Relief maps with self-shadowing

Edit Nay 25: Should now work on more cards, particularly X800 Package also included again.
Please let me know, if it doesn't run on your Shader Model 2 compatible graphics card

Webplayer (1.1 MB):
http://proglet.com/software/Relief_Pro/Relief_Pro_0_0_5.html

Note: You can play with some settings by hitting <TAB> in the webplayer.

The package is included below.
Edit: ONLY WORKS ON PRO
For Indie users, I previously posted a simpler version without shadows: http://forum.unity3d.com/viewtopic.php?t=10927

To use:
- Create a userdefined layer (I have used layer 8 and called it ReliefObjects)
- Create an object with a renderer attached
- Put the object on the ReliefObjects layer
- set the material to use the Relief/Pro/Spec shader
- Attach the "Relief" script to the object
- Set the "Shadow light" variable to point to an object. The nearest light would be a good one.

The shader will look weird in the editor view or the game view, if they are both open, and you may have to maximize the game view (to hide the editor view) for the effect to work properly.

The self-shadowing only works with one light. The calculated shadow is applied to all pixel lighting.

 

That looks really amazing. Alas, the webplayer doesn't work on my X800. Instead, I get black lines and noise artifacts, and flickering rectangles. None of the controls seem to have any effect. The indie version worked without issues, and while it's still amazing, it's not quite as amazing as this version looks.

--Eric

 

I only see a pink background on my Macbook :/

Needs more fallback shaders?

 

MBP - all great.
The floor looks painful to walk on... I kinda cringed as I walked around.

 

Damn. I had similar problems earlier on myself because of a race condition. Trouble is, if the GPU is deeply pipelined, the result of the first trace passes are not written to texture memory before the next pass tries to read the intermediate results. I solved it on my machine by actually having separate camera render calls for each phase.

Or it might just be an X800 bug

Have you tried switching all options off, so you just have the most simple trace?

Actually it should fallback to bumped specular. That works fine under emulation.
I should probably do some more explicit checking of the intermediate shaders and switch the main shader, if one of them fails.

 

For those who had problems running this demo, please try again. I probably over-optimized a bit too much the first time around.
A number of fallback shaders more in the style of the indie version are now included.

If the demo doesn't run on your machine, please try checking the console log, and see if it says anything.

Note that fragment shaders are necessary for the effect to work, but it should at least fallback nicely.

 

Awesome! Works fine on my iMac w/ X1600.

 

Looking real funky on my mbp running a radeon x1600 and OS X 10.5.2. This is what I see:

http://www.getdropbox.com/gallery/18971/0/?h=61e159

 

Yup, just tried it here on a MBP running 10.4.11 and get the same result.

 

No problems here. MBP Peryn Generation, OSX 10.5.2

Looks very nice, thank you for sharing

This kind of things is what makes me consider Unity Pro even more than I already do ... dooohhh :?

 

Didn't they ship a new graphics card with that one?

 

NVIDIA 8600M GT 256MB here

Given you had 256MB before they would be identical.

 

Hmm. On paper, your card should be capable.
I can' teally figure out version numbers on graphics cards, but I believe that x1600 is newer than R9000, which works under emulation.

Do you see any error messages in the log?

I think this type of problem might have to do with timing, pipelining and stuff like that. I am going to start a thread in ShaderLab about it.

 

That is just plain cool. Nice work. My crappy work computer has an ancient vid card with 64 MB of VRAM but it can still run this at the "fast" setting. And it looks great! (Sure I'm only pulling 3 fps, but at least it runs!)

Please keep working on this idea. This is major. Thanks for sharing.

 

This is from my console (Intel Core Duo 2 MBP with ATI X1600 graphics):

Code (csharp):

1. np_entry: NP_Initialize called
2. np_entry: NP_GetEntryPoints called
3. 080522 09:30:48 Datafile [url]http://proglet.com/software/Relief_Pro/Relief_Pro.unity3d[/url]
4. 080522 09:30:48 Loading Unity Player
5. 080522 09:30:48 loader: unity function UnityBundleStart load failed
6. 080522 09:30:48 loader: unity function UnityBundleSetupScreen2 load failed
7. Player: init engine
8. 080522 09:30:51 Loader: Start Success
9. host: proglet_2ecom Pref folder: /Users/ImAUser/Library/Preferences

 

... and I'll add that I don't have this problem with my 17" iMac that uses the same graphics chip. I've posted a bug report on the off chance that it might be a Unity issue rather than an Apple drivers issue.

 

Ok. I thought I might get away with it, if I never looked at neighbouring pixels. (Yeah the depth-blur shader does do that after all).

To anybody else listening: I am going to try a ping-pong technique with two rendertextures that Aras suggested in ShaderLab. I'll probably get the time for it within a day or two.

 

Just thought I'd chime in and say that this looks quite cool indeed. Nice work!

 

Doesn't work here : I tried everything to escape from this welcoming place, but with no success :evil:

Seriously, it works perfectly (iMac alu 24', Radeon HD2600). Thanks a lot for sharing your amazing work !

 

Sah-weet! Awesome job.

 

Too bad i cannot test this stunning shader (still missin UnityPro)...
I can a imagine a thousand cool things to do with this code-beauty.

Really great work!!!

I am looking forward to utilize this as soon as possible

Cheers Frank

 

43 fps on my x1600 MBP. Very cool!

 

Huh, yet on my x1600 MBP it looks like the attached screenshot. I'm running 10.4.11 and double checked software update to make sure I'm current.

 

No errors.

 

Ok, BigKahuna and AngryAnt, I think I know what is happening on your machines. Please try this one:

Edit: Link removed. Go to the first post in the thread for an up-to-date version

I haven't updated the package yet. First I want to see, if this works.

 

Whatever you did, it fixed it! Looks cool and I'm getting roughly 50 fps on my MBP.

 

Goody. It was Aras, who told me, that I was using some reserved variables in a wrong way.

Are the 50 fps fullscreen? Because in the webplayer window, the rate is usually capped at 50 fps.

 

Oops, yeah you're right but full screen isn't getting me faster frame rates. "Fast" setting gets 20-23 fps full screen and roughly 50 fps windowed. "Medium" gets roughly 18 fps full screen and 40-43 fps windowed.

 

Windows XP - SeaMonkey Browser - NVidia GeForce 8600GT
Same result with both the html versions:

 

Thanks for testing. I am currently working with Eric5h5 to fix some more problems, and so far we are progressing, so no need for more tests right now. A new version will be coming.

 

Fix version working. The rendering is a bit fuzzy around the edges of the shadows through.

 

I think at least some of that is because the heightmap was made by an amateur texture artist. On the walls for example, the edges of the stones are actually as jagged as the shadow looks.

 

Ok. Latest version ready.
The link is in the first post in the thread.

It contains

- Fixes for ATI X800
Thanks to Eric5h5 for helping me debug on the X800
- Better precision at sharp angles
- 5 "debug spheres" that illustrate the intermediate steps of the shader in the texture buffer.

Yes, you have to look closely to see the difference between the leftmost three.

Package also included again.

Let me know if your gfx card still has problems. Or if problems have returned.

 

Seems to work equally well on both my 17" iMac x1600 and 15" MBP x1600, but you removed the relief map from the floor, right?

 

Works worse than the initial one at least unless I missed the "line per line" shadowing on the rounded parts where the shadow is meant to fall.

 

@Bigkahuna: Yep, I turned if off on the floor. It _did_ look painful to walk on.

The spheres were not in the original. And I can see what you mean. It is a common problem with shadowing, that calculations become badly defined when the direction to the light is at a 90 degree angle to the surface normal. You see similar types of problems with the built-in shadows.

Something for me to think about.

 

Looking very nice now. Great piece of work.

 

Ok, I have a new way of doing shadows without the "marching line"

Do you like this one

http://proglet.com/software/Relief_Pro/Relief_Pro_0_0_6.html

better than the old one

http://proglet.com/software/Relief_Pro/Relief_Pro_0_0_5.html

?

 

0.0.6 looks better. The effect displays more clearly.

Really good work!

Cheers

Frank

 

Try using Firefox or Safari. I don't think the browser would really make a difference, but it could be that they run the plug-in differently.

 

Same here.



Nvidia 8600 GTS Dual SLI

 

For both as I move the camera, the relief warbles.



Vista 64-bit, NVIDIA GeForce 8600 GTS DUAL SLI

 

Well, based on this screenshot, it looks as if it is hardly running at all. As if perhaps rendertextures were disabled. Do you know if you have rendertextures working in e.g. projects that use reflective/refractive water?

 

I'm running on a windows box. Send me links to any players you want to see screenshots for.

It's possible my graphics card is just lacking a minor capability. Since I can run Crysis decently.

 

In the island demo
http://unity3d.com/gallery/live-demos/tropical-paradise

the water should look something like this (with reflections)

And I think your graphics card is plenty powerful. But I think there was some problem with some Nvidia drivers leaking memory (and therefore getting rendertextures disabled). Just wanted to check if that was it. If not, probably a bug in my code.

 

The terrain reflects in the water beautifully.

 

Well, I guess I have a bug then.
I don't expect to have time to look at it anytime soon, but if I do fix it, the fix will be posted to this thread.

 

You know it could be an issue with your shader pipeline. Maybe you are referencing a parameter in the fragment program that wasn't initialized or passed through the vertex program.

I've noticed when you do that, you'll see weird camera effects.