2D games with 3D elements

[alucardX]

3 hours ago, agradeneu said:

3D calulation in a 2d game might help with fast rotation, tilting effects of texture images and so on. Or some simple 3d objects, rather small but with delicate texture detail.

 

I think you understand my idea pretty well. Having something like this might just make the game visuals pop.

[Songbird]

This is not the same as the original poster's request, but Phase Zero is as far as I can tell a completely 2D sprite-based engine with a voxel map for the terrain. It's a really cool effect and the game is almost too fast when you play it -- easy to get disoriented.

[alucardX]

1 hour ago, Songbird said:

This is not the same as the original poster's request, but Phase Zero is as far as I can tell a completely 2D sprite-based engine with a voxel map for the terrain. It's a really cool effect and the game is almost too fast when you play it -- easy to get disoriented.

I somehow always forget about Phase Zero. The effects in that game are really awesome.

[Austin]

4 hours ago, alucardX said:

I think you understand my idea pretty well. Having something like this might just make the game visuals pop.

I can't think of a single Jaguar game that does what you are talking about, but I know what you mean.

 

I always appreciated Knuckles Chaotix on the 32X, where it sparsely incorporates polygons into the 2D visuals and gameplay. Most notable is on the carnival themed stages, where there's a polygonal lift. Pretty cool stuff. I wish more games of this era did that.

 

[alucardX]

20 minutes ago, Austin said:

I can't think of a single Jaguar game that does what you are talking about, but I know what you mean.

 

I always appreciated Knuckles Chaotix on the 32X, where it sparsely incorporates polygons into the 2D visuals and gameplay. Most notable is on the carnival themed stages, where there's a polygonal lift. Pretty cool stuff. I wish more games of this era did that.

 

This kind of thing is exactly what I'm thinking. Using wire frame models in some places would be cool too.

[Shaggy the Atarian]

13 hours ago, agradeneu said:

Every 3D game has some sprites (e.g. explosions or all objects like in AVP).I think he meant something like a 2D shooter/environment with a few stand out 3D objects, 

Like Thunderforce V. That havent been done yet on the Jag. (No polys in Breakout 2000 IMO)

B2K just comes to mind as at the beginning of each stage it first populates the bricks with a vector-looking wireframes.  Since the camera is static and nothing interesting happens with the bricks otherwise, hard to say.  But I get what he means now

 

 

And now to keep in spirit with the real theme of the thread

 

 

[Jag64]

If I was lazy, which I am, and knew the slightest about Jaguar coding, *which I don't, I'd just try to replicate that type of action by pre-rendering the 3d objects and calling them to on a track or spline. Maybe I'd do that with a camera. I don't know, *I'm not a programmer. Either way, I'm getting an on-rails scroller vibe. That's not a bad thing. Just think there might be a few ways to pull that off easily. And it could look nice. I feel like there is a sht ton of number-crunching coders on here, but are there not many artists? This is an Atari site, so if you click a thread, there's probably four different dudes with synthesizers; audio's covered. I'm eager to see your idea as it develops and hope some of the more Jag-versed members can maybe get you on the path to seeing it released.

 

*?

5 minutes ago, Shaggy the Atarian said:

B2K just comes to mind as at the beginning of each stage it first populates the bricks with a vector-looking wireframes.  Since the camera is static and nothing interesting happens with the bricks otherwise, hard to say.  But I get what he means now

 

Man! Just thinking of all the fun themes you could populate a new Breakout clone with. But also, I really like that vector/static mash-up. Someone could use that to make some fun cyberpunk themed games.

Edited May 6, 2022 by Jag64
Add a reply quote instead of creating a new reply

[agradeneu]

1 minute ago, Shaggy the Atarian said:

B2K just comes to mind as at the beginning of each stage it first populates the bricks with a vector-looking wireframes.  Since the camera is static and nothing interesting happens with the bricks otherwise, hard to say.  But I get what he means now

 

 

And now to keep in spirit with the real theme of the thread

 

 

I dont think someone would write 3d code just to do ....nothing with it;-)

 

She appears a little modest compared to what we are used to in this thread 

[alucardX]

On 5/6/2022 at 6:33 PM, Jag64 said:

I'm eager to see your idea as it develops and hope some of the more Jag-versed members can maybe get you on the path to seeing it released.

I am throwing a couple of concepts around right now. Thanks for the encouragement.

[ArneCRosenfeldt]

On 5/6/2022 at 2:04 AM, Stephen said:

Please keep 75% to 80% of the polygons dedicated to the boobs

A lot of spans of those polygons will span less than 4 pixel. The manual says that you have to software render (JRISC) the first 3 pixel on the GPU. This could just as well happen on a 32x. At least with only 4px you can fix all increments at 0.0 which simplifies the setup. I still have not understood Gouraud. You know the normals of the faces, not those of the vertices. At least with LoD use the high detail faces joined at the vertex. Why I love hexagons, a slightly irregular mesh is better for LoD and for the average function: Some vertices divide by 4 and some by 8, but none by 6. Though we need fast Quake normalize anyway?

 

I would even say that you could use a regular grid for the surface and by -- yeah quadtree -- we identify regions where all faces face the camera and moving along +x on screen moves along one of two axes in the grid ( one of the 4 diagonals ). The latter is to reduce the number of tests in our:  Raytracer . So basically, you Bresenham check the two exit edge candidates ( costs two adds per pixel ) . Keep all vertices in SRAM so that we can easily move to the next phase. Indepentently, every time 4 px have been calculated, calle StoreP . Set high GPU priority to make sure that the hi register was already flushed.

 

This allows to let the blitter do other stuff in the background, but we have no z-buffer. Z-buffer on GPU is certainly much slower than on the blitter.

 

Quote

Enables an "extra" source data read at the start of an inner loop operation

Theoretically (ha), we could not enable SRCENX and SRCENX in the blitter and use 00 - Add phrase width (truncate to phrase boundary) to write 8 px at once. We could just let it write phrase aligned. We only need z to sort relative to the background .. far away. Thus z needs only be sampled every 8 px. I think in some modes z incr. is added also in the last iteration. Z-increment is a single 32 bit register in constrast to F02258 ..  F02267

Edited May 9, 2022 by ArneCRosenfeldt
more

[Jag64]

4 hours ago, ArneCRosenfeldt said:

A lot of spans of those polygons will span less than 4 pixel. The manual says that you have to software render (JRISC) the first 3 pixel on the GPU. This could just as well happen on a 32x. At least with only 4px you can fix all increments at 0.0 which simplifies the setup. I still have not understood Gouraud. You know the normals of the faces, not those of the vertices. At least with LoD use the high detail faces joined at the vertex. Why I love hexagons, a slightly irregular mesh is better for LoD and for the average function: Some vertices divide by 4 and some by 8, but none by 6. Though we need fast Quake normalize anyway?

 

I would even say that you could use a regular grid for the surface and by -- yeah quadtree -- we identify regions where all faces face the camera and moving along +x on screen moves along one of two axes in the grid ( one of the 4 diagonals ). The latter is to reduce the number of tests in our:  Raytracer . So basically, you Bresenham check the two exit edge candidates ( costs two adds per pixel ) . Keep all vertices in SRAM so that we can easily move to the next phase. Indepentently, every time 4 px have been calculated, calle StoreP . Set high GPU priority to make sure that the hi register was already flushed.

 

This allows to let the blitter do other stuff in the background, but we have no z-buffer. Z-buffer on GPU is certainly much slower than on the blitter.

 

Theoretically (ha), we could not enable SRCENX and SRCENX in the blitter and use 00 - Add phrase width (truncate to phrase boundary) to write 8 px at once. We could just let it write phrase aligned. We only need z to sort relative to the background .. far away. Thus z needs only be sampled every 8 px. I think in some modes z incr. is added also in the last iteration. Z-increment is a single 32 bit register in constrast to F02258 ..  F02267

[ArneCRosenfeldt]

Muse means that for every NSFW post you upload some code on GitHub . I am happy with my typeScript project, and it seems that I'll earn some money with BASIC the next 6 years ..
Also, I did not get the mechanical simulation to work. I want perfect anisotropy. So either use random particles or this Stress and Strain Tensor and eigen values. Also you want stiffness on the skin. And you need a lot of samples => run on Jerry and dial back on the sound. You know how the manual says that Jerry can push 16 bits per scanline to Tom if DSP priority is below OP priority (use a timer with PLL, or what?)? So order by scanline already on Jerry so that Tom is finished with the top vertices in-time and all with low latency ( minimal scanline advance ).

[ArneCRosenfeldt]

3 hours ago, ArneCRosenfeldt said:

Muse means that for every NSFW post you upload some code on GitHub . I am happy with my typeScript project, and it seems that I'll earn some money with BASIC the next 6 years ..
Also, I did not get the mechanical simulation to work. I want perfect anisotropy. So either use random particles or this Stress and Strain Tensor and eigen values. Also you want stiffness on the skin. And you need a lot of samples => run on Jerry and dial back on the sound. You know how the manual says that Jerry can push 16 bits per scanline to Tom if DSP priority is below OP priority (use a timer with PLL, or what?)? So order by scanline already on Jerry so that Tom is finished with the top vertices in-time and all with low latency ( minimal scanline advance ).

for the eigen vector we need the eigen values. As long as they are not degenerated, one can use Householder transformation to iterativly follow them on every time step. When they come out of degeneration, we need to find the root of the characteristic polynom. Polynom up to 4th order can be solved analytically. We need SQRT for this. A lookup table is enough because we iterate for precision.

 

Shear modulus is what we are interested in. Compression modulus is responsible for the speed of sound. Unless we want sound, we can iterate slower. Still, ideally, simple JRISC code optimized to run at as short time steps as possible. Then add some over relaxation or matrix solver. Right now I am working at a matrix solver for physical fields. I want to experiment with a hybrid of LU and iterative.

[alucardX]

4 hours ago, ArneCRosenfeldt said:

for the eigen vector we need the eigen values. As long as they are not degenerated, one can use Householder transformation to iterativly follow them on every time step. When they come out of degeneration, we need to find the root of the characteristic polynom. Polynom up to 4th order can be solved analytically. We need SQRT for this. A lookup table is enough because we iterate for precision.

 

Shear modulus is what we are interested in. Compression modulus is responsible for the speed of sound. Unless we want sound, we can iterate slower. Still, ideally, simple JRISC code optimized to run at as short time steps as possible. Then add some over relaxation or matrix solver. Right now I am working at a matrix solver for physical fields. I want to experiment with a hybrid of LU and iterative.

And how do you feel about your mother?

[Lost Monkey]

8 hours ago, ArneCRosenfeldt said:

for the eigen vector we need the eigen values. As long as they are not degenerated, one can use Householder transformation to iterativly follow them on every time step. When they come out of degeneration, we need to find the root of the characteristic polynom. Polynom up to 4th order can be solved analytically. We need SQRT for this. A lookup table is enough because we iterate for precision.

 

Shear modulus is what we are interested in. Compression modulus is responsible for the speed of sound. Unless we want sound, we can iterate slower. Still, ideally, simple JRISC code optimized to run at as short time steps as possible. Then add some over relaxation or matrix solver. Right now I am working at a matrix solver for physical fields. I want to experiment with a hybrid of LU and iterative.

Uh... does it tits?

[Gunther]

 

 

Edited May 22, 2022 by Gunther

[+Stephen]

4 hours ago, Gunther said:

 

 

See - play to the system strengths.  60FPS looks beautiful, buttery smooth.  Great particle effects, nice 8-way scrolling.

[Gunther]

This effect would look excellent on a dungeon crawler/r.p.g.

It reminds me of effects used in the flying sections of Genesis game, Batman and Robin.

A 2.5D paltformer would be excellent.

It's a shame ATARI was trying to rush-it out unfinished, as I'm sure it would've been a great game.

The developer page - http://www.leonik.net/dml/sec_lw.py - is an interesting read.

 

 

 

Edited May 23, 2022 by Gunther

[+cubanismo]

That does look cool. Was a build ever released to the general public? The note at the bottom of the developer's page doesn't leave me hopeful, but thought I'd ask.

[JagChris]

2 hours ago, cubanismo said:

That does look cool. Was a build ever released to the general public? The note at the bottom of the developer's page doesn't leave me hopeful, but thought I'd ask.

I think he said he was going to provide a download link but if he didn't he probably just forgot.

 

You can DM him if you wish.

 

 

[Isgoed]

14 hours ago, cubanismo said:

That does look cool. Was a build ever released to the general public? The note at the bottom of the developer's page doesn't leave me hopeful, but thought I'd ask.

If I recall correctly, some years ago was it Carl of Songbird Productions who had a demo/alpha/beta of this game on an alpineboard, but lost it?

 

Please correct me if I am wrong.

 

 

 

[Gunther]

 

 

 

 

Edited May 28, 2022 by Gunther

[ArneCRosenfeldt]

On 5/10/2022 at 8:13 PM, Lost Monkey said:

Uh... does it tits?

People seem to do it in Blender now all the time and thus it is not so impressive. Maybe better leave Jerry to the music and homebrew kid friendly ( no drugs, nor guns .. ).