Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 


Forgot your password?
Close
typodupeerror
×
Quake First Person Shooters (Games)

Brian Paul to join Precision Insight 81

Posted by Hemos from the fun-with-graphics dept.
physic writes "Brian Paul, the maintainer and original author of the free OpenGL library called Mesa. will be joining Precision Insight to work on Linux/Mesa fulltime. Mesa is best know to the linux gaming community as the library that allows Quake3 to run under linux on 3dfx, nVidia TNT2, and Matrox G200/400 video cards. "
This discussion has been archived. No new comments can be posted.

Brian Paul to join Precision Insight More

Brian Paul to join Precision Insight

Comments Filter:
  • So how long until a Tux Q3 skin comes out?


  • Ehm (Score:1)

    by Adnans ( 2862 )
    Hasn't he been working for PI for a couple of weeks now?
  • Its almost a month since 1.08 q3test has been released, and they still don't have a linux port. WTF?!?!?!??! I'm going threw quake3 withdrawl.

  • I used to use a penguin skin for CTF, but it stood out really well and I kept getting fragged.

    Or maybe I'm just lousy. Hmm.

    --
    QDMerge [rmci.net] 0.21!
  • It has been out for quite a while. The beginning of August as I recall.

    - |Daryll

  • Is this possible? I'd like to play quake1/2/3 in linux at a decent speed. IE: not the current GLX.. and don't really want to wait for the XF4.0 DRI stuff.. is anyone coding mesa for TNT1/2?
  • Hrm, last I checked Q3 wasn't exactly playable with Linux+TNT. What's going on with TNT/GL development these days? Will it be in XFree 4.0?
  • They're waiting for the DRI. Nobody going through a wire protocol encode/decode is going to be very fast until it hits (my Voodoo 3k chunks on my dual P3/450 occasionally :( ).
  • No. We spoke to him over SIGGRAPH at the start of August. He took some time to think about it and visit our office in Steamboat Springs.

    He did actually accept the offer a little over a week ago, but his start date isn't until October.

    - |Daryll

  • The DRI is out in the 3.9 series of XFree releases, so they can start working on it.

    The Voodoo cards do not go over the wire. They are essentially direct rendering already. (Albeit not integrated with the window system).

    We demo'd Q3ATest at SIGGRAPH running in a window as our first full OpenGL implementation within the DRI.

    - |Daryll

  • I talked to Frank and he said that q3a was peaking at around 100+ fps with steady rendering rates in the high 40s. Sounds nice ;).
  • If you're such a fan of linux and quake you would have known that the linnux version came out along time ago...
  • ... you could be bringing Brian one step closer to a lawsuit.

    With the way SGI is embracing Linux, I doubt it.
  • Um, so what's a brother gotta do to get q3test running with a Matrox Millenium G200 AGP?

    I did download Mesa at one point, and glx, but then I gave up when it wanted me to replace some libraries and recompile X. It was, and still is, a bit over my head, and struck me as the type of thing I would screw up, in the process permanently hosing my X Window System.

    Is there an easier way?
  • 1.08 for Linux has been out for almost a month. Don't trust www.quake3arena.com, trust the archives.
    ftp://ftp.cdrom.com/pub/idgames/idstuff/quake3/lin ux/q3test-1.08-glibc-2.i386.tar.gz is dated Aug 11, 1999.

    --
  • There has been lots of work for the Matrox G[24]00 cards recently. Matrox has released nearly all (if not totally all) of the specs for the cards (hundreds of pages worth) which makes programming for them even easier than the TNT cards (which release source but not the best in the way of specs). With some of the most recent code for X / GLX / AGP / Gx00 WARP and stuff, we are reaching pairity with Windows -- according to John Carmack, we have even better performance than Windows on Quake2demo2, and have about 90% of the Windows Quake3 demo speed. And we're improving all the time. So go for those G400 boards... they're darn fast and getting faster!

  • Cool (Score:1)

    by florin ( 2243 )
    There's nothing better than doing exactly what you like to do and then getting recognition and finally, paid. Congratulations Brian for his excellent work on Mesa and for turning it into his own job.
    1. Get Mesa 3.0 (cvs wan't working last I tried), compile, install.
    2. Get latest cvs of glx. compile install.
    3. replace link from libMesaGL.so.3 to libMesaGL.3.0 with link from libMesaGL.so.3 to libGL.1.0.
    4. Make sure your X config has both glx.so in the modules list and a 640x480 setup (the latter got me for a while).
    5. Run q3test.
    6. If you're anything like me, get slaughtered:)

    BTW: FAQ [openprojects.net]

  • With saturday's (NZ time) cvs, glx gets tolerable performance on my dual 300 celery, generally playable but kinda laggy in spots, until my mouse drops a byte and X and my mouse no longer agree on the beginning of the packet. When I crank the cpus up to 450, things are noticably better, including the mouse, I think. However, I don't leave the cpus overclocked as I don't have adequate cooling to run with the case closed and I don't like leaving it open (too many little fingers in the house:).
  • Which is a better way to land a job with your hobby: Writing a development library or writing an app. Together with the guy who wrote Wine, we've seen two people this year land jobs coding their development tools for income but when we look at application programmers the closest was the XMMS programmers getting free soundcards from 4Front but no jobs. Is Linux supposed to be a development platform or what?
  • I dunno. I play Q3test on a TNT2 all the time, I've never had a speed complaint. Ok, well, say I'm in a room with 12 or 15 others and we're all shooting rockets and stuff, it does slow down a bit. But is that the card, or the server?

    PS-- my machine's a dual P3 450 running RH 6.0. Wonder how much that has to do with it...

  • I know Brian personally - we went to college together (name dropping at /. - heh!). He is a very talented individual. The smartest individual I know in regards to computer graphics. I wish him the best personally and professionally.

    Now, I wonder what he did with all that code for the 3D object designing tool he made that was developed under SunOS (a 68020 machine!) for the Pixar that UW-Oshkosh had back in '89... =)

    Yes, kiddies - Pixar use to make hardware! Boy do I feel old...
    --
    "All that is visible must grow and extend itself into the realm of the invisible."
  • > Is there an easier way?

    On the developers' mailing list they had just decided to clean things up and make their first release, when they got a one-two-three punch from the release of the XFree86 pre-4.0 kit, specs on the WARP engine, and about a week's outage on their CVS server. The server is back up again, but they seem to be trying to take advantage of some of the new stuff before they can up a release kit.

    Just a guess, but I suspect they'll have a kit out sometime during the autumn.

  • > this is not an implementation of OpenGL

    For whatever it's worth, recent CVS versions of Mesa have been building a libGL.so rather than libMesaGL.so as was done previously. I won't pretend to know the story behind the change.

    > That would require a hefty license

    Does it really? I thought it would have to go through some kind of validation process, which likely as not isn't free, but I've never heard mention of a "hefty license".

    I wouldn't be surprised if they and SGI were working together to get a "Genuine OpenGL" certification for Mesa, in order to get a standardized cross-platform 3D toolkit into the widest possible play before some other company (ahem!) takes over the field.

  • ...amen to that.

    If you've got skill, you can compensate for the lag (I used to be 28.8)

    As long as you're connection isn't bursty its still the king. If it is bursty, well, the others suck pretty bad too...

  • Mesa has also been important to Macs (Score:3)

    by webslacker ( 15723 ) on Monday August 30, 1999 @04:14PM (#1715931)
    For a while, Mac users have been able to use PC Voodoo2 cards with their Macs thanks to Mesa (except for Diamond's MonsterII cards AFAIK). Being a Mac owner and wishing for better 3D than my Rage128, I can certainly appreciate the work these guys did.

  • This isn't just Quake (Score:3)

    by Caballero ( 11938 ) <daryll@noSPAM.daryll.net> on Monday August 30, 1999 @04:29PM (#1715933) Homepage
    I also wanted to remind people that this isn't just about Quake. The work we're doing on OpenGL is about 3D graphics for Linux. Digital content creation, CAD/CAM/CAE, scientific visualization, medical imaging, and yes games.

    We showed Quake at SIGGRAPH because it was an easy thing to leave running as a demo, not because games are the only, or even the most important OpenGL application.

    - |Daryll

  • You think Quake was the game to complain about lag with? haha we all know that the BEST FPS was of that year was Duke Nukem 3d. I could rpg people with 5 second lag jumps...wow i miss that game. Anyone care to comment on the lack of DN4ever?

    -brandon
  • Mesa is an implementation of the OpenGL API. It is easy to slip and call it OpenGL, but that's really not correct. No, SGI isn't going to sue Brian, but people should respect the distinction. Brian is careful about it.

    To officially be called OpenGL the product must pass a set of conformance tests and you must pay a licensing fee to SGI. Then you may call it OpenGL. That's why Mesa is called Mesa and not OpenGL.

    In the future, I suspect Mesa will pass all the conformance tests, but it still won't be called OpenGL because no one will pay for the license.

    - |Daryll

  • This is slightly off-topic. But 1st let me thank all the ppl that have worked on the linux port of X11 and OpenGL. Thank you, Thank you, Thank you, but I want more.:-)

    I work on various unix workstations (including SGI, HP, Sun, and linux) and find that the commercial versions of unix have greater support for more visual classes than linux. I realize that linux has the "disadvantage" of trying to support a very wide range of graphic cards. I also realize that the lack of support (e.g., documentation) from various graphic card manufacturers makes things very difficult.

    My X Window and OpenGL code sucks. I'm not a professional programmer but some low-life scientist that writes a lot of inefficient code; you know, cheese-ware full of holes. Giving me a wide range of available visual classes really helps. Pseudo-color overlay planes is also nice (coming soon in 4.0!).

    My point is that I would really support anybody that makes my life easier. Writing code that has to take into account the availability of all the various visual classes and default depths is a real pain in the butt. I have to think that other ppl with legacy code would also want to have this. The economics of the problem also indicates that the cost of relatively cheap graphic cards easily offsets the cost of rewriting the programs.

    I ask you to have some compassion for me as I have to use Motif.:-) OTOH, I'm not so proud to ask how I can better solve my problem. Nonetheless, congrats Brian; I'll start watching Precision Insight more.

  • hmm, thinking about it, the BSD Daemon would make a great quake skin, for the axe he could use his pitchfork and such... *evil grin* maybe if i get bored enough one of these days...


  • I would just like to take this opportunity to thank the guys at Precision Insight, SGI, and XFree for the great work they have been doing on the direct rendering infrastructure. I think this is one of the most exciting changes to X for linux, and after hearing the talks on the subject at linux expo, and talking directly to a few of the guys, I honestly believe they are doing a great job.

    Daryll makes a good point above that people should realize this impacts not only game playing, but could create a larger general awareness of linux from larger ISVs, and the users of their products.

    The latest news that Brian has joined PI has me thrilled. This was the exact right move. Congratulations to all involved in the project.

    Eric
  • mesas biggest drawback is the lack of trim nurbs.
    thats what will keep maya etc. off XF86-4

  • Everyone can say anything they want about 3dfx, about glide, and how much the TNT2 is a better card.

    All that matters is support, and right now if you run linux, your only REAL choice is a 3dfx. Gives you the nice /dev/3dfx, works flawlessly with q3, and everything else.

    I'm the first to admit the TNT2 is a better card, but if the card is only "soft of" supported in mesa, X, etc, it's a moot point..Sit around and wait until it is? or get a 3dfx card, and enjoy what there is now?.

    Linux world isn't like the Windows world, where you can simply compare hardware and get the best. You have to factor in the linux support as well. And what usually ends up happening is you end up with the 2nd or 3rd best, but with the 1st in linux support.
  • PI is working with the Xfree86 [xfree86.org] team on a hardware accelerated GLX module for Xfree 4.0. The support wasn't in the pre-release [xfree86.org] 3.9.15 but they said [xfree86.org] that it would be in an upcoming pre-release. The stuff that suse [www.suse.de] is working on is called MLX [www.suse.de] and it hasn't been updated in months but it still works(sorta, full screen only) on my Diamond FireGL 1000 pro.
  • With SGI putting so much faith in Linux, it would be idiotic for them not to certify Mesa as open GL compliant. that is unless they want to re-implement openGL from the start with Linux. It wouldent suprise me it once Mesa is up to specs, and has desent hardware acceleration, that SGI would certify it for no fee. SGI is still a leader in graphics, and for them to deny the certification of a critical 3D API for the OS that they are betting on would be relatively stupid...
  • I've been running Q3A reasonably happily on my TNT 2 Ultra for a couple of weeks now. The machine is a Celeron 450a.

    The GLX renderer isn't amazing -- I have to turn off lightmaps, and I run in 640x480 (haven't really messed w/higher res modes, the card may have plenty of fill to do them...). I'm still waiting for DRI to get full performance out of the board.

    Nonetheless, q3 ought to run acceptably on even a TNT with a few more of the settings tuned down...
  • Congrats, Brian!

    Does this mean Mesa is going to get more better faster? =)

    --
  • True Linux support comes in the forms of vendors releasing specs for their chipsets so that open source drivers can be produced. This is why I bought a Matrox G200 (couldn't affort a G400) when I got my new computer: Matrox released the spec. Admittedly not all, but as things have been progressing, they've released more and more of the specs.

    According to John Carmack in a posting on the glx-dev mailing list, the Linux G200 driver is almost as fast as the Windows driver (this is with his special, experimental, tree; soon to be merged (WID)) and he expects to to exceed the speed of the Windows driver soon. I thus do not regret my purchase. The driver is a little buggy (I get some strange effects when I die in certain modes in q3test), but very usable and fast enough for my dual celery 300a (without SMP support in Mesa) and even better when I crank up to 450.

    You can keep your binary driver while I bask in the glory and warm fuzzies of my rapidly developing OS driver.

  • Amen and Hallelujah Brother. I don't think people really understand what a difference integrating 3D into all the distros will make. All you folks who make 3D on my Linux desktop a snappy, pleasant experience where I can code my own visualization for free are hero's in my book.
  • Hehe, weeelll, wouldn't you agree that digital content creation, CAD/CAM/CAE, scientific visualization and medical imaging are all relevant to games? CAD is good for designing those models and maps; scientific visualisation helps you get things, especially SFX, to look right; medical imaging is to get body models anatomicly correct, especially post rocket):>; and digital content creation is, umm, the game itself(?).

    Oh, I agree, video games drive the market (and don't tell me that those medical imaging systems don't get played with, who could resist playing with such a `toy'?), but the rest is very usefult for creating the games. They're also useful for affording to play the games.

  • I remembering hearing at some point that quake 3 was multithreaded? Is it now, or will it be?(supposedly?) thanks
  • If I understand correctly, the former x11amp folks are working for 4front now. And yes, Linux is a superior development platform. Heck, even a superior gaming platform once the hardware companies start shipping those DRI modules for XF86 4.x.
  • Agreed...

    As a graduate student in graphics at UC Berkeley, and I would love to be able to recommend installing some cheap Linux boxes to do 3D work in our lab. But until the graphics support reaches the same level as SGI or Windows (i.e. accelerated 3D in a window) I don't feel that Linux is useful enough for us.

  • People have been spending billions of sweat and toil earned dollars on video games. Just try to convice me that people do the same with 'digital content creation'...

    Dude, you have no clue how many billions of dollars are involved with the digital content creation, simulations and such. Don't even get me started on the space simulations part or the chemistry stuff.
  • I'm not dissing games. I play Q3A and you might see me on a server every once in a while. I do think games are good for the spread of Linux.

    Under Linux today, I suspect games are not the largest market. That's because Linux isn't getting used on the home desktop as much. I know there are a lot companies that want 3D under Linux for other applications.

    The reason I brought it up at all was that the posting was labelled as being about Quake and the threads were all about Quake. Digital content creation means the games can be built under Linux. It also means film production can be done under Linux. I worked on a film that approached $2B in worldwide gross. We did use Linux on it, but only as a render farm. I want to see that change. I know from first hand experience that it's a sizable market that is desperate to use Linux. I know people doing SciVis/Medical work with my 3dfx drivers and want better support. All those are important as well.

    The bottom line is that it all needs good OpenGL. I want people to keep in mind the other capabilities that it will enable. Don't focus entirely on games.

    - |Daryll

  • Yes. And a very good one at that. (IMHO, of course)
    --
  • if i recall correctly, 4Front sponsers the xmms team, and they pay them to port xmms to other systems, I think. more info is on the xmms or 4front webpages.
  • ..I got him mixed up with you Daryll ;-) Thanks for the kick-ass work
  • > With the way SGI is embracing Linux, I doubt it.

    With the way SGI is embracing MESA, I doubt it.

    Over a year ago, they were saying that they were doing everything to get MESA be openGL compliant. The only thing they were refusing him was official OpenGL approval. That last issue might be eroding nowadays.

    Roger.
  • Q3Test is playable with a TNT (here ELSA Erazor II) under X11. I've installed the X Server with the TNT from nvidia.com.

    Actually, It runs better than Q2 here.

  • OpenGL Conformance and Mesa (Score:4)

    by Oddhack ( 18073 ) on Monday August 30, 1999 @11:21PM (#1715965) Homepage
    There's a common misconception being expressed in this thread. The conformance tests cannot be used to test an API or a collection of source code, because conformance is a property of the actual driver binary and hardware being used. It is highly unlikely that (for example) software Mesa, the Matrox driver based on Mesa, and the Nvidia driver based on Mesa will all have the same conformance test results. Thus it is literally impossible to say that "Mesa is conformant", only that a particular driver based on Mesa is conformant.

    That said, we did give Brian Paul access to the conformance tests for his own personal use, as an aid to improving Mesa.

    There will be a great deal of OpenGL activity on Linux in the next few months, from SGI as well as others. Stay tuned. BTW, if anyone is thinking about going to the Open Source / Open Science [bnl.gov] conference at Brookhaven National Lab in October, I'll be speaking on OpenGL and Linux there (mostly a status update aimed at researchers, though).

    Jon Leech
    OpenGL Core Engineering
    SGI

  • Brian,

    Congratulations on the new job - I know Avid Technology will miss you. Thanks for putting so much effort into Mesa.

    Both you and all at Precision Insight and SGI deserve a cheer for chasing high-end 3D.

    Now, get cracking! ;-)
  • Lack of trimmed nurbs in Mesa won't keep maya or other professional packages from Linux.

    When (not if, it's gotta be when) AW releases Maya for linux they will almost certainly require you to have one of the licensed OpenGL implementations from Xi Graphics or Metrolink. I'd bet SGI will also release their own implementation of OpenGL on linux (probably only supporting their own hardware).

    After all if you're paying $5000US for a maya license $200 extra for a OpenGL implementation isn't really significant.
  • Couple of questions:

    How much would it cost to get a particular driver conformant?

    Shouldn't the OEMs be working on getting their graphics cards comformant? What can we do to "nudge" them along? ;-)
  • Pixar was the name of the computer Lucasfilm built (or had built?) to render, among other things, the Genesis sequence in Wrath of Khan, and the stained-glass knight in Young Sherlock Homes. They did a presentation about the computer at the "West Coast Computer Faire" in S.F. back in 84 or 85 (I'm not sure exactly).

    Man was that awesome! Upwards of 1000x1000 resolution with a billion colors, all packed into a case barely bigger than a family refrigerator :-)

    chris
  • > but it still won't be called OpenGL because no one will pay for the license.

    Any idea what the scale of the license is? Is it a per-copy license, or a one-time fee?

  • The problem is not the large number of graphics cards that need to be supported, but simply that those graphics were first and foremost designed with Windows in mind. The concept of multimode visuals doesn't exist on Windows, so why complicate the hardware ?
  • We're trying to make sure everything we do is as independent of OS as possible. There are some other folks that are trying to port it to BSD for example, and they made some good suggestions for things we can do to make their lives easier.

    With that said, the customer demand right now is for Linux. The companies that are paying us to do the work are paying us to do Linux, and the vendors that want to write applications want to put them under Linux. So, the focus is still there.

    - |Daryll

  • It is most cool that Brian can now work on Mesa as part of his job. This is a Good Thing for the community as well as him. I went to college with Brian, and he was cranking out kick-ass graphics code even way back then (over a decade ago)... he will do amazing things at PI, I am certain.

    Thad

  • Licensing costs vary quite a bit depending on the exact circumstances, so that can't be easily answered.

    The IHVs do work on conformance for their Windows drivers; they can't pass WHQL certification without it. Similarly for the workstation vendors on IRIX, HP/UX, Solaris, etc. On Linux or other OSes this hasn't become an issue yet; I don't know of any IHVs who are supporting their own OpenGL drivers for these platforms. This will change (is already changing at some companies), and at that point they'll put the effort in to pass conformance as part of their driver work.

  • (Sorry, I couldn't resist the urge to name-drop also.) The editor Brian Paul made for modelling Pixar's old 'Go' files got passed on to me as a Practicum project a bit later on (I went to the same school, but Brian graduated before I got there. I think I saw him in person only once, when he came back to the tiny UW-Oshkosh to give a talk to the CS club.) Anyway, the problem was that 'Go' was no longer being used much and everything had shifted to 'Renderman'. I was charged with the task of altering the editor so that it read/wrote Renderman .rib files instead of Go files. This was supposed to be a fairly dead-simple project but it turned out to be harder than expected because the innards of the editor were tied to the 'go' format rather tightly. For example, 'go' was left-handed like all older graphics systems were, but Renderman is right-handed. Simply finding all the places were this flipped the math was an interesting task. (I often made the mistake of flipping the math in two places along the pipeline such that I flipped it back to lefthanded again.)

    The second major obsticle was that 'go' used camera and lookat point notation while Renderman simply stored the transformation matrix and nothing else. This meant that it was possible to recreate the camera point and lookat *direction*, but not the exact lookat point. This was a problem because the editor used the lookat point as the point to swing around when rotating the camera, and thus if you loose this information, the scene swings wildly when you rotate the camera and looks more like a 'pan' than a rotate. The solution was to assume that the scene was 'near' the world's origin point, and thus when recreating the lookat point, you pick the point along the lookat vector that is closest to the origin. (Simple math - find the point along the lookat vector where you get zero for the dot product of the lookat vector and a vector from the origin to the lookat point. A dot product of zero indicates perpendicular vectors, so that must be the closest point to the origin.)

    Anyway, the project passed on to someone else when I left (I can't remember his name, but I remember the face), and it was mostly working then and he was going to add new features to it to take advantage of Renderman features that were not in Go.

    I hope that long-winded answer helps satisfy your curiosity. I didn't meet him in person up close, but from fiddling with his source code I'd agree with your notion that he's quite a 3D guru. Some of that math was pretty neat and there was some things where at first glance it seems the algorithms shouldn't have worked but they did anyway. (Further examination showed that Brian was just being more clever than I could handle.) That guy is pretty darn cool. I wish him the best of luck in his new job.

  • I can not wait when visualization of 3d molecular structures will be fast with Linux. Thanks to all of you for the work!
  • A tux skin for Q3A would be pretty cute. I can see a Linux clan vs. a BSD clan battling it out. Cute mascots battle it out on Q3A over who's cuter. :) Well, IMHO Tux is cuter. In fact, Tux wins over Pokemon anyday.
  • SGI is backing the XF86-4.0/mesa thing. im pretty
    sure they want it to run with that. its well known
    that only a small portion of linux users use
    xig or matrox. why would AW want to alienate them?

    also, since games are already using mesa with acceleration, the conformance precedent is set.

    maya is about $7k for complete and (i think)
    $16k for unlimited.

    the commercial implementation is significant.
    the source of mesa has saved me alot
    of time (and thus, my client alot of money)
    i would certainly not want to be forced into
    a non free (in speech sence) implementation.

Slashdot Top Deals

Money will say more in one moment than the most eloquent lover can in years.

Close