Quantcast

Page 3 of 3 FirstFirst 123
Results 31 to 45 of 45

Thread: Quake II engine being rebuilt and optimized for a Stock Atari Falcon030

  1. #31
    Hero of Algol
    Join Date
    Aug 2010
    Posts
    7,918
    Rep Power
    180

    Default

    Well, as a complete outsider of the Jaguar scene I really don't feel like buying all that he says until seeing it applied on something much more complex and closer to a real game than the early outdated build of his H.E.R.O. remake that I can see on YT.
    I mean, 60 fps on that isn't exactly impressive IMO. But IDK if he has something much more polished now.

    Also, I would like to read more details about the new algorithm for drawing lines that he's talking about and how his 68K-based setup actually works.


    On a system with relatively hardwired 3D hardware such as the PSX I wonder if such a new/faster line drawing algorithm would benefit the PSX or if you're just stuck with whatever the 3D hardware may have built into it.
    I think we would need a ton more info about this supposed faster algorithm and then set a scenario to be able to discuss its possible advantages on other systems like the PS1.

    AFAIK, the PS1 hardware did offer routines for 3D primitives drawing and stuff like that, but you could and should always "complement" the hardware stuff with software. One example is the texture warping. Some game engines put some effort on trying to mitigate it as much as possible.
    If you compare Wipeout to some other pod-like racers on the PS1 you'll notice that most of the time the Wipeout games would present far less texture warping, polygon seaming and clipping issues than the other ones. That's achieved by software AFAIK.
    The same for Quake II vs many other FPS games on the system.

    Going back to your question, it sounds like this new algorithm could be more interesting to systems like the 32X which are have no hardware support for 3D but, then again, that's all very vague assumption since we don't know exactly what he's talking about (what exactly the algorithm does and how it does it; what kind of hardware resources you need to have in order to take advantage of it, which games could make good use of it, etc.).



    Another idea he had that he has started a blog on is procedural texturing. He wants to see what he can do in this regard to get Doom 3 style textures on the Jaguar. Once again these techniques would most likely be more useful on the 32x rather than the Sega Saturn.
    Again, it all sounds very, very vague to me.
    I read his post about the "Procedural Texturing" thing and it sounds interesting but very superficial. A really simple single scene mostly static demo featuring textures generated by his procedural texturing approach barely resembling Doom 3 textures would be the bare minimum to show that his ideas could hold any water in real hardware.

    I'm very skeptical towards big promises combined with "high level programming" and "old hardware".
    History shows that the best results achieved in the homebrew scene with these old consoles are usually by people who don't promise much but have deep knowledge of the hardware and some crazy passion for low-level programming.

    Hopefully I'm completely wrong.

  2. #32
    Nameless One
    Join Date
    May 2014
    Posts
    99
    Rep Power
    11

    Default

    Quote Originally Posted by Barone View Post
    AFAIK, the PS1 hardware did offer routines for 3D primitives drawing and stuff like that, but you could and should always "complement" the hardware stuff with software. One example is the texture warping. Some game engines put some effort on trying to mitigate it as much as possible.
    If you compare Wipeout to some other pod-like racers on the PS1 you'll notice that most of the time the Wipeout games would present far less texture warping, polygon seaming and clipping issues than the other ones. That's achieved by software AFAIK.
    The same for Quake II vs many other FPS games on the system.
    OK thanks that is what I was wondering. I did not know if you had an option to control something like line drawing more directly on the 3D hardware.

    Again, it all sounds very, very vague to me.
    I read his post about the "Procedural Texturing" thing and it sounds interesting but very superficial. A really simple single scene mostly static demo featuring textures generated by his procedural texturing approach barely resembling Doom 3 textures would be the bare minimum to show that his ideas could hold any water in real hardware.
    Procedural texturing has been used with quite a bit of success even with some of DMLs early projects. It's not a new thing. I think the trick is finding the right procedure to generate the textures you want.

    I'm very skeptical towards big promises combined with "high level programming" and "old hardware".
    History shows that the best results achieved in the homebrew scene with these old consoles are usually by people who don't promise much but have deep knowledge of the hardware and some crazy passion for low-level programming.
    He is capable of low level work it's just not his passion. And the line-drawing algorithm I don't believe is system specific so it's not something that you need deep understanding of the Jaguar itself for.

    Perhaps these things weren't researched too well in the games industry due to the crunch of time/money/deadlines. And then with the advent of faster hardware and more memory the need was just not there. This could be a good thing.

    Every little bit helps. Better generic algorithms developed by someone like VladR can be used by hardware specific people like DML whose time would be split researching things like that.

    Many hands make light work.

    Hopefully I'm completely wrong.
    You might not be. We'll see.
    Last edited by A31Chris; 10-05-2015 at 01:54 PM.

  3. #33
    16-bits is all he needs Master of Shinobi matteus's Avatar
    Join Date
    Apr 2006
    Location
    UK
    Age
    38
    Posts
    2,344
    Rep Power
    62

    Default

    I'm puzzled by how people on that forum can be raving about a side scrolling game that shows no 3D camera movement! I could develop something on the MD identical to what is being shown there?!?


  4. #34
    Nameless One
    Join Date
    May 2014
    Posts
    99
    Rep Power
    11

    Default

    Quote Originally Posted by matteus View Post
    I'm puzzled by how people on that forum can be raving about a side scrolling game that shows no 3D camera movement! I could develop something on the MD identical to what is being shown there?!?
    That's an old build. Later versions show more of the side to side 3D rotation. Those builds just don't seem to be up. And we're not 'raving'.

  5. #35
    Hero of Algol
    Join Date
    Aug 2010
    Posts
    7,918
    Rep Power
    180

    Default

    Quote Originally Posted by A31Chris View Post
    OK thanks that is what I was wondering. I did not know if you had an option to control something like line drawing more directly on the 3D hardware.
    You're nowhere near as limited as on the 3DO AFAIK, but Chilly Willy would be the guy to give you a proper picture of the situation.


    Quote Originally Posted by A31Chris View Post
    Procedural texturing has been used with quite a bit of success even with some of DMLs early projects. It's not a new thing.
    I know it's not new, hehehe. Seriously, I know what procedural texturing is.
    But there's a huge gap between "it can be done" and "it's actually useful" and another huge gap to "this is really the best approach to do it".


    Quote Originally Posted by A31Chris View Post
    I think the trick is finding the right procedure to generate the textures you want.
    I think some other questions will popup later on and they're not easy to answer IMO:
    1) Assuming we're using the 68K for the main loop and most of the logic, how do we use the GPU to produce the textures by itself? Or we'll be doing it using the DSP chip which was supposed to be used for audio?
    2) Is the above setup even practical in terms of bus/general resources concurrency?
    3) The tools we have at our disposal allow us to push those crazy setups?
    4) Is any of the known algorithms for procedural texturing even practical (as far as Doom 3-like textures go) within the Jaguar hardware limitations?
    5) Can we get any sort of playable frame rate (let's say 10 fps as the bare minimum) for a 3D game out of all this craziness?
    ...
    Stuff like that.


    Quote Originally Posted by A31Chris View Post
    He is capable of low level work it's just not his passion. And the line-drawing algorithm I don't believe is system specific so it's not something that you need deep understanding of the Jaguar itself for.

    Perhaps these things weren't researched too well in the games industry due to the crunch of time/money/deadlines. And then with the advent of faster hardware and more memory the need was just not there. This could be a good thing.
    I really wouldn't assume that. Stuff like algorithms for drawing primitives have been studied non-stop for decades and decades by academic Computer Graphics researchers and students.
    I'm not saying that by 1994/95 the industry had tried every single algorithm in the book with the Atari Jaguar but it does make me raise an eyebrow when someone claims to have found a better algorithm for line drawing which is supposedly non-hardware specific "just like that".

    It's completely different when a guy like Douglas Little begins to research on something and shares each step of his progress, without making big promises and aiming at better translations/adaptations of a 3D engine to a given platform.

    Anyway, I wish him luck if he's really committed to what he proposed and I'll be happy to see any kind of results coming from it, because I find these experiments really interesting.

  6. #36
    Nameless One
    Join Date
    May 2014
    Posts
    99
    Rep Power
    11

    Default

    Quote Originally Posted by Barone View Post
    I know it's not new, hehehe. Seriously, I know what procedural texturing is.
    But there's a huge gap between "it can be done" and "it's actually useful" and another huge gap to "this is really the best approach to do it".
    I dunno. It's new to me. DML used it and it was useful. He has videos of it. Whether it's the best approach I guess is the nature of these experiments. It definitely takes longer to setup that's for sure and find an algorithm that produces what you want. His texturing on the HERO demo is procedurally done. With the limited memory of the Jaguar it may find a middle ground.

    I think some other questions will popup later on and they're not easy to answer IMO:
    1) Assuming we're using the 68K for the main loop and most of the logic, how do we use the GPU to produce the textures by itself? Or we'll be doing it using the DSP chip which was supposed to be used for audio?
    The DSP chip can be used to run non-audio code as well as audio code back-to-back in the hands of a skilled programmer.

    Quote Originally Posted by Adisak Pochanayon
    fwiw the DSP code I wrote for NBA Jam was amazing. Ran audio in one context and supported full async code in the other. You could literally run two local programs (audio mixing + user code) locally on DSP.
    Not many developers are that talented though. High Voltage Software were and apparently Scatologic. This appears to be the origin of Scott LeGrands claim that the Jaguar has more computation power than the original PSX.

    In a perfect world the GPU will be coordinating with the blitter to draw textures. How much that locks up the GPU I am not sure.

    However VladR is currently not using the GPU or DSP. Just the m68k and the OP/Blitter.

    2) Is the above setup even practical in terms of bus/general resources concurrency?
    3) The tools we have at our disposal allow us to push those crazy setups?
    4) Is any of the known algorithms for procedural texturing even practical (as far as Doom 3-like textures go) within the Jaguar hardware limitations?
    5) Can we get any sort of playable frame rate (let's say 10 fps as the bare minimum) for a 3D game out of all this craziness?
    ...
    Stuff like that.
    I know! It will be interesting to watch.

    I really wouldn't assume that. Stuff like algorithms for drawing primitives have been studied non-stop for decades and decades by academic Computer Graphics researchers and students.
    I'm not saying that by 1994/95 the industry had tried every single algorithm in the book with the Atari Jaguar but it does make me raise an eyebrow when someone claims to have found a better algorithm for line drawing which is supposedly non-hardware specific "just like that".
    I hear ya. I thought about that too. It occurred to me that there must have been a ton of academics even back in the 60s and 70s working on the mainframes and the primitive graphic systems to find a faster way to draw a line so yeah it does raise an eyebrow. However if they found a better line drawing algorithm it doesn't seem to be public knowledge.

    But it is a big claim. He say it's 'multiple times faster' than the regular Bresenham.

    Anyway, I wish him luck if he's really committed to what he proposed and I'll be happy to see any kind of results coming from it, because I find these experiments really interesting.
    Me too. Its fun to watch.

  7. #37
    Hero of Algol
    Join Date
    Aug 2010
    Posts
    7,918
    Rep Power
    180

    Default

    Quote Originally Posted by A31Chris View Post
    I dunno. It's new to me. DML used it and it was useful. He has videos of it.
    Yes, I'm aware of those videos and they're mind-blowing in a lot of ways IMO.

    But DML's approach heavily relies on the Falcon's DSP, which seems to be quite powerful (it's rated around 16 MIPS; Mega Drive's Motorola 68000 is rated around 1 MIPS, for a reference).


    Quote Originally Posted by A31Chris View Post
    Whether it's the best approach I guess is the nature of these experiments.
    Yep.


    Quote Originally Posted by A31Chris View Post
    The DSP chip can be used to run non-audio code as well as audio code back-to-back in the hands of a skilled programmer.
    Not many developers are that talented though. High Voltage Software were and apparently Scatologic.
    This is interesting, I didn't know about NBA JAM Tournament Edition using it.
    I know that the Jaguar version of the game is pretty solid, it has some advantages over the PS1 version. It has some little glitches on the lateral borders of the screen though, it seems to be doing some crazy stuff.



    Quote Originally Posted by A31Chris View Post
    This appears to be the origin of Scott LeGrands claim that the Jaguar has more computation power than the original PSX.
    While it makes some sense on paper (if you consider the MIPS rating of all Jaguar processors) I really doubt it would materialize on the vast majority of games.
    There's a lot of intercommunication/bus concurrency issues which come into play when you try to maximize the resources of a system like the Jaguar and that's why I usually don't buy claims such as that one.


    Quote Originally Posted by A31Chris View Post
    In a perfect world the GPU will be coordinating with the blitter to draw textures. How much that locks up the GPU I am not sure.
    However VladR is currently not using the GPU or DSP. Just the m68k and the OP/Blitter.
    And that's the deal breaker to me.
    DML's achievements rely on DSP-assisted rendering and main loop running on a 68030 (which is quite more powerful than the Jaguar's 68000 and it's also clocked a bit higher)
    IMO the only way to match or surpass on the Jaguar what DML has achieved on Falcon is to use the Jaguar's GPU and/or DSP otherwise it's on a huge disadvantage.


    It's really, really sad how poorly the Jaguar's design was treated by Atari. Had it been properly debugged prior launch and released with proper support of development tools it could have been really great IMO. I find it's design way more interesting and better than the 3DO's and 32X's, it's really a shame what happened.
    Last edited by Barone; 10-11-2015 at 04:50 PM.

  8. #38
    Nameless One
    Join Date
    May 2014
    Posts
    99
    Rep Power
    11

    Default

    Quote Originally Posted by Barone View Post
    It's really, really sad how poorly the Jaguar's design was treated by Atari. Had it been properly debugged prior launch and released with proper support of development tools it could have been really great IMO. I find it's design way more interesting and better than the 3DO's and 32X's, it's really a shame what happened.
    It really is. I have talked to many of the old Atari guys and they seem to have had a very rigid mindset. Third party developers recount stories of things they got the Jaguar to do and when reporting it to Atari their reply was 'That's impossible!' And then it seems the issue would be dropped. It appears Atari wasn't interested in a follow through or to discuss it 'Wow really? Hey let's see what you did!' Impossible seems to have been the keyword Atari loved so much.
    There is an risc gcc Atari had developed for the Jaguars GPU. Unfortunately the resulting asm tripped a hardware bug in the GPU. The Atari guys it appears just threw their hands up at the bug. Well there's no workaround for that so that's that. Even 20 years later when they discussed it on Facebook there's no indication that they even really tried to tackle the problem. Two companies had compilers for the Jaguars gpu that they used successfully. One used Atari's that is mentioned above. And another John Carmack created for Doom. Even 20 years later Atari was unaware of either development. Until recently.

    High Voltage Software got hold of Atari's GCC for GPU compiler(the how of that is an interesting story in and of itself since they weren't suppose to have had it) and blew right through the bug problem. They had an amazing setup on the Jaguar that they had developed using the risc gcc and a workaround for the gpu bug problem. Their development style cut the m68k out of the equation altogether after startup. They developed a gpu manager system that pre-processed the resulting risc gcc asm and then the binary that split up their programs into 256 byte chunks for working with a run-time gpu manager that they developed. The run time gpu manager loaded the chunks into gpu memory and time stamped them. When a new chunk was needed from main ram the gpu manager blew out the least recently used chunk. HVS claims they were getting an 88% cache hit rate efficiency on the Jaguar with their setup.

    More on that story which gets more technical for those interested:

    http://www.3do.cdinteractive.co.uk/viewtopic.php?t=3496
    http://www.3do.cdinteractive.co.uk/v...hp?f=35&t=3492

    More on the risc gcc:

    http://www.3do.cdinteractive.co.uk/v...&t=3356#p37117

    This is the last thing they did using that setup. It's unfinished but it shows the performance levels they achieved with it.

    https://www.youtube.com/watch?v=XG904E964Ts

    A 3D guy who has been working in the field since SGI workstations and before looked at that video and was impressed with the performance HVS achieved with it considering the hardware. So much so his 3D work on the Jaguar is targeting a resolution of 320x240. Though that level may not be possible without HVS's custom setup.

    High Voltage Software was going to sell their setup to Atari but Atari folded before they could. And until about two years ago Atari had no knowledge that anyone had used their risc gcc successfully. I have links to the Facebook conversation if anyone is interested. It is an interesting read.
    Last edited by A31Chris; 12-17-2015 at 11:11 PM.

  9. #39
    Hero of Algol
    Join Date
    Aug 2010
    Posts
    7,918
    Rep Power
    180

    Default

    I'm digging you last post and your PM (thanks a lot for that).

    And this:

  10. #40
    16-bits is all he needs Master of Shinobi matteus's Avatar
    Join Date
    Apr 2006
    Location
    UK
    Age
    38
    Posts
    2,344
    Rep Power
    62

    Default

    Unfortunately I can't spread anymore rep!


  11. #41
    Nameless One
    Join Date
    May 2014
    Posts
    99
    Rep Power
    11

    Default

    WOW now THAT"S a demo!

  12. #42
    Hero of Algol
    Join Date
    Aug 2010
    Posts
    7,918
    Rep Power
    180

    Default

    This demo needs to be posted here:


  13. #43
    Raging in the Streets goldenband's Avatar
    Join Date
    Dec 2009
    Posts
    4,516
    Rep Power
    96

    Default

    Whoa.

  14. #44
    Nameless One
    Join Date
    May 2014
    Posts
    99
    Rep Power
    11

    Default

    Quote Originally Posted by saturndual32 View Post
    Amazing. I just wish he had picked the Jaguar instead of the Falcon, as the lucky Atari machine he is pulling miracles on, hehe.
    A quick heads up. Word is DML has reacquired an Alpine developers board a few months ago for his Jaguar. This is most likely more related to tinkering with his old work on live wire than anything Quake related.

  15. #45
    Road Rasher
    Join Date
    Jun 2011
    Posts
    484
    Rep Power
    16

    Default

    Quote Originally Posted by A31Chris View Post
    A quick heads up. Word is DML has reacquired an Alpine developers board a few months ago for his Jaguar. This is most likely more related to tinkering with his old work on live wire than anything Quake related.
    Thats great news!
    And about it not being Quake related...anything can happen!
    Thanks for the heads up.

Thread Information

Users Browsing this Thread

There are currently 1 users browsing this thread. (0 members and 1 guests)

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •