SNES Doom Amiga Back-up Disks

[Retro1234]

I mean I don't know the full story but did Doom actually run perfectly on a NEXT 68030?

[Karlos]

Originally Posted by Retro1234

I mean I don't know the full story but did Doom actually run perfectly on a NEXT 68030?

I don't know about 030, but it was developed on a 68040 next box. The DOS release was a port.

[Retro1234]

I mean to say the Amiga wasn't powerful enough seems crazy, the Amiga 68000 might not of been powerful enough but an equivalent 68040 etc

[Retro1234]

Also anyone got any videos or anything of it running on a NEXT box ?

[Karlos]

Originally Posted by Retro1234

I mean to say the Amiga wasn't powerful enough seems crazy, the Amiga 68000 might not of been powerful enough but an equivalent 68040 etc

Again, he didn't say that. He explicitly said Doom needed the full power of the 040 on a machine with chunky pixels already and that trying to render to a planar architecture was a non starter.

He's still right about that - show me a single Amiga texture mapped 2.5D FPS game that renders pixels straight to bitplanes. They all either use copper hacks to simulate an ultra low resolution chunky display *or* depend on rendering to a chunky buffer and converting that directly.

[Thorham]

Originally Posted by Karlos

depend on rendering to a chunky buffer and converting that directly.

Does that matter if it works?

[Karlos]

Originally Posted by Thorham

Does that matter if it works?

It matters in context of when he said it, yes. Generally, rendering to a buffer that you then have to transfer again to another one (just to be able display it) in software is a massive bag of suck. You just wouldn't do it if you weren't forced to at gunpoint. You do it because you generally have rendering tricks that need the fastest memory accesses to be affordable, especially if you need to do things like transparency that require reads back. You'll notice Doom doesn't have that, the only transparency it has are completely see through texels.

Now imagine that your transfer operation isn't even a simple memory copy for that second transfer you didn't even want to do in the first place, but some god-awful scattering of bits from perfectly sensibly arranged bytes to completely different memory locations. However you cut it, what the said was accurate given the state of the art in 1993. There were absolutely Amigas that could have ran a bonafide Doom port then, but very few people had them.

We got lucky with C2P, not because it's a good solution in practise, but because our chip ram write bandwidth is so damn slow that the cost of the operation can be hidden behind the pending writes. If you arrange your code correctly, you can get the conversion almost for free - but only compared to a direct copy to chip.

The graphics architecture we had, just wasn't intended for pixel by pixel CPU-driven rendering. It was intended for hardware blitting, scrolling, beam racing and all the things that made it, and continue to make it awesome. Brute force pixel computation just isn't how it's supposed to be used.

[oscar_ates]

And commodore could not improve that chipset in 10 years except chucking the aged 68020 in it

[Thorham]

Originally Posted by Karlos

It matters in context of when he said it, yes.

It doesn't. You do what you must to get it to work properly, and work properly it does regardless of what anyone said about it back then.

That said, no, the solution isn't exactly ideal (needing c2p sucks).

[pipper]

Also, be aware that the C2P technique was largely unknown in '92/'93 and it took years for the demo scene to refine the algorithm down to copy speed. Asking for Carmack to think of that and implementing it perfectly on a platform he didn't particularly value is a stretch.

[Thorham]

Originally Posted by pipper

Asking for Carmack to think of that and implementing it perfectly on a platform he didn't particularly value is a stretch.

Absolutely, but he was still wrong

[Karlos]

Originally Posted by pipper

Also, be aware that the C2P technique was largely unknown in '92/'93 and it took years for the demo scene to refine the algorithm down to copy speed. Asking for Carmack to think of that and implementing it perfectly on a platform he didn't particularly value is a stretch.

This. Also, Carmack was a high level language guy primarily, algorithm first. When it comes to the hard cycle counting optimisation tasks identified through profiling, seasoned experts like Abrash were hired. I just couldn't picture Carmack being interested in figuring out C2P tricks.

[Karlos]

Originally Posted by Thorham

Absolutely, but he was still wrong

What's your definition of "work properly", btw? His was to run fluidly on 486 level PC in 256 colours and be somewhat playable on 386. I don't think a 10fps AGA port on 020 is what he had in mind.

[AestheticDebris]

Originally Posted by Karlos

There were fully playable Amiga doom ports within days of the source going open.

But only because there had been 3-4 years of developers trying to figure out every trick under the sun to make Doom style engines render on Amiga with reasonable performance. It wouldn't have been feasible to devote that kind of effort into a commercial port which would have, at best, very limited sales.

[Thorham]

Originally Posted by Karlos

What's your definition of "work properly", btw? His was to run fluidly on 486 level PC in 256 colours and be somewhat playable on 386. I don't think a 10fps AGA port on 020 is what he had in mind.

68030 50mhz in low detail mode? Should be similar to a suitable enough 386.

[Thorham]

Originally Posted by AestheticDebris

But only because there had been 3-4 years of developers trying to figure out vlevery trick under the sun to make Doom style engines render on Amiga with reasonable performance.

There's no trick. All you need is c2p and that's just a bunch of transposes. Still takes time to figure out how to do efficiently, of course.

[Megalomaniac]

It's true that Doom's entire concept was something that did not suit Amiga architecture. Commodore had little hope of seeing Doom coming - it was still a year away when the A1200 launched, you had Wolfenstein 3D but that wasn't a true killer app.

Besides, if you can get 10fps on a stock A1200, you can easily get 20fps (which is at least 'somewhat playable') on a 25Mhz 030 with some fast RAM (which I guess would be equivalent to a good 386SX or a minimal 386DX?), with 256 colours. And plenty of people considered 10fps acceptable in those days for 3D games. Plenty of PC games (and some Amiga games, e.g. Red Baron) had minimum specs on the box with which you'd be lucky to get anywhere near 10fps.

And, while I know I've said this before, Pc reviews usually ignored the fact that most people had PCs nearer the minimum spec than the recommended one, whereas Amiga reviews (while usually accepting of games which needed hard drive installation to not have too much loading for most people) usually went too far the other way and ignored the existence of faster Amigas, or accelerators.

[Karlos]

I don't think 10fps on stock A1200 is realistic for an official doom port in 1993/4. And by stock, I mean an A1200 with 4MB or so additional fast RAM needed to just run the thing.

[Marconelly]

Originally Posted by Thorham

Doom level 1 exists for the Dread engine, including levels of elevation.

I'd really like to see/try this! Is it available for free? I didn't realize Dread engine supported different elevations.

[Thorham]

Originally Posted by Marconelly

I'd really like to see/try this!

Doom part at 5:56

[ Show youtube player ]

Originally Posted by Marconelly

Is it available for free?

More info is in the video.