[MDX]Pick a screenshot and I'll write an article on the most popular one.

The third one!

The third one is rigid body physics, entities collide with themselves and any planes in the BSP they intersect. It's related to DirectX because the planes themselves contain faces which need to be tested, to find out of the collision object is above the face, or somewhere else on the plane.

I find parallax mapping the most interesting. Even though I don't have problems with the topic myself, I would like some new insights about optimizing the technique, both in performance and quality (preferably, but not necessarily, at the same time).

Have you got info on "true bump mapping" (per-pixel displacement mapping), whereby a volumetric data set is used in ray-tracing bumps onto flat geometry by using a pixel shader? I would like to discuss about how to efficiently implement displaced edges so that when the geometry cuts out, the edge would also be displaced and would kill the pixels that fall outside the simulated bumps.

I have thought about displacing the original geometry by a "maximum bump height" factor * normal, and then just cutting the extraneous pixels off by using either texkill or alpha value of 0 with alpha testing. However, this technique is not water-proof in my opinion.

For parallax mapping, I am not using real volumetric data as I am targetting ps 2.0 hardware.

I am using a blurred heightmap to retrieve max height, then sampling from texcoord - (ViewDir * TBNMatrix). It works fairly well, but I'm going to look into other ps 2.0 methods, if I find any good ones.

Edit: In the screenshots above, the texture is clamped to the edges. I clamp texcoord + (ViewDir * TBNMatrix) and add how much was clamped on the x and y axis to the new texcoord.

This way, a texture cannot "overflow" off the sides.

Excellent initiative!
I'll vote for the third screenshot as well.

2 Looks like an improvement on 1 (which looks like water through a polaroid lense), but I'd go with either of those. Admittedly, I'm not an MDX programmer so it's of little (direct) use to me - but it's a good idea [smile]

As a general comment, your bump/specular mapping looks a little "wet" (especially 5 and 7). That is, it looks a little fake/shiny/plastic. Changing some of the parameters for the specular and other more metal-like lighting models might well improve it a bit. Concept is probably sound though.

Cheers,
Jack

I vote for the last one!

Well I'm obsessed with water simulation, so I'm going to have to say the first two. Cool looking stuff. I don't really care that the app is in VB.NET cause hey, a shader's a shader right? Where are you going to post the article you write? Gamedev?

Jollyjeffers - Thanks for making my images a bit smaller, I was wondering how to do that. My specular looks shiny/plasticy? Look at the 2nd screenshot. ;)
Not so wet scince I've started clamping the specular to a "dirty" texture, based on how rough a surface is.

Water simulation - I'm actually not rendering any water. ;)
My engine sees it as a surface with a shader (the specular in screenshot 2) and scince bDistortion is set to True, a refraction map taken out of another defined shader. So this technique could work on plasma walls, dirty glass, etc without me adding anything to my core engine. I'm using a post-processing method for refraction, fog and any other deferred effects I'll think of. I'm targetting ps 2.0 so just to be safe, I'm limiting myself to 8 textures per shader pass and I'm using single-pass shaders only. My pp deferred technique only uses 3 stages currently, distortion map, distortion stencil map and a depth map from the camera for fog.

Where will I post it? On gamedev.net, most likely.

All these screenshots are from my engine at various points in time, the water specular I'm still working on. The screenshots in order of dev work are 2,1,3,5,6,7,4

Quote:

Original post by CadeF Jollyjeffers - Thanks for making my images a bit smaller, I was wondering how to do that.

S'ok - Add a width="__%" attribute to the >img src=""< tag if you want things smaller [smile]

Quote:

Original post by CadeF My specular looks shiny/plasticy? Look at the 2nd screenshot. ;) Not so wet scince I've started clamping the specular to a "dirty" texture, based on how rough a surface is.

Okay then - my bad! Didn't realise it was intentional. My current project is exploring a number of lighting models, and one of the big critiscms of Blinn-Phong (and similar) is that it leads to plasticy/wet/artificial type effects. Thus I just put 1+1 together and assumed it was accidental...

Quote:

Original post by CadeF I'm limiting myself to 8 textures per shader pass and I'm using single-pass shaders only.

May I ask why? For ps_2_0 (if memory serves me well) you should be able to sample from 16 unique samplers at most 32 times. Sure, performance tails off when you get close to that limit - but an artificial limit of 8 in only a single pass might be unnecessary...

Quote:

Original post by CadeF Where will I post it? On gamedev.net, most likely.

Good choice [grin]

Cheers,
Jack

ps 2.0, scince I'm targetting a Radeon 9600 and a Geforce FX 5200 as my min cards. 8 textures because any more get slow on those cards. Fortunately, I don't need more than 3 textures in any shaders yet.

All of it would be beneficial to me. :) I'm not a newbie with DX, but I haven't produced anything that looks like those shots.

I've improved my Parallax technique now, added a new screenshot at the top.