Optimizing Wolf3D-style rendering

[Don_Adan]

I dont know, if unfinished Wolf3D clone can be useful for something. Anyway here are source code and demo of "Wilczy Szaniec", coded by Hali/Union. https://www.ppa.pl/rodzynki/wilczy-szaniec.html

[rothers]

Originally Posted by AndNN

ACTUALLY THIS HAS LEFT A BAD TASTE FOR ME AND I WONT BE BACK AGAIN...

Well that escalated fast.

I've never seen Wilczy Szaniec before, the youtube video looks good. Is it in 8 colours?

[LaBodilsen]

So happy to see this thread springing back to life, only to discover it's because someone have a bad taste of pride.

@AndNN: instead of claiming first idea/copyright on some idea, that in it's very definition is very broad. how about helping getting this game on to Amiga, and make a collaboration? and then getting your credits in the actual game, because you made some code for it, and not for the idea it self.

I've read your earlier threads, and they where full of many ideas, but very little work on code. Britelite atleast but some code forward, and didnt even complain when i disassembled the code to try out some stuff myself.

and btw: your code using fastmem, is not for this thread, as this is targeted A500 +512kb fakefastmem.

i don't wanna argue with you, and as mentioned if people send you PM's calling you shit. that is unacceptable. I can almost feel your pain, when i read your posts. But try and step back a little, and then show these naysayers that what you say is actually the truth, be providing proof of it. we need all the help we can get, if this game is ever gonna be made for Amiga500

[sandruzzo]

I think there is no need to have rotate chunky buffer. With some little arrangment we can have linear texture mapping instead colum. If we divede a wall, into small walls, we can shoot less rays, and have some kind of affine mapping with correct every n pixel...

[britelite]

Originally Posted by sandruzzo

I think there is no need to have rotate chunky buffer. With some little arrangment we can have linear texture mapping instead colum. If we divede a wall, into small walls, we can shoot less rays, and have some kind of affine mapping with correct every n pixel...

I think you're confused as to why the chunky buffer is rotated 90 degrees, as it has nothing to do with the raycasting. It's just to make the rendering of the walls faster, as you can use post-increments for drawing the stripes instead of offsets.

But the raycasting could of course be made faster with a combination of raycasting and interpolation.

[sandruzzo]

Originally Posted by britelite

I think you're confused as to why the chunky buffer is rotated 90 degrees, as it has nothing to do with the raycasting. It's just to make the rendering of the walls faster, as you can use post-increments for drawing the stripes instead of offsets.

But the raycasting could of course be made faster with a combination of raycasting and interpolation.

If you fallow my idea you can draw wall faster without rotating. I'm not confused, I understand it.

[britelite]

Originally Posted by sandruzzo

If you fallow my idea you can draw wall faster without rotating. I'm not confused, I understand it.

In that case you really have to elaborate on your idea, because to me it just seems you're mixing up two different things.

EDIT: or are you suggesting that texture-mapped triangles should be used?

[sandruzzo]

Originally Posted by britelite

In that case you really have to elaborate on your idea, because to me it just seems you're mixing up two different things.

EDIT: or are you suggesting that texture-mapped triangles should be used?

The Idea is That: mixig the two world if worth it. Not triangle per se, but quads. With simple linear interpolation, you can texture a wall

[britelite]

Originally Posted by sandruzzo

The Idea is That: mixig the two world if worth it. Not triangle per se, but quads. With simple linear interpolation, you can texture a wall

I don't see how that would be faster than my current solution when it comes to rendering speed, so I'm eagerly waiting for your proof-of-concept demo

[DanScott]

Originally Posted by sandruzzo

The Idea is That: mixig the two world if worth it. Not triangle per se, but quads. With simple linear interpolation, you can texture a wall

In theory, it sounds ok, but how do you deal with where 2 different wall textures meet, where the 2 adjacent rays intersect some way either side of 2 different textures.

Corners and edges could also be a problem.

I fear that any code that looks for (and deals with) these cases could probably negate any speed gain, most likely will slow down the whole thing overall.

[britelite]

Originally Posted by DanScott

In theory, it sounds ok, but how do you deal with where 2 different wall textures meet, where the 2 adjacent rays intersect some way either side of 2 different textures.

Corners and edges could also be a problem

You could speed up the raycasting by, for example, casting every 4th ray first, and then check if they happen to land on the same wall in which case the 3 points in between would be interpolated. If they're not hitting the same wall, then cast the middle ray and check again and possibly interpolate, or then cast the remaining two. Worst case scenario would be slightly slower than just straight up casting all rays, but in most circumstances it would at least gain a slight speed boost.

[DanScott]

That makes sense.


But the pixels still need calculating and plotting, and your method of writing the strips is highly optimal, more so than could be achieved by interpolating across??

I guess sandruzzo can try this out and see if it works

[britelite]

Originally Posted by DanScott

I guess sandruzzo can try this out and see if it works

Indeed, and I'd be happy to be proven wrong

[sandruzzo]

Since we know that move.b x,(a0)+, is faster than move.b x,40(a0), this is the key of linear interpolation.

Just my 2 cents

[sandruzzo]

Originally Posted by britelite

Indeed, and I'd be happy to be proven wrong

I don't prove any wrong friend. You're doing a hell of job so far. Just to see if this idea could help.

[britelite]

Originally Posted by sandruzzo

Since we know that move.b x,(a0)+, is faster than move.b x,40(a0), this is the key of linear interpolation.

Yes, and that's why I'm doing move.w (a1)/(a1)+,(a0)+

[sandruzzo]

Originally Posted by britelite

Yes, and that's why I'm doing move.w (a1)/(a1)+,(a0)+

Maybe since your engine is allmost done you could try this solution and see if it works

[britelite]

Originally Posted by sandruzzo

Maybe since your engine is allmost done you could try this solution and see if it works

I don't need to try it to know that rendering quads is slower than the way I do it. If you take a moment to think about how you interpolate u/v in quad, you will probably realize it too.

I do agree with you that there are ways to make the raycasting part more efficient, but I don't agree on your choice of rendering.

[rothers]

Would there be any advantage to precalculating a certain number of zoom levels on each texture? Assuming there is memory to spare.

Just it looks like Wolfenstein uses the same wall texture over and over again.

[britelite]

Originally Posted by rothers

Would there be any advantage to precalculating a certain number of zoom levels on each texture? Assuming there is memory to spare.

I've already done that (mip-mapping)