Slashdot Mirror
SGI Open Sources GLX
An AC was the first to inform us that Silicon
Graphics has released
GLX as Open Source for Xfree86
in order to stimulate the number of hardware-accelerated 3D
drivers supported by Xfree86. Interestingly Red Hat and
SGI are funding new
driver work
to be done by
Precision Insight
on a multiple pipe 3D architecture extension for X. Precision
Insight were the people who brought the NeoMagic drivers to
Xfree86 under contract to Red Hat. Of course it comes with its
own
license: anyone care to comment on it?
Grab the code here.
Not only is this great for gaming, but also rendering, 3D-CAD, etc. Anyone ever played with 3DTop under Win9x? Nice concept. Imagine a mix of that with E or WM or KDE [guess who's not playing favorites here ;)]. A whole new category of Window Managers, 3D ones. I'm sure someone would want to play with it, i know I would.
Go SGI!
I guess I won't dump the g200 for a 3dfx flavor after all. There is hope...
Smart move.
Now how about giving Brian Paul a little help, so that he can make Mesa into a fully OpenGL-compliant API.
If only /. could influence nVidia ( of the Riva TNT fame ) into supporting their cards ( 3D side of the entire Riva family ) on X for Linux. That would bring a smile to my face. :^)
:-(
I'd be willing to help with the coding as well. I wonder how many other would also? I have asked nVidia but no response except "ask your card manufacurer".
Do commercial UNIX(tm) implementations of X have
it?
aren't they the ones who obfuscated the code?
drnkn mstr
The GIMP and the GNOME canvas both provide anti-aliasing by drawing fonts to a buffer, smoothing the buffer, and then drawing the buffer to the screen. This is expensive, and not The Right Way.
Perhaps XF86 4.0 will have an anti-aliased fonts extension. I don't know whether or not other implementations already do.
This is great news. I just get the feeling that the more that gets added to X, the harder it will be to give X the rewrite I think it desperately needs.
great. all i need now are multihead drivers for xfree86 and i will be set
Dont forget BANSHEE for us poor bastards who cant
afford TNT, banshee is still fast!
www.3dfiles.com has some awsome 3d screen savers for windows, that Comsmos-1 one is awsome, which X ahad that
-Cheekyboy
Is there a Mesa driver for it, or just proposed?
I agree with your agreement!
Free GLX, Creative Labs writing linux drivers for all its hardware, Civ3 for linux? I see serious change on the horizon...
I am, absolutely! I am now close to accumulating enough CAD tools to seriously siggest migrating some of my employer's CAD applications to Linux.
/. password AGAIN :-( ]
While there will doubtless be some that will just "flame away" (the Win95 teenage Rambos that frequent this site), this is a fantastic win, and something I wouldn't have dared dream about a month ago when I first built and installed Mesa.
I'll be doing a test build this weekend on 3.3.3.1
Looks like my ATI hardware 3D will be useful after all.... Ya-Hoo!!!!!!!!
License terms look resaonable, too, at a casual glance. This is a BIG BIG win, that a lot of folks may not yet fully understand, but they will...
Paul
reichp(at) ameritech.net
[darn -- lost my
I wrote an email to Nvidia to encourage them to either do something or to release the specs. However I have never gotten a reply. :-(
Still.. It is true the SVGAServer truly rocks with the TNT.
The Falcon
Woohoo! This is great news, people. I was so shocked I almost peed my pants.
This is great! one of the reasons I chose linux was to do some 3d stuff and programming. I almost left linux when the graphics weren't as good as windows. I was shocked. NO! MICROSOFT CAN NOT BE GOOD AT HIGH PERFORMANCE GRAPHICS! MS SUCKS AAT ANYTHING THAT SUPPOSeD TO BE FAST AND HIGH END!I was afraid that NT was going to kill everything. I was going to write a grahics kernel driver and have it run in kernel mode but only if the user mode counter part (which would run on top and comminicate directly to the kernel part) detects that it doesn't operate with the hardware directly. This is how NT 4.0 works with graphics. NT 3.51 did all the rendering in user space outside of the kernel and it was terriblely slow and linux works like this today. :-(
With these improvemnets I can finnaly get NT off my system and stop hearing all this games and graphics fuss by the ms lovers who are trying to kill every workstation with unix on it at work. I hope this will silience the unix critics who all state that ms writes the best and tightest code in the world. Just look at quake and 3dstudio max. yada yada yada.
All backpeddling and realism aside, this is a big
possitive step for linux. God bless SGI!
This is the best news in a long time! Now I don't mean to be impatient, but when when when?
The century ends in 1999, but the next millenium doesn't begin until 2001?
This is good news. Good work SGI!
The only thing that troubles me is the increasing number of OSS licenses. It doesn't matter as long as they mix well, but do they really? Can you mix code from GPL, MPL, QPL, EPL, GLXL, BSD, Artistic? It is my understanding that you can't, i.e. you can't take pieces of code from programs with different licences and put a new together, because both pieces would require you to publish the new software under their respective licence. You can't merge two projects with different licences (the whole KDE debacle) etc.
This means that while we do have lots of OSS software, we still always have to be very careful about how we mix and match it. I would be happier if there was only one licence I'd have to care about. But I guess you can't have it all.
Not being a lawyer, I'm confused by these matters. Anyone care to explain?
This is such fantastic news I can hardly contain myself. I didn't think it would happen so soon... in fact, I was pessimistic about it happening at all.
Thank you SGI, you have made the right decision. Go SGI! Go RedHat! Go Precision Insight!
One step closer to World Dominatation. Who else is going to join the Linux Army in the battle that we will undoubtably win?
Cheers
AndyM
........ and have it monitor the forums for any mention of the word "millenium" (where it will point out that it actually starts in 2001), any mention of the word "gender" (in wich case it will point out that "sex" is the correct term), any mention of "gnome" (in which it will point out that KDE is better), and mention of "KDE" (in which it will point out that gnome is better), any mention of Linux (in which case it will point out that *BSD is better), any mention of RedHat (in which case it will point out that SUSE/Debian/Caldera is better), any mention of SUSE/Debian/Caldera (in which case it will point out that RedHat is better), any question marks in place of quote marks (in which case it will point out that the poster should not be using microsoft products), any grammatical errors (in which case it will point out that the poster should go back to third grade), and in the unlikely event it doesn't find any of the above it will make some sort of winging complaint that the story is irellevant or there are two many of that type of story.
I think it will save many slashdot posters a lot of trouble.
It's not good enough because GLX isn't free. It's not good enough because GLX isn't free. It's not good enough because GLX isn't free.
Oh wait, GLX has the Gnomer seal of approval so the standard whine doesn't count.
Yeah, it's real simple. If the wording of an actual license is totally irrelivant. What you need to look for is the Gnomer Seal of Approval. See, it's fine to use Netscape and MySQL (both with restrictive non-GPL licenses), because the Gnomers say it is 'k001' to do so.
However, don't use Qt because Qt isn't free. Qt isn't free. Qt isn't free.
> So why aren't the SANE scanner backends > integrated into the kernel? They send whatever
> scsi commands they want to a scanner via the > scsi generic interface. surely we can't allow > that in user-space!!
Ah, but the SANE backends aren't beating directly on the hardware I/O ports and other fun things like the XFree86 drivers do. In the case of the scsi drivers, a relatively raw (but controlled!) kernel interface to the hardware is used.
> unix is good precisely because the all the > functions aren't integrated into the kernel.
> because the kernel implements a simple safe > interface to hardware, and leaves the rest to > userspace.
That's quite true. But XFree86 has no such interface to use, unless you have a server that's willing to run over top of fbcon and/or KGI.
> ie simple framebuffer that can do some simple > functions like change video modes, reset the
> card, etc. And leave the accelerated and 3d
> graphics to X or *whatever*.
If the hardware is designed well enough, you can get away with that, but not all hardware (a good example would be the S3 ViRGE) is.
> If the whatever cocks up, the kernel can always > recover.
That's not always true, especially on poorly-designed hardware like the ViRGE (alas, all too much video hardware is not entirely "safe", unless you're blessed with SGI produkt).
Very little video hardware is possible to reset to a known state (blindly, at least), and in some cases just tweaking accel registers on a poorly designed card can do fun things to the bus, from which the kernel can physically not recover. _in those cases_, you're going to _need_ (you do care about stability, right?) a _thin_ kernel layer relating to acceleration that basically keeps track of state and does some limited sanity checking of the accel accesses.
That's basically what GGI's KGI and some of the better fbcon acceleration proposals are about.
For the most part, they don't involve an abstraction layer (and since they're in-kernel they shouldn't try, for something as complex as graphics accels!). For "safe" hardware, they all are supposed to basically provide a pass-through (i.e. direct mmap() to userspace for MMIO) for the use of the "whatever".
Even in microkernels, the little kernel that's there is still sits on the hardware, protecting it from direct (i.e. i/o port level) access. Think about that -- it's done for a reason.
Hullooouu!
You people are dumber than snot.
The twentieth century starts 1900 - as there is no year 0, but right after you are born, you are in your first year.
The nineteen-nineties go from 1990 to 1999.
The twentieth century goes from 1900 to 1999.
The second millenium goes from 1000 to 1999.
The third millnium goes from 2000 to 2999.
the twenty-first century goes from 2000 to 2099. Once the clock ticks over from 1999 to 2000, you are in the 21st century and the third millenium. Party like it's 1999.
Very, very true.
> true. but what is wrong with a *trusted* process
> like X accessing raw i/o, or a
> nearly raw kernel interface? the only
> thing we need to do is to make it safe.
SUID's are _very_ hard to write securely. You can
do much better job in kernel, because it's not
run by the user. And don't forget, that there's
not X graphics only.
Briefly, kernel should sit on the hardware and only allow safe things to do.
Edheldil
I have an Onyx and work and am thankful for it. There are e few problems with it though. IRIX is a propriatary and deficient unix. It is hard to administer, doesen't come with a full suite of programs unless you pay a LOT of money.
Right now our lab is looking to buy some more machines. It will probably be more linux boxes from VA Research(or some similar place). If SGI were to ship their visual PC's with linux including X support and a full gl implementation we would probably snap up several.
We will not bring any NT boxes to this lab becuse of software costs, security and lack of stability. Most people don't know that all NT upgrades have to be checked out and recoded for their machines by SGI. Therefore they are going to be released some time after their release for regular NT. This may lead to problems with lack of hotfixes and delays in fixpacks. Linux could change this.
With linux I would know that this machine would still be useful in 5 years. I would not be at the mercy of SGI when they drop support for this machine. Any fixes that need to be made can be coded be me if need be. The kernel has support for the machines. Now if SGI was to support X, gl and tweak things a bit, they could create a new market for their machines.
Piotr
The QPL was intentionally put forth in an attempt to make extremely murky the meaning of Open Source standards, and the Opensource people took the bait. I said once before that in about 20 years nobody is going to be able to tell with absolute certainty what it is to be Open Source software because the QPL opened the door to the idea that non-GPL'd stuff == open source. By that time it will have gotten so murky and muddied that they'll be debating whether or not Microsoft End User Licences (EULA's) meet the "new, common sense interpretation" of open source!
Yes, I predict in a generation, those who stick to the GPL standard will be the digital kooks on the wingnut fringe of the Linux scene. Most of you guys are just here for the free software of the moment, and the Star Wars news, and you probably do not care about the ramifications of accepting software with strange, non-GPL licenses as open source. One day you'll remember what I said here when it all comes true and wished you'd heeded this warning.
The moral of the story? If it cannot accept the GPL - IT AIN'T OPEN SOURCE!!!! THROW IT BACK!!!
This totally makes Linux an order of magnitude :-| :-)
more viable for high-end 3D stuff... and that's
all i'm saying.
Chugging along slowly but surely is Berlin -- a replacement for X.
X Window System, by Jim Gettys
If it allows you to distribute patched versions, it's open source. If it puts restrictions on that, it ain't.
(Hey I'm listening and heeding!)
If only matrox would release the documentation for it... :/ :(
Info coming soon they have stated on their website for close to 6 months now
Don't forget that Creative Labs is planning to support 3D video drivers for all their products, including the Graphics Blaster Riva TNT. To find out more information from GGI developer Jon M. Taylor, check out his post to the GGI mailing list.
It's not a bad license. It's mainly a BSDish license, but no advertising clause; the license must be reproduced if the software is distributed in source form, and ALSO if the software is distributed in executable form. Great job SGI, this can really but put to some use!
Brian Fundakowski Feldman
I don't think there could be better news. Now lets get XFree fully OpenGL!
Doesn't Raster have something about buying dinner for anyone who could make some X calls use hardware acceleration? I call dibs!
Posted by Bill, the Galactic Hero:
First, their Visual Workstations aren't just standard PC's -- there's some thought in the architecture
Second, they really support Linux and OSS.
I was wrong about them. Buy SGI.
Supposedly, you can get primitive 3D support with a bunch of hacks starting here. It involves using ancient versions of XFree and Mesa, so I haven't tried it. However, you may be able to learn something about the card by reading the sources...
:)
I'm really hoping that XFree 4 will have 3D acceleration for the G200 card, but Matrox (who I annoy on a monthly basis about this) still refuses to release any information or write any drivers. The card is still really nice in X with 16M of video RAM, though
F0 07 C7 C8
Baloney. X is awesome. It does have its problems, but a "complete rewrite" is not in order. I assume you mean a "complete rewrite" of the X protocol definition since a "complete rewrite" of any particular server or Xlib implementation really doesn't mean anything.
X is proven technology that has been around for 15 years. I believe that the only people who are clamoring for things like a "complete rewrite" (and those who constantly bitch about X for one reason or another) are those who don't really understand it well enough to appreciate it.
I believe he was refering to X itself, not the XFree86 implementation of it in particular.
The concern that X is getting too big is valid. While it may be humanly possible to write a large, complex software program that's relatively bug free, no one has done so yet.
TedC
This rocks. This puts Mesa in a different position, though, that's for sure.
-o
Ok. I've noticed I've been happier with SGI
lately than I have been angry at them. It's odd.
SGI seems to be trying (hard or not) to do the
"right thing" and one often questions why the heck
they aren't taking the stingy backstabbing who
cares what we want route. I figure they'd be
a little more dedicated towards capitalizing...
I see three things:
1> SGI is really dumb and don't realize that this
could hurt them in the longer run. (they're not
exactly in the most financially stable state
right now)
2> SGI is doing this out of the kindness of their
hearts and doesn't seek any returns at all.
3> SGI sees this as a plan to profit, and in the
process gives us exact what we want.
Idealistically, it'd be the second case.
Realistically, it's the third case, which if it
is, has proven to me that capitalism really does
work for everyone! hooray!
Three cheers for SGI!
"We love SGI." :)
Isn't Redhat just so Eviiiillll......NOT
Make that a *big* pat on the back!
It seems to me that he did more than anybody to stimulate these developments, by writing Mesa. We should all be grateful for all of his very hard work.
The future sure looks bright!
-- Mike Greaves
Graphics in userland is the way to go for anything that may be used heavily in server roles. Unix has always followed this wisdom.
NT 3.5 was properly engineered. NT 4.0 is not.
-- Mike Greaves
SGI has made a good decision here, one that should help both 3D accelerated applications on top of XFree86, and SGI.
I'll be thrilled if I can purchase one of SGIs Visual workstations in the near future, install Linux on it, and be able to take full advantage of the graphic horsepower of the machine.
Thanks to everyone at SGI that made this possible!
...it started at about 500 or so. And the monk was off, so what he intended as year 1 was actually about -4.
So, it's just a number anyhow. Can we just party at big round numbers?
ati? i wouldn't count on it. not any time soon anyway.
according to ati, they are not releasing 3d drivers forn non windows platforms, and they consider their 3d implemetations to be completely proprietary information. maybe they would change their mind if enough people asked them, but i wouldn't count on it any ime soon.
other things aside, the ati's (at least mine) are also really good 2d cards for x, which is what i mostly care about anyway
If I don't put anything here, will anyone recognize me anymore?
Actually, had we started the calender at the year 0, the same way we normally count our age, then it would in fact be the new millenium on 1-1-2000. Does anyone know if that's the case? Or did the year number start at 1?
--
-- Knowledge shared is power lost. -- Aleister Crowley
Given that SGI has press releases saying that they are going to support Linux, something like this is needed for their new Intel machines.
This is a very positive development and I hope there is a driver available for my Matrox G200 in the future. I've been meaning to go through my OpenGL books, this will help inspire me to actually get Mesa and start working on some 3D stuff (haven't used gl in years).
the good ground has been paved over by suicidal maniacs
Four days after Dave asked for people interested in improving Linux hardware support to write more accelerated GL drivers in his mini-interview on happypenguin.org. What synchronicity. Here's to every video card we've got, including - Permidia2 - ATI RagePro - G200 - RivaTNT - i740 - and soon Voodoo3 all working in accelerated GL by the millenium!
I am downloading now, so I don't know what is in the package. If it's the whole of Open GL and is really Open Source licensed it would kill Mesa.
SGI used stochastic dither antialiasing in their renderer, which is patented by Pixar (I left there last week to form a new company and am technically "on vacation" at the moment). That's just one of the areas where "infringement" might be a problem. I hate software patents.
Bruce
Bruce Perens.
Do I see open source, network-transparent, hardware-accelerated 3D graphics in X on the horizon?
There was no year 0 A.D.; the 1st Century A.D. was from the year 1 to 100, inclusive. The 1st Century A.D. had to start with the first year A.D., which we write as 1 A.D.
Someone wrote in to PC magazine a while back about how stupid this was, and that common sense shows that the century starts with 2000. He proceded to give an argument that disproved his own point; he said something to the effect of "Everyone would agree that your 100th penny is part of your first dollar." That is, of course, entirely true, which means that your 2nd dollar starts with your 101st penny. Centuries work in exactly the same way.
This in no way precludes having a great party on 31-Dec-1999. I fully plan to have a huge party on that date, to celebrate the "rolling over of the odometer", as it were, and another huge party a year later, to celebrate the end of the century.
Agreed - although X is relatively well designed, the line between client and server is drawn at some very bad places. It's also not easy to make X support multiple monitors as well as MacOS, short of figuring a way to make the root window use the shaped window extension. The question of spanning heterogeneous displays is also very awkward - what happens when a window crosses from a 32bpp display onto a 16bpp display (or even a 1bpp)? Most window managers don't grok the concept either, so the default behaviour for error dialogs would be every bit as bad as NT. Yes, a replacement for X would be nice, but I have no time to work on such things for the next several years at least, unless someone wants to make that my full-time job. ;^)
Man, I think I'm going to start supporting SGI more. Hmm, maybe my pocketbook might be big enough for a new Visual Workstation, running Linux of course, hmm....
SGI plans to support Linux on it's new intel-architecture workstations -- supporting advanced 3-d on Linux will mean porting GLX. Every commercial Unix has already licensed GLX, so the giveaway won't cost SGI anything. Since SGI has gotten a great Unix for free to bundle with their machines, it only makes sense that they give away GLX. Yes, you'll be able to use GLX on other Linux machines, but SGI believes that some people will be willing to pay for machines with really kick-ass hardware.
In closing, go SGI!
--
This signature left intentionally blank.
Mesa is already mostly there, but it costs to take the compliance tests, so he never claims to be.
-Erik
http://www.sgi.com/software/opensouce/
:)
-Erik
The first millenium had 999 years. Yep, we just cut one off. Just like that. Retroactively even, since of course the calendar changed quite a bit.
I know some pencil-necks are quivering with impotent rage that not every convention of mankind is mathematically precise, but most of us accept it and plan to party this year like the next begins the next millenium. Because according to society's ever-sloppy definition, it is. Try to get over it.
I've finally had it: until slashdot gets article moderation, I am not coming back.
Seems like SGI is taking cues from IBM's turnaround-they are listening to their customers again.
I used to work in a large all SGI shop. To sell the Visual Workstations to management, there would have to be assurances that the existing administrative routines be preserved.
And that takes NT, an OS designed from the ground up to be incompatible with non-MS systems, out of the picture entirely.
Thus, those machines will need an alternative operating system that can play with _any_ existing configuration. I've got one on the tip of my tounge-Linux.
Less than a year ago I had marked SGI for dead by 2001. They are extending the lease on life every day it seems.
I found one thing that some people won't like - the acronym is GPL....;)
Out-by-one errors are so *pervasive*!
Same again, Embarrassed Anonymous Coward.
Perhaps you meant, by 2000. Or Y2K if you must.
NT 3.51 did all the rendering in user space outside of the kernel and it was terriblely slow and linux works like this today. :-(
in kernel graphics is not faster than user-space graphics.
Why does everyone think that graphics *acceleration* has to be in kernel space? "Wheee let's throw everything into the kernel that has the slightest thing to do with hardware" and we'll end up with a monster kernel ala NT4!
(kernel/s/ i know, NT4 is a micro-kernel arch).
So why aren't the SANE scanner backends integrated into the kernel? They send whatever scsi commands they want to a scanner via the scsi generic interface. surely we can't allow that in user-space!!
get a clue people.
unix is good precisely because the all the functions aren't integrated into the kernel. because the kernel implements a simple safe interface to hardware, and leaves the rest to userspace.
ie simple framebuffer that can do some simple functions like change video modes, reset the card,
etc. And leave the accelerated and 3d graphics to X or *whatever*.
If the whatever cocks up, the kernel can always recover.
KISS!!!!
This is good news BUT, in order to get decent 3D accelleration, we HAVE to keep after the hardware companies.
I work for one of these companies (although i cant say which one) and let me tell you they are all in the "considering it" phase, but they really need to know that there is customer interest before they invest man-hours writing X drivers.
This is the most important thing the Linux/FreeBSD/etc. community can do at this point. The more _polite_ and _persuasive_ emails the hardware companies recieve, the better. None of these companies want to be "the first to try it" and then get burned. It must be proven to affect their bottom line.
Wow...sounds like linux may be getting real multimedia development afterall.
heh
"It's Brazilian"
I wouldn't say this is that much of a boost for gaming. The current mesa drivers for 3dfx would do a much better job. GLX is an X server extension and some libraries that translates opengl calls to a data stream, and then converts them back to opengl calls at the X server end.
You would get better performance by cutting X and GLX out of the loop and using a console opengl implementation such as mesa/3dfx. This would be true for any opengl implementation.
Personally, I'm hoping this will help someone embarrass Matrox very, very badly. They promised that they would have a working OpenGL driver in September (about a month after the Millenium G200 was released and the month that the Mystique G200 was released). October, still nothing, but they say it's almost ready. Same thing in late November. Then they release an alpha driver in December which actually works worse than the D3D wrapper in most cases. What had they been doing all that time?! But the worst part is that they were lying to their customers (including me) all the time. I really doubt they realistically believed they could have that thing out in a month in August. Their customer relations were very, very bad. And they treated the people at www.matroxusers.com as their own customer service department, one they didn't have to pay. I hope someone comes out with an OpenGL driver in Linux for the G200 very quickly to further shame Matrox. It's sad; they used to be a good company. I will never buy another card from them, though.
No, _they_ didn't get in working quickly at all. I asked Matrox about X support, and they told me to look at Xi. Suse ended up releasing the driver, not Matrox. And I never said the card itself was bad. It's actually quite decent. It's just their customer service and driver support that sucks. (They also completely botched K6-2 compatibility; no, not just support for 3dNOW; the things make Super7-based machines crash all the time in Windows--even more than they would do otherwise)
GLX is not a replacement for Mesa; GLX is what connects OpenGL (or Mesa) with the X server. In the near future, GLX and Mesa will be hooked up into XFree86, giving a full software renderer.
Precision Insight is also working on enhancements to GLX which will be the
basis for a hardware driver kit that can be used to build drivers for specific boards.
Speaking of G200s - I'm running XFree86 3.3.3. on an 8MB OEM card at 1600x1200, and occasionally get white horizontal flickers while moving big windows. Anyone else have experience with / solutions for this? Basically a good setup, there's just this minor annoyance.