how to get it. Please use it and help testing!
Bet on it. If the information get freed it is just a matter of time that it will be implemented.
This is EXACTLY what most GNU / Open source advocates are looking for. Here's a software standard that makes software easier to write. Who cares what OS / Hardware you're writing for? It's all a standard, just re-compile for the other platform.
Yes but who gets to define the standards?
My general observation of the computer industry suggests there are several variants of what are commonly called "standards" (oh and toss in the word open somewhere to really confuse matters).
- evolved standards, resulting from a bunch of people trying to solve a specific problem and iterating towards a solution (e.g. Gl->OpenGL->Web3D?)
- committee standards, a technical specification agreed ahead of time (e.g. Fortran, IETF, W3C sorta)
- defacto standard, an implementation that so completely fills a role/niche that it is universally adopted (e.g. NFS)
- imposed specifications, a base implementation released and through fortunate timing or market dominance, is defined as a standard for conformance
The problem is that it is extremely hard to create universal standards that work well the FIRST time. Usually at least two or more implementations hitting real-world bottlenecks are necessary to define a flexible standard that covers enough variances of the problem space for it to be widely adopted. Plus major issues in allowing it to evolve over time without any preconception of future technologies (witness Fortran 66, 77, 8x, 90, 95, etc). This learning curve puts a serious crimp on progress, especially as the hardware tech evolves faster than software development.
Anything that gives CHOICES is OK in my book ...
I would qualify this by observing that it takes bloody-mindedness and sheer Machiavellian talent to get any largeish group with a multitude of self-interests to adopt a common code of practice (see legal conformance or tussels with metric/imperial measurements). Choice of standards that have no external force or non-biased referee insisting on interoperability leads to fragmentation and slower adoption. Combined with the pressures of commercialisation and the prospect of dominating a profitable market, good technical standards can often be distorted, subsumed or replaced by plurity of suboptimal choices.
As they say, standards are good, lets have lots of them.
LL
For my part I would be already glad, if they would open up their electropaint screen saver sources. :)
Fun aside, please remember that SGI is undergoing a big change. Having a Silicon Graphics workstation has been a bit like having a Mercedes-Benz car. This was very capable (not always the fastest, but always with stunning graphics and that special flair) and expensive hardware, and the software was not what a PC user would call cheap either.
So they need to make the transition from relativly low volume/high priced to high volume/low priced.
Once the demand is there, the software will follow. In some months from now we have really powerful 3d on the free operating systems, people will take it seriously and consider using these platforms for serious work, asking for serious tools. It will happen.
This is EXACTLY what most GNU / Open source advocates are looking for. Here's a software standard that makes software easier to write. Who cares what OS / Hardware you're writing for? It's all a standard, just re-compile for the other platform.
Yes but who gets to define the standards?
My general observation of the computer industry suggests there are several variants of what are commonly called "standards" (oh and toss in the word open somewhere to really confuse matters).
- evolved standards, resulting from a bunch of people trying to solve a specific problem and iterating towards a solution (e.g. Gl->OpenGL->Web3D?)
- committee standards, a technical specification agreed ahead of time (e.g. Fortran, IETF, W3C sorta)
- defacto standard, an implementation that so completely fills a role/niche that it is universally adopted (e.g. NFS)
- imposed specifications, a base implementation released and through fortunate timing or market dominance, is defined as a standard for conformance
The problem is that it is extremely hard to create universal standards that work well the FIRST time. Usually at least two or more implementations hitting real-world bottlenecks are necessary to define a flexible standard that covers enough variances of the problem space for it to be widely adopted. Plus major issues in allowing it to evolve over time without any preconception of future technologies (witness Fortran 66, 77, 8x, 90, 95, etc). This learning curve puts a serious crimp on progress, especially as the hardware tech evolves faster than software development.
Anything that gives CHOICES is OK in my book ...
I would qualify this by observing that it takes bloody-mindedness and sheer Machiavellian talent to get any largeish group with a multitude of self-interests to adopt a common code of practice (see legal conformance or tussels with metric/imperial measurements). Choice of standards that have no external force or non-biased referee insisting on interoperability leads to fragmentation and slower adoption. Combined with the pressures of commercialisation and the prospect of dominating a profitable market, good technical standards can often be distorted, subsumed or replaced by plurity of suboptimal choices.
As they say, standards are good, lets have lots of them.
LL
Quake 2 is a pretty good example ;).
"Pray arm me further by your reply" Winston Churchill
Web3D? You should have a look here:
http://gel.urstudios.com/
The problem is that it is extremely hard to create universal standards that work well the FIRST time. Usually at least two or more implementations hitting real-world bottlenecks are necessary to define a flexible standard that covers enough variances of the problem space for it to be widely adopted. Plus major issues in allowing it to evolve over time without any preconception of future technologies (witness Fortran 66, 77, 8x, 90, 95, etc). This learning curve puts a serious crimp on progress, especially as the hardware tech evolves faster than software development.
Well spoken. It might interest you folks that one of the main developers behind Direct3d has another 3d API out of the door that is a nice C++ library for Win32 and Unix:
QDraw
It will be used for the above mentioned Web3D format. Works very nice under FreeBSD. Alas I recommend hardware 3d acceleration (I use a TNT) for some of the samples. Backends are OpenGL and our special friend aalib!
We owe much to id Software, without them OpenGL would have not been established widely on Win32. Now they have OpenGL, we have OpenGL and game publishers and other software companies will have it easier than ever to create powerful applications for both platforms.
Further it is a joy to have John Carmack working on the open glx Matrox driver. This guy is dedicated to the subject and certainly not only money driven.
Linux is uniting *NIXes. Solaris and FreeBSD already have the means to run some Linux programs, this makes Linux binaries "almost" a default format.
Well, as a former OS/2 user I have very mixed feeling on this subject, as OS/2 emulation for Win16 was so good that nearly no native app was created, in the end the death for that operating system.
FreeBSD's Linux emulation is very good. So you can run Netscape, RealAudio Player,Myth2 demo, Quake3 under that emulation layer. Like in the case of OS/2, Linux apps might actually run better under FreeBSD in some regards. You even get the same Linux headaches (like clib/glib madness) under the emulation. :-)
However it is just an option and not the native ABI and we will strive to keep porting and developing for FreeBSD. So we have native hardware acclerated XFree86 (2d) and Mesa (3d) versions and apps.
Drop me a mail at 3d@freebsd.org if you need more information.
Of course the basic idea of the effort mentioned in this Slashdot article, interchangeability of OpenGL libs is important for FreeBSD as well, and we will try to find an appropriate solution and join discussion. Maybe it will differ in some aspects (our shared lib versioning scheme is different than the one from Linux, so we might not be able to use a solution one to one) but I believe it will not be too far from it away.
How the heck are you suppose to make a display subsystem implementation platform agnostic? If the psycho behind OpenBSD wan't to model OpenBSD's OpenGL after Linux then more power to him.
is there any pc hardware that takes advantage
of opengls 2d accelleration (glDrawPixels etc)
i know sgi boxen are good this way...
It just does their customers no good for them to port to Linux today. Their customers won't buy it today. Customers need a stable environment that works. Although we're starting to see implementations that work, stable is not a word I'd use to describe them.
Have patience. I'm convinced it'll happen. It is moving along at a fairly rapid pace. If vendors don't have products out by next SIGGRAPH then I suspect you'd have a real reason to complain.
- |Daryll
I forgot to mention in my other post. If you are having problems like that with the shutter glasses it may be because they are not set up correctly. I dont know what kind you are using but when we switched to the IR emitters from http://www.nuvision3d.com our whole setup worked better. These use a tiny fraction of the power that the StereoGraphics crystalEyes glasses we had before used. If you are using the IR emitters make sure you have more than one. Buy enough to make sure the glasses always recieve a signal no matter where the user is facing.
Be insightful. If you can't be insightful, be informative.
If you can't be informative, use my name
Well during AWGUA 99, they said (Mark Sylvester) that depending on the user response, they may port their apps to Linux. So if you Alias Wavefront licenses are current, bug them about it. There was a loud cheer just when they announced the availability of the Maya and Composer batch renderers. Though, I think realistically there won't be any serious consideration until Xfree 4.0 and DRI are included in the distros. These apps need some serious hardware OpenGL accelaration.
Still the apps are coming. Houdini 4.0 will run on Linux (using the Xi OpenGL I believe). There is Blender which could me called a mid-level package, and there is the roto and ras_track from Hammerhead. So it might just be a matter of time. Hopefully by next SIGGRAPH will get some real good news. Here is a link to the OpenGL BOF for reference:
Openl BOF minutes
the LSB uses the convention libfoo.lsb.N where the
.6 in libc.so.2.0.6)
lsb.N reflects the version of LSB in use. However, it's not clear we should be tied to a particular version of
LSB.
WTF is this? Why does LSB screw up library names like this? Where does the version of the library go? What about library bugfixes? (like the
slightly offtopic: Has there been any progress on a Java OpenGL API standard? I recall news about perhaps magician becoming the standard, but no word since then.
My Java Quake-wannabe development is suffering :)
The Real Magic is in the software - the CAVE Library automagically manages all the walls and headtracking for you. When you write a CAVE program, basically you just give the library a callback that draws the world, and the library will call it each time it needs to draw a wall, setting up the correct projection magic so it's drawn properly to your head position. It's really
pretty slick.
I've been playing with the idea of designing my own CAVE type setup using a group of 4 to 6 cheap networked PC's, each with a 3D card and an LCD projector. It doesn't seem like it would be that difficult to do; has anyone tried this? If not, are there any "gotchas" that would keep it from working? The only ones I can think of are possible network delay desynchronizing the walls, or perhaps the view-positioning math is harder than I think it is...
I don't care if it's 90,000 hectares. That lake was not my doing.
Well, A|W can't port and release their products on Linux until there's a decent OpenGL implementation there. (No, I'm afraid mesa doesn't count -- a software only GL is *way* too slow for the stuff Alias does)
Doesn't Windows have OpenGL drivers provided by each video card maker? For example Nvidia vs 3dfx?
Yes. SGI's GLU implementation is written in C++.
I would like to see linux-like portability and ease of use with SGI's outstanding cases...
"Be nice, veer left, and never stop thinking" Iain Banks - Walking On Glass
Right. This is the same idea as the LSB, but targeted specifically at OpenGL and trying to work on a faster timescale, since 3D acceleration is starting to happen in a big way on Linux.
What you've described is a CAVE. They're not $50 million, more like $250k-$500k, depending on how decked out your SGI is.
They're actually pretty simple - you get a couple of real good projectors, some big pieces of plastic film to use as a screen, and just project onto them. Then you get a magnetic tracker (they're not that much) and maybe a little joystick-like thing, and you've built a CAVE.
The Real Magic is in the software - the CAVE Library automagically manages all the walls and headtracking for you. When you write a CAVE program, basically you just give the library a callback that draws the world, and the library will call it each time it needs to draw a wall, setting up the correct projection magic so it's drawn properly to your head position. It's really
pretty slick.
The big bummer with the CAVE library is that it's pretty expensive - about $30,000 (educational and with-hardware discounts are availble, though that drop that to about 15,000) It's availble from VRCO. There isn't any free software that does anything quite like it out there...
-Erik
What do you mean pseudoparanoid? Why is a reality wall any worse than a TV?
"Be nice, veer left, and never stop thinking" Iain Banks - Walking On Glass
GLboolean represents a yes/no value. It has to be an addressable quantity since you can pass pointers to arrays of them to some GL calls. The smallest addressable quantity is a byte, so that's a reasonable choice.
Anyone else noticed that nVidia is the only graphics card manufacturer to cooperate on the project ?
nVidia is going to be the way ahead for Linux graphics acceltating (well that is apart from SGI of course).
now, our community is splintering with announcements like this
Please calm down and have another look. There's almost nothing here that other free unix OS's can't use. Your charges are laughable. There was no implication in this that any code will be non-portable.
No part of this says it will not work on FreeBSD/OpenBSD/AcmeBSD. What's Windows got to do with anything?
Do we really want people like this harming our comminity?
The word 'harming' is ridiculous. In fact, it shows no respect for the many well known developers who are (and have been) writing this software FREE. What is your agenda?
You haven't lived till you have played DOOM in one of those things. We have one here at school that they use for among otherer things showing off how nifty our schools computer resources are. And one of the things they like to show off is a doom port... Its really killer. I loved it, and I had no problem with the 3d glasses, even though they were the old RF kind. The RF joystick was a bit strange to get used to, but once you get the hang of it, it kicks ass.
mov ax, 13h
int 10h
Its not enough to say that things can be used for other OSes if you want.
The intent up front should be to provided an OpenGL on multiple OSes. Why is it that all the Non RedHat people get all excited about releasing products "for RedHat Linux"? Becuase if there is no intention of helping other platforms then it is more lilely that they will fall to the side in the future.
I just do not want to see people start to develop for specific platforms. I think that it helps all of us if we develop for Free Software.
-jason
Perhaps an Apple G4 cluster running Black lab Linux?
Why not?
TML
Chill. This has no effect on portability of code to BSD. No code will result from this (other than maybe a simple test suite to verify that the libraries and headers are in the right place and export the right symbols). It's just a way to organize libraries and header files on a specific OS. Mesa, for example, is not going to be any less portable to FreeBSD because it puts the libraries in one place as opposed to another when built on Linux.
I suspect that commercial interests will skip supporting things other than Linux (which is why I'm against binary only stuff!) and the open source people will work at making sure (or will have someone come along that will) that it works on anything anyone is interested in.
XiG and MetroX may not be interested in doing their stuff for anything other than Linux- but I know that the GLX acceleration project has at least one FreeBSD person working on the Matrox G200 support for *BSD.
Add to that support for the Matrox G400, NVidia TNT2, and several others I'm not quite yet at liberty to mention (Work on these chips have just begun, and I don't want to get false hopes up!)- all with open source code and no black magic that would keep someone from implementing it under *BSD, etc. under any chipset that will work with an adapter.
I am not merely a "consumer" or a "taxpayer". I am a Citizen of the State of Texas
Ok you have mesa trying to be OpenGL compatible and you've got OpenGL trying to be Mesa compatible now. Why don't the SGI guys just beef up OpenGL support in Mesa?
~GoRK
Not knowing otherwise, I'm wondering if one has to pay $$$ to license OpenGL from SGI to implement on Linux? Or are they releasing it as free/open source software?
Without knowing that status, while it seems cool that the "Official" OpenGL is making it's way to Linux, it'd be a shame that it couldn't be included in all the distros.
After using both linux and SGI in an academic setting for many years, I can envision linux-like portability and ease of use with SGI's outstanding graphics in a single system. One can only hope.
-- Moondog
hmmmm... i would think that those walls, with a little bit of creative flooring would make the beginnings of a HoloDeck...
wouldn't THAT be a hoot? i think i would go mad, though; surrounded by my dream(s), but not being able to directly interact with them....
When politicians are involved, everyone loses.
This is EXACTLY what most GNU / Open source advocates are looking for.
... after all, if you're selling hardware whose claim to fame is best-of-breed OpenGL performance, wouldn't you want software to be available for it?
... but if the platform doesn't matter, then the best software will be more affordable, since it will sell better ( being the best ).
...
Here's a software standard that makes software easier to write. Who cares what OS / Hardware you're writing for? It's all a standard, just re-compile for the other platform.
This will help both developers and companies like SGI
A lot of software is made for x86 ( Windows) because it's a popular platform, and thus a large market
Anything that gives CHOICES is OK in my book
The fact that no one understands you doesn't mean you're an artist.
i'm gonna withhold my happy dance until SGI agrees to port Alias and their other tools to linux instead of just using linux boxen for rendering. what good is spiffy OpenGL support in Mesa if there's no decent way to create content to display with it?
The picture of the Reality Center wall is how I've always invisioned the "Geek compound" or whatever Rob calls his dwelling. I can picture CmdrTaco and Hemos standing in front of their wall, viewing a graphical represenation of current Slashdot traffic, using voice commands to config httpd, ban trolls, and stories, etc.
Somehow, though, I suspect reality is much less sexy than my imagination. ; )
when you can have a whole room?!
THAT seems just like the Holodeck to me, except it's too small (1-5 people?)
Now if only I had $50 million dollars...
Glückwünsche, haben Sie Slashdot ermordet, indem Sie zum korporativen Druck beugten und Subskriptionen einlei
I was in one of these immersive environments about a year and a half ago, and they are extremely cool. It is hosted at Virginia Tech (Balcksburg, Va.) and they have public open houses if you are nearby. My program (Architecture) was experimenting with Virtual Environments, and I am sorry to say I graduated before I got to see the facility come completly online. It is a shared facility, so multiple programs can schedule to use it. If you ever get a chance, check them out. Too bad they are way out of budget for most buisnesses that could use them. Here is a URL if you feel like checking out a real world application of these systems at Tech,VELab and one for the CAVE.
Who is the master of foxhounds, and who says the hunt has begun? -Pink Floyd
I was reading the standards, and noted that GLboolean was 8 bits. OK, so this is C, not C++, and 8bits is probably the smallest efficient size, but this adds a new demension to "fuzzy logic."
I was kidding, really.
Actually, the C++ standard provides the bitset container for this. Certain classes (such as iostreams) do require a large number of booleans. Looking at my compiler (Metrowerks) it appears can use less than one byte per bit, although other implementations may vary. Of course, OpenGL is not written in C++.
I worked at the UIC EVL lab for awhile doing some programming. Never looked at the ImmersaDesk (which was also developed there) but the CAVE was simply the most amazing 3D I've ever seen. Having 3 huge walls plus the floor getting projections along with the LCD glasses was totally amazing. I know exactly why people get excited....cuz it kicks ass.
We have 2 of them over here with immarsadesk/glove/3dshutter (think reverse projected TV set type things with a 12 foot screen and wireless control for a room full of 3D shuttered battery powered glasses)..i *dislike* them intensely. They give people a headache from the 3D effects, very uncomfortable to work with due to limited RF range (if the glasses dont sync correctly you see dual fuzzy images which give splitting headaches). I fail to see what the fuss is about..probably by people who never use one daily. just cause a technology looks cool doesnt mean its useful...or worth using on a daily basis. im more productive with old fashioned SGI 20 inch monitors at 1600 x 1200 32bpp than with a reality centre/immarsadesk screen.
Yes, they do have drivers for each card that is different.
But OpenGL is NOT a driver. It's an API that is used to write 3D rendering applications with. It doesn't concern itself with drivers. Under Windows, there's only two different OpenGL layers to muck with- the Win95 one and the WinNT one. With Linux, you've got at least twice to three times as many. What they're asking for here is to make the OpenGL implementations identical as far as calls, behavior, etc. and to provide as common as is possible driver layer API.
I am not merely a "consumer" or a "taxpayer". I am a Citizen of the State of Texas
It's a document written by a bunch of concerned OpenGL developers/users that attempts to standardize such mundane (yet important) issues as:
- "What is the OpenGL/Mesa library *called*?",
- "Where do the header files live?",
- "How does a BINARY program figure out what OpenGL extensions there are?"
Don't read into this more than there is!I see this as a good thing for Linux. Means that hopefully they will moake ports of the rest of there stuf that uses OpenGL to Linux too. I hope that this is not just a promise and they follow thru.
SGI technology in Linux can be good, it will hopefully get more people to embrace Linux, and improve Linux as an OS.
Linux is uniting *NIXes. Solaris and FreeBSD already have the means to run some Linux programs, this makes Linux binaries "almost" a default format.
I am looking forward to next year and the things to come for Linux, and hopefully more features added in in the 2.4 kernel. I wonder what 3.0 will be like and if it will make it that far?
Only 'flamers' flame!
this is in section 2.2 regarding assembly.
does anyone
actaually use C++ to implement low level libraries
like this? (id call it device driver, but the
example refers to GLU which does not
interact with hardware)
seems like this leads to many problems.
PC based CAVE like displays are just starting to be developed. PC graphics are now becoming fast enough. The walls need to be gen locked or synced which is a problem with current PC hardware. This can be solved by interlacing the frames so left eye and right eye are alternating lines. This cuts the resolution in half but emmitters are not required. LCD projectors I dont think update quick enough yet but many other projectors work. Many Caves have used PCs for the tracking. The math and rest of the software is not too much of a problem.
I'll take the bait and respond to this troll.
How is this any better than the Microsoft "Will only work on MS Windows" philosophy?
Umm, this is an open standard which anyone can use. If I wrote an API for Linux made the it an open standard would I get flamed for not writing the BeOS and BSD versions of the API? It's a public API and how it works is well known. If you want it for BSD or BeOS do it yourself....
SGI has only made it easier for people or companies to make openGL API's on other OS's.Be insightful. If you can't be insightful, be informative.
If you can't be informative, use my name
Does this make the world look more and more like the world of Fahrenheight 451? I posted this once before when sony announced it's robotic dog, and there were stories of robotic fish that were indistinguishable from real ones (except for the eyes). I state it once again. If (and when) this comes to the mass market, what next? Western populations are apathetic as it is. People dont read as much as they used to, relying more and more on the boob tube to provide sensory deprivational entertainment. Even the books that people DO read are hopelessly bland, PC, and uncontroversial ones like "Chicken soup for the soul). What is happening to this world????? Just some thoughts from a pseudo paranoid. -Laxative
What school is this?
Cooperation between all the Free Software camps is an important part of our world. All the major Free projects strive to work on all FreeUnixen: Netscape, Gnome, Mesa, etc...
But now, our community is splintering with announcements like this:
Operating systems other than Linux. BSD etc. are welcome to use whatever comes out of this project, but we are explicitly not trying to solve this problem for every free OS in the world.
To all of you who have said how easy it is to write portable code and to make it run on all platforms, they are basically saying that you are wrong. Do we really want people like this harming our comminity? How is this any better than the Microsoft "Will only work on MS Windows" philosophy?
Linux is part of something much larger.
-jason