rts/Map/SM3/Plane.h wrote: typedef float3 Vector3;
SM3
Moderator: Moderators
Re: SM3
doh! :D
thanks Kloot!
i was looking for these:
i forgot how typedefs work again :D
just did not yet use them often enough.
thanks Kloot!
i was looking for these:
Code: Select all
class Vector3
struct Vector3
typedef Vector3
just did not yet use them often enough.
Re: SM3
I realize that you guys are about to take a bold step, entering dangerous terrain of hard labor, suffering and fear by doing the coding needed to make SM3 become reality. At least it seriously looks like that is happening.
I can only express my great gratitude over your fearless journey into the unknown and give you a sincere...
GOOD LUCK!!
(You may need it!)
I can only express my great gratitude over your fearless journey into the unknown and give you a sincere...
GOOD LUCK!!
(You may need it!)
Re: SM3
It may be of note that jelmer originally built SM3 in a standalone application so eh could test without meddling with the mess that is spring rendering. This may explain some quirks such as the above.
One of the reasons he put forth for buggyness and slow progress was that porting the code over was annoying in that lots of rendering code in spring did not preserve state among other issues.
One of the reasons he put forth for buggyness and slow progress was that porting the code over was annoying in that lots of rendering code in spring did not preserve state among other issues.
-
- Posts: 916
- Joined: 27 Jun 2009, 01:32
Re: SM3
Considerung Kloot's efforts here and on making the engine handle over all needed input for normal mapping & stuff for models is there any way we could get an option for the new map format which would allow us to use normal or even parallax mapping with decals for some nice looking ground scars?
Re: SM3
Such an option wouldn't need to be exclusive to any new format. Decals (explosion scars, building plates, etc) are rendered as overlays independent of the map's type (SMF, SM3), adding normal + parallax mapping support could be done right now with a bit of work.
Re: SM3
IIRC tobi or maybe trepan/lurker, someone, posted a reply to the last time someone foolishly asked for parallax mapping, demonstrating how you get to use a huge tonne of your system resources, for no observable gain under most conditions.
Parallax mapping in conclusion, is worthless and hogs resources apparently. The post had pictures and explanations though, maybe someone should re-post it.
Parallax mapping in conclusion, is worthless and hogs resources apparently. The post had pictures and explanations though, maybe someone should re-post it.
-
- Posts: 916
- Joined: 27 Jun 2009, 01:32
Re: SM3
While it's one of the most performance eating techniques it also depends a bit on which exact method you use as "parallax mapping" in fact is a quite broad term. In the end it's up to a general request anyway to apply a custom shader to the decals so you can do all the things you want from animating stuff (like doing a strange distortion so it looks like molten) or doing some normal mapping lighting fake or do a displacement mapping (-> parallax style) to get some real geometry for a stronger and perspective safe effect...
Re: SM3
So far as I can tell, whoever posted that rant is an idiot. I can't even tell that they implemented a true shader there.
Parallax mapping + normals + specularity control (which I've already implemented into my Unit shader) looks spectacular, and while it's not free, it's certainly cheap, and with a SMF++ format, it would be pretty efficient.
We "just" need a SMF format that loads a second DDS DXT5 set of tiles, and a MapConv that does two sets of tiles- one for the diffuse texture, and the other one for the normalmap / depthmap.
Maps would get quite a bit bigger, filesize-wise, but that's the only serious disadvantage. The resulting geometry only has to be sent to the shader once. And if the shader is GLSL and you can specify what shader you can use... well... that would make a lot of things possible. You could use the alpha channel for glowmaps, reflectivity, etc. (assuming that we were able to get access to the cubemaps, that is).
And yeah, using normalmaps on groundscars, at least the LUPS variety, is very easy, I just haven't had time to implement it yet, and won't have time until I finish moving my content into the modern age.
Parallax mapping + normals + specularity control (which I've already implemented into my Unit shader) looks spectacular, and while it's not free, it's certainly cheap, and with a SMF++ format, it would be pretty efficient.
We "just" need a SMF format that loads a second DDS DXT5 set of tiles, and a MapConv that does two sets of tiles- one for the diffuse texture, and the other one for the normalmap / depthmap.
Maps would get quite a bit bigger, filesize-wise, but that's the only serious disadvantage. The resulting geometry only has to be sent to the shader once. And if the shader is GLSL and you can specify what shader you can use... well... that would make a lot of things possible. You could use the alpha channel for glowmaps, reflectivity, etc. (assuming that we were able to get access to the cubemaps, that is).
And yeah, using normalmaps on groundscars, at least the LUPS variety, is very easy, I just haven't had time to implement it yet, and won't have time until I finish moving my content into the modern age.
Re: SM3
I think he's just saying that the Parallax part is unnecessary - the rest of the shader stuff (normals + specularity) is almost-as-good and cheaper. I can't speak to the cheapness, but I see his point on the other stuff.
Either way, his screenshots show that it is good for a landscape, which is what we're talking about here. It may be wasted on unit skins, but it's definitely a good thing for the map itself.
Either way, his screenshots show that it is good for a landscape, which is what we're talking about here. It may be wasted on unit skins, but it's definitely a good thing for the map itself.
Re: SM3
The guy who originally posted that rant thats just been linked to is the guy who wrote the infinity universe engine, who knows MASSES more graphics kung foo than you do argh, this is the guy who has kilometer long spaceships taking off of planets going into orbit seamlessly, and dealing with multiple shadow systems and immense scales, aswell as procedurally generating graphics and objects on the fly.
Re: SM3
Here is a demonstration of why parallax is cool.
1. When not to use parallax: object is mainly very flat planes, and we don't want to create new geometry via shader:
As you can see, if you compare to my earlier shots, it's pretty much the same as a normalmap, but with a (very small) difference in performance.
2. Parallax is Ok, for minor details- resulting accuracy vs. lighting is better, and it's practically free:
Here, the bumps stick out and feel more natural and detailed. Pretty worthless at a distance, though.
3. Parallax is awesome, if you want to create whole surface geometry from your bitmaps and are deliberate about it. This was an accident with CrazyBump, but it serves as a good example- in the ARB shader, those "awnings" that are noticeably swollen outwards with parallax are flat- that entire section of the model is... two triangles.
Any other questions? Anybody else still not getting it? That guy is testing parallax, concluding it's worthless, and he obviously has no clue how to use it, or why it's fundamentally different than normalmaps. Parallax is what you use, when you want to auto-LOD meshes with incredible details, guys. And it's what you want to use when you want to, say, have a map's drawn geometry conform to the true-resolution heightmap, instead of an approximation.
But the key thing here is that it is different, requires a different approach to get the best use out of it, and it's really not expensive at all, assuming that it's just parallax, and not the giant bells-and-whistles of what I'm trying to do with Unit shaders.
It's just another tool, basically.
1. When not to use parallax: object is mainly very flat planes, and we don't want to create new geometry via shader:
As you can see, if you compare to my earlier shots, it's pretty much the same as a normalmap, but with a (very small) difference in performance.
2. Parallax is Ok, for minor details- resulting accuracy vs. lighting is better, and it's practically free:
Here, the bumps stick out and feel more natural and detailed. Pretty worthless at a distance, though.
3. Parallax is awesome, if you want to create whole surface geometry from your bitmaps and are deliberate about it. This was an accident with CrazyBump, but it serves as a good example- in the ARB shader, those "awnings" that are noticeably swollen outwards with parallax are flat- that entire section of the model is... two triangles.
Any other questions? Anybody else still not getting it? That guy is testing parallax, concluding it's worthless, and he obviously has no clue how to use it, or why it's fundamentally different than normalmaps. Parallax is what you use, when you want to auto-LOD meshes with incredible details, guys. And it's what you want to use when you want to, say, have a map's drawn geometry conform to the true-resolution heightmap, instead of an approximation.
But the key thing here is that it is different, requires a different approach to get the best use out of it, and it's really not expensive at all, assuming that it's just parallax, and not the giant bells-and-whistles of what I'm trying to do with Unit shaders.
It's just another tool, basically.
Re: SM3
http://en.wikipedia.org/wiki/Parallax_mapping
to the texcoords based on incident angle and a texture sampled
height at the original texcoords.
Are you talking silly shit again, Argh? It's just a minor adjustmentand he obviously has no clue how to use it, or why it's fundamentally different than normalmaps
to the texcoords based on incident angle and a texture sampled
height at the original texcoords.
Re: SM3
On my computer (with ubuntu, 9600gt and restricted nvidia drivers) SM3 has been reliably crashing since couple of weeks ago. I made a small patch here:
http://github.com/oliveoil/spring/commi ... bb350f89e1
Can someone with access push this upstream please?
I know about one more crash at the moment, when terraforming is done. Will try to look into that as well.
http://github.com/oliveoil/spring/commi ... bb350f89e1
Can someone with access push this upstream please?
I know about one more crash at the moment, when terraforming is done. Will try to look into that as well.
Re: SM3
anyone else getting these dark shadows?
- Attachments
-
- screen023.jpg
- (410.57 KiB) Downloaded 116 times