Klax, Atari ST to Jaguar conversion

[Shamus]

 

Try beating the klax now. I hear it's becoming resistant to antibiotics!

I knew a guy who had the klax once. It wasn't a pretty sight.

[sh3-rg]

I think the problem is we've never had what could be considered a truly AAA game from a developer who knows the real ins and outs of the Jaguar to see what it's real potential was. Instead, what's filled that void has been endless speculation, debate, and what ifs. If there truly was something that we could point to that indicated, "yes, this is pretty much what the console's potential is and not too much more," I think a lot of that speculation would dry up. For the countless systems that did have their potential clearly realized, there's never any such talk. I'm just not sure if the Jaguar will ever realize it, but who's to say what will happen in 5 - 10 years with an unexpected breakthrough here and there? Certainly it's taken at least as long for some other platforms to have homebrew creations that meet or exceed the best commercial releases than the Jaguar's close to 20 years since its commercial demise. In the mean-time, it's clear that the often baseless speculation will continue.

 

I'm not sure that's really fair to say there was no AAA from people who really understood the architecture. Well, whatever passed for AAA in the early 90s, and from people who knew more about Jaguar then than maybe only 2 or 3 people know today or will ever know going forward.

 

Super BurnOut. Have you read the story of how this game came about? All the hoops the guy jumped through to produce the end product. Sprite-based game pushed to the limits of the Jaguar's capability thanks to a God-tier understanding of the object processor and potential pitfalls of the Jaguar. They might have been working night shifts with borrowed hardware, but the result stands up there with anything around that time.

 

Rayman. The guys who wrote this weren't sat around resting on their laurels. I'm sure you're heard the phrase "GPU in main". Despite what we've witnessed over the past 10 or 12 years, these guys were making use of it back in the day, as were the Virtual Xperience guys (the ones the Super BurnOut devs time-shared their hardware with). It's not like devs back then didn't know how to get more out of the Jaguar than what was written in the handbook.

 

Tempest 2000. Minter's blitter "trick" - another example of someone going above and beyond what the Atari Book Of How To Make Jaguar Games.

 

People back then knew how to tinker with hardware and in the Jaguar they had a machine that was all about tinkering.It's probably unfair to think that all devs back in the day simply ported stuff over or settled for a bare minimum and that their understanding was lacking. You had some of the best Atari ST and Amiga guys working on it for periods, people who really knew how to write code to the metal and fuck with hardware in ways nobody ever envisioned. For me that's enough, I don't have any problem saying that what we got in little pockets probably wasn't far from reaching a mythical "full potential". The Jaguar doesn't secret away all this power behind layers of abstraction, it's says there, says "haaaaaaayyy!" and looks at you seductively. There's two ways people take that when presented with it - they either shit their pants and scream for something higher level that will hide away all the nuts and bolts or they relish the opportunity to get their fingers in all those warm, moist crevices do do their worst. Like I said, there are a few tricks and tips for getting better performance and I gave examples above, but I don't think there's a hidden box of many more just waiting to be discovered... there's not a massive scope for digging and bypassing to get to somewhere secret and special... Jaguar just stands there with it's trousers down and you either want to entangle yourself in its junk or you don't

 

When you simply look at 3d polygon engines, you should read no further than kskunk's excellent post where he describes how he and others have sat down, done their due diligence and aimed to produce an engine in the most Jaguar-centric way, having a deep understanding of the hardware. He goes on to explains why the esoteric architecture has a knack for pegging your plans back time after time. If that guy shrugged and said enough is enough, I don't need to ponder any further.

[LinkoVitch]

Who here has beaten KLAX? I beat the Lynx version many moons ago...

 

I never thought it would be the sort of game that had an ending, I always thought it would just be one of those that just got faster and faster. Is there a special end of game screen or anything? or just an unsatisfying "Well done" message?

 

Jaguar just stands there with it's trousers down and you either want to entangle yourself in its junk or you don't

 

Crap, does this mean I should start taking the DSP out for a fancy meal more often? I've been all up in it's junk for ages! :D

[JagChris]

I disagree with a lot of this. Oliver never said he pegged the hardware in Super Burnout. In his second game that never was released he began leaving the 68k behind in his programming.

 

IWhen you simply look at 3d polygon engines, you should read no further than kskunk's excellent post where he describes how he and others have sat down, done their due diligence and aimed to produce an engine in the most Jaguar-centric way, having a deep understanding of the hardware. He goes on to explains why the esoteric architecture has a knack for pegging your plans back time after time. If that guy shrugged and said enough is enough, I don't need to ponder any further.

 

Kskunk is an awesome guy, but there is something to be said for finishing something even if the journey does not meet your expectations.

 

There are many who ran into the same things Kskunk ran into and finished their engines and produced a product we all see. Cyril Cogordan.(Robinsons Requiem) Lee Briggs(World Tour Racing) These two were all alone and in the dark. The team who did Hover Strike and the sequel. Francois Yves Bertrand with Fight for Life. Eclipse(Iron Solder 1 and 2) And of course Scatologic. And all of them who have actually finished their 3D engines have said that more can be gotten out of the machine than what we've seen so far. Except for FYB. I don't remember him saying anything one way or the other on the subject.

 

With all due respect to Kskunk IMO he should of stuck it through and finished his engine. Even if things turned out different than expected. Perhaps his thinking needed to change a little rather than just walking away when things aren't going exactly as he planned.

 

Overall I agree with Bill. It would of been fun to see more of what the Jaguar could of really done in all areas because I'm convinced we have not seen all of it.

Edited February 27, 2014 by JagChris

[JagChris]

Rayman. The guys who wrote this weren't sat around resting on their laurels. I'm sure you're heard the phrase "GPU in main". Despite what we've witnessed over the past 10 or 12 years, these guys were making use of it back in the day, as were the Virtual Xperience guys (the ones the Super BurnOut devs time-shared their hardware with). It's not like devs back then didn't know how to get more out of the Jaguar than what was written in the handbook.

 

What were the Virtual Xperience guys using it on? Val D'isere? It was used on Rayman? Where did they say they used it?

That's 2D. My understanding is the GPU in main is only a real boost when using it for stuff like true 3D polygons.

Edited February 27, 2014 by JagChris

[Bill Loguidice]

I'm not sure that's really fair to say there was no AAA from people who really understood the architecture. Well, whatever passed for AAA in the early 90s, and from people who knew more about Jaguar then than maybe only 2 or 3 people know today or will ever know going forward.

Great analysis. I guess my greater point was not so much the lack of AAA titles - I think that's clear. I think it's also clear (and you provided great examples) that we can point to specific games that make good use of some of the Jaguar's features. I suppose my REAL point was being able to point to a game that uses the vast majority of the Jaguar's features and really pushes the platform to say (at least) 90% of its absolute potential. Your example of Super Burnout was good - it's a technically sound game - but does that really push the Jaguar to its limits? I somehow don't think so. You pointed to Rayman -- even on less 2D friendly systems it was replicated,and, while it looks nice (really nice), I can't point to any one thing about it that pushes the Jaguar to my non-technical eyes. Tempest 2000 is obviously a classic and just about perfect on the platform, but again, it's deceptively straightforward audio-visually and again was easily replicated on other platforms.

 

One of the games that comes to mind for me is Zero 5. Though it's basically background-less, it seems like a technical showcase. Is that pushing the Jag to its limits? Battlesphere Gold? I don't know. Is there something you can then point to in your opinion that meets that criteria, that this is more or less the ceiling of the Jaguar?

[LinkoVitch]

What were the Virtual Xperience guys using it on? Val D'isere? It was used on Rayman? Where did they say they used it?That's 2D. My understanding is the GPU in main is only a real boost when using it for stuff like true 3D polygons.

Main ram is slow, very slow compared to the SRAM on the riscs, even just that fact makes gpu in main always slower than running in local! It never will offer a speed increase! It is only likely to offer a boost to convenience given larger code pace, or if perhaps you are using gpu ram for something else! This is ignoring the fact that you are now also using the sole bus of the system to run code over, whilst you do that nothing else is using it. If a higher priority CPU comes along then it will likely have to change memory pages too, more delay!

 

Better solution would be to only run code in local, and blit in any extra chunks when they are needed. You can get an aweful lot of code and stuff into 4K anyway!

[Gaztee]

I never thought it would be the sort of game that had an ending, I always thought it would just be one of those that just got faster and faster. Is there a special end of game screen or anything? or just an unsatisfying "Well done" message?

 

 

Crap, does this mean I should start taking the DSP out for a fancy meal more often? I've been all up in it's junk for ages! :D

Well that explains why you have such smooth skin on your face these days

[sh3-rg]

I disagree with a lot of this. Oliver never said he pegged the hardware in Super Burnout. In his second game that never was released he began leaving the 68k behind in his programming.

These are his exact words:

 

“I split the whole screen into 8 zones...every sprite was split in pieces to fit each zones. That worked really well, and we never saw any video glitch with it, even when we had thousands of sprites / tiles moving / scaling at the same time. At that point, the whole game was developed in GPU.”

 

“A lot of games were only using the 68K and a little of the "GPU" and used the DSP for sound” according to Nallet; lending credence to the notion that it was developer incompetence and not hardware concessions that made development for the Jaguar difficult. “I knew I had to push the GPU and DSP for Super BurnOut to stand out.”

 

Kskunk is an awesome guy, but there is something to be said for finishing something even if the journey does not meet your expectations.

I'm not sure what that means, are you saying kskunk didn't release anything so his findings are irrelevant? That's one opinion I suppose. I prefer to look to this:

 

Years later, I sat down with the docs, assembler, and logic analyzer. I knew untapped power lurked inside,waiting all these years for someone, finally, to unlock it.

 

After trying, I get it now. The Jaguar is a kit of ingredients by somebody who had never made a cake. The docs describe how great each ingredient will taste in a hypothetical 3D cake, but it's all from scratch, so only talented chefs need apply! Midway into mixing, you realize there's not enough flour and only a teaspoon of sugar and the eggs are rotten. Sure, a really great chef will still come up with something delicious. But it's not going to taste like Sony's instant brownie mix, and it's going to be a lot more work.

 

It's amazing how many tasty games we have, especially once you've seen the ingredients up close... Posted Image

 

- KS

all of them who have actually finished their 3D engines have said that more can be gotten out of the machine than what we've seen so far.

EDIT: That's typical coder talk right there, how many times have I read that over the past 20 or 25 years? "If I had my time over..." "I know how I'd do it better next time..." there's always a next time... and a next time... and... It's the Molyneux syndrome.

[sh3-rg]

What were the Virtual Xperience guys using it on? Val D'isere? It was used on Rayman? Where did they say they used it?

ZzyorxxII or whatever it was called. There's a French interview out there somewhere.

 

That's 2D. My understanding is the GPU in main is only a real boost when using it for stuff like true 3D polygons.

True 3D polygons, as opposed to those bootleg polygons of questionable dimensions down the market? ;0)

 

I'm not going to answer it, I can't erase a decade or more of indoctrination in one post

[LinkoVitch]

 

ZzyorxxII or whatever it was called. There's a French interview out there somewhere.

 

 

True 3D polygons, as opposed to those bootleg polygons of questionable dimensions down the market? ;0)

 

I'm not going to answer it, I can't erase a decade or more of indoctrination in one post

 

I went and did science

 

Rather than derail this thread with it I have stuffed the related stuff to do with GPU in main over here : http://atariage.com/forums/topic/222589-gpu-in-main-science/

[+CyranoJ]

Science and facts will never beat out mythology and cults. Ask the Pope

 

GPU in main has but ONE use.

 

A C compiler targeted specifically to GPU in main.

 

Which doesn't exist.

 

And never will.

 

So it's useless.

 

RIP.

Edited February 27, 2014 by CyranoJ

[sh3-rg]

 

Great analysis. I guess my greater point was not so much the lack of AAA titles - I think that's clear. I think it's also clear (and you provided great examples) that we can point to specific games that make good use of some of the Jaguar's features. I suppose my REAL point was being able to point to a game that uses the vast majority of the Jaguar's features and really pushes the platform to say (at least) 90% of its absolute potential. Your example of Super Burnout was good - it's a technically sound game - but does that really push the Jaguar to its limits? I somehow don't think so. You pointed to Rayman -- even on less 2D friendly systems it was replicated,and, while it looks nice (really nice), I can't point to any one thing about it that pushes the Jaguar to my non-technical eyes. Tempest 2000 is obviously a classic and just about perfect on the platform, but again, it's deceptively straightforward audio-visually and again was easily replicated on other platforms.

 

One of the games that comes to mind for me is Zero 5. Though it's basically background-less, it seems like a technical showcase. Is that pushing the Jag to its limits? Battlesphere Gold? I don't know. Is there something you can then point to in your opinion that meets that criteria, that this is more or less the ceiling of the Jaguar?

I get where you're coming from. See the above quotes re: Super BurnOut, that's why I feel it's the best example of a game that ticks every box. 60frames, looks great, plays great, sounds great, exclusive and basically lives up to all the promise of the games we played years earlier on other systems that were pale imitations of real arcade experiences. But more than that, it scores on the technical side, even if maybe it doesn't appear to do so to those who play it and dismiss it as "just" a 2D game. It's the ultimate great Jaguar game doing what Jaguar was designed to do best. It's a shame so many wanted it to do mostly other things, Atari included ;0)

 

The thing about a fabled "90%+ power" game is it's approaching the question of game creation from a really awkward perspective. Starting with tech first isn't a comfortable position to be in, it's not conducive to creativity, but there are rare exceptions. It's healthier and more enjoyable to have ideas first, think how they can be realised within the target platform, tinker, see what works what doesn't and eventually your alpha is your beta is a release. Impressive tech + some tricks with such a processor + this great thing somebody came up with that you heard about + you know that thing that allows for 50x the amount of blopperflops per jubberbyte that the other thing does... I don't know what kind of game that would result in...

 

 

but I still bet it'd sell 100 carts!

 

 

Zero 5 is a good example, you're right. Not something I'd instantly think of as I don't own it and I personally don't like it having played it a few times. That was created by one guy, an Amiga demo coder, pretty much by himself, only working to a vague idea of what the game should be. There are some parallels to the development of Super BurnOut. Universally seen as technically impressive but this one is enjoyed by a smaller group.

 

First and foremost, game has to do game right and everything else has to fit around that. I guess that's an opinion many on AA share being a site built around a love of the older and simpler experiences that had to rely on that first and foremost... and probably why many of those who don't look in here see no reason to

[JagChris]

 

 

EDIT: That's typical coder talk right there, how many times have I read that over the past 20 or 25 years? "If I had my time over..." "I know how I'd do it better next time..." there's always a next time... and a next time... and... It's the Molyneux syndrome.

 

So its coder talk. How does that disqualify what is said? Its the very point to this whether the Jaguar has been pushed its length or not. I don't understand why you would discount that.

[JagChris]

It was used for Surrounded without a C compiler so it does not appear to be useless. Gorf, Tbird and AOwl where the only ones who did 3d so other than that from what I understand there is no real benefit to using it.

 

GPU in main has but ONE use.

 

A C compiler targeted specifically to GPU in main.

 

Which doesn't exist.

 

And never will.

 

So it's useless.

 

RIP.

 

 

 

Science and facts will never beat out mythology and cults. Ask the Pope

 

This is condescending and emotionally leading to the audience. This is how flamewars start. Be nice.

[JagChris]

I'm not sure what that means, are you saying kskunk didn't release anything so his findings are irrelevant? That's one opinion I suppose. I prefer to look to this:

 

Yes we are all aware of what Kskunk said but it still chooses to ignore the point that others have dealt with the same kitchen of ingredients, slugged it out to the finish line and when the dust settled they stated the Jaguar wasn't finished yet. And I'm not talking about any supposed benefits of GPU in main. This is BEFORE all that. Even without it they claimed more could be had.

 

You are choosing to look at one guy who did not slug it out to the finish line. I'm looking at others who were just as brilliant and finished.

Edited February 28, 2014 by JagChris

[JagChris]

Main ram is slow, very slow compared to the SRAM on the riscs, even just that fact makes gpu in main always slower than running in local! It never will offer a speed increase! It is only likely to offer a boost to convenience given larger code pace, or if perhaps you are using gpu ram for something else! This is ignoring the fact that you are now also using the sole bus of the system to run code over, whilst you do that nothing else is using it. If a higher priority CPU comes along then it will likely have to change memory pages too, more delay!

 

Better solution would be to only run code in local, and blit in any extra chunks when they are needed. You can get an aweful lot of code and stuff into 4K anyway!

 

Really? You are saying that for 3D it would not offer any speed increase the way its used by AtariOwl or Gorf for Surrounded. And I'm not talking about running the time critical stuff out in main. You mean to say that AtariOwls work could be duplicated without using GPU in main? Or the performance of Surrounded can be matched without it?

[+CyranoJ]

Maybe that could have been a lot better if it hadn't used it, because frankly, I don't consider that a good example of anything other that proof that code can run in main. I could code that entire game, in RAPTOR, in about 30 minutes, using sprites. And it would run at a solid 60fps.

 

Also, doesn't it use Atari's render code? I'd assume it does, or Gorf wouldn't have spent all those years asking someone else to write him a better one. Pretty sure none of that runs GPU in main. So, it's only the logic that runs on the GPU in main, not any of the 3D engine? Correct?

 

As for being condescending, nope. It's called an opinion.

 

It was used for Surrounded without a C compiler so it does not appear to be useless. Gorf, Tbird and AOwl where the only ones who did 3d so other than that from what I understand there is no real benefit to using it.

 

 

This is condescending and emotionally leading to the audience. This is how flamewars start. Be nice.

Edited February 28, 2014 by CyranoJ

[JagChris]

Sprites? Actually that would be cool to see! I wouldn't mind seeing that at all.

 

However the point is for it to be in 3D and how GPU in Main ram is a benefit in that instance. If you can match Surroundeds' performance using polygons without GPU in Main then you will have proven your point and my hat will be off to you guys. Until then I'm sorry. I'll be of the opinion that for 3D the Jaguar will benefit from using the techniques used by Gorf and AtariOwl.

 

Maybe that could have been a lot better if it hadn't used it, because frankly, I don't consider that a good example of anything other that proof that code can run in main. I could code that entire game, in RAPTOR, in about 30 minutes, using sprites. And it would run at a solid 60fps.

 

As for being condescending, nope. It's called an opinion.

 

 

It's a condescending insulting opinion. Its meant to emotionally charge the reader into thinking 'watch out, this guys in a cult!' with all the negative imagery that conjures up. So there is a chance the reader will have already formed an opinion of me without actually reading what I'm trying to say.

[+CyranoJ]

Please read above. Uses Atari Render Engine, does not draw polys using GPU in main. Your argument is invalid.

[JagChris]

Well it uses a modified version of the Render engine and they had to manually do the GPU in Main but yes it does use it. I was a tester on it.

 

Please read above. Uses Atari Render Engine, does not draw polys using GPU in main. Your argument is invalid.

[+CyranoJ]

You draw them, I'll code up "Pick a random number between 1 and 6" - let's see who takes longer? hehehe.

 

 

Sprites? Actually that would be cool to see! I wouldn't mind seeing that at all.

 

 

[+CyranoJ]

Would you like me to disassemble it? Or would proof that the render code runs on the 68000* shatter the universe?

 

*just a guess here, not looked - but whatever.... the Atari Render code for sure does not run GPU in main.

 

Oh wait, a spinning Christmas Tree is telling me not to do that. LOL

 

Well it uses a modified version of the Render engine and they had to manually do the GPU in Main but yes it does use it. I was a tester on it.

 

Edited February 28, 2014 by CyranoJ

[JagChris]

CJ if you can prove that it is not running in main in whatever way you see fit than that's fine as well. Absolutely.

 

Would you like me to disassemble it? Or would proof that the render code runs on the 68000* shatter the universe?

 

*just a guess here, not looked - but whatever.... the Atari Render code for sure does not run GPU in main.

 

Oh wait, a spinning Christmas Tree is telling me not to do that. LOL

 

 

The overall gist that I can remember of the modified renderer that JagMod was working on and using was that it was adapted to run in main for a majority of the code. Some of the renderer was indeed still acting like it was not running in main and was waiting around for some things to get done that it did not need to wait on anymore. I believe Jagmod said it would be easier to rewrite the engine from scratch than modify the renderer anymore to completely shed all its old bad habits that were holding things up.

 

If you can disprove any of this conclusively and shed light on some kind of deception that JagMod and Gorf were perpetrating on us then please let us know. I would really like to know the truth if we were being hoodwinked.

Edited February 28, 2014 by JagChris

[+CyranoJ]

You know what? I can't be arsed.

 

If reading facts about code that doesn't even access memory, running in main ram, being 90% slower than running in local doesn't persuade you this is all a cock-and-bull smokescreen, then nothing will convince you.

 

If you can't see that running from local SRAM at the speed of the GPU is faster than breaking out over the bus using much slower main ram and halting the other processors while your doing it, nothing will convince you.

 

If you can't see that a 3D engine is exactly the same as a 2D engine, is the same as any string of instructions... as far as bus arbitration goes, then nothing will convince you.

 

If you turn off the 68000, and run your GPU code in main ram, then yes, it's faster than the 68000. It is not faster than running code in local, though. That is probably why AO did that. His code probably didn't fit in local, and turning off the 68000 and running it in main probably boosted it a tad.

 

If you run your code in local, well golly gosh! It'll run pretty darn quickly - and you can also use the 68000 for something else. Doesn't that sound wonderful? Maybe it's why they did that?

 

As for your leading comments about deception, intended purely to stir the pot up - I've not said, anywhere, that JagMod or Gorf were trying to be deceptive - they fully admitted the game logic was running from GPU in main. And I believe them. The render code, however, is firmly from Atari and I don't believe for a nanosecond that they modified it to run GPU in main. If that were the case, Gorf wouldn't have wasted a decade waiting for someone more capable to code him a new render engine, because he'd already have one. This point invalidates your argument that Surrounded is proof that GPU in main is good for 3D. Completely.

 

I'm not wasting my time here on this subject any more. This forum already has 10+ years of utter, total and complete nonsense written on this subject by 'The cult of GPU in main' that it doesn't need any more input, from anyone.

 

People who know (Including AO, Linkovitch, Dr Typo, SCPCD, Zerosquare and yes, myself) have multiple times said how foolish this belief is. Yes, there are edge cases, but edge cases don't write game engines. I can't think of a single reason why anyone who follows science and logic should think otherwise and continue to believe in the untapped potential of the Jaguar GPU in main myth.