Jump to content

  • Log In with Google      Sign In   
  • Create Account


Question about depth bias.

  • You cannot reply to this topic
5 replies to this topic

#1 cephalo   Members   -  Reputation: 528

Like
0Likes
Like

Posted 23 June 2014 - 09:16 AM

I recently used the depth bias field in the rasterizer state description to get rid of some z-fighting issues that I had. I drew some stuff, then changed the rasterizer state to a depth bias of -10, then drew other stuff. This eliminated any z fighting perfectly.

 

I have a question though. My next step will involving some more co-planer drawing, and I could simplify some things if I could allow these quads to overlap in places. Does the depth bias apply only to reads of the depth buffer? I'm not sure how to formulate this question... let me express what I want to do.

 

Can I, in the same draw call of many, possibly overlapping instanced quads, ensure that whatever is drawn after, or before, always wins the depth test? I just need one quad or the other to always win. It could be those drawn first or those drawn last. I'm unsure of exactly how the depth bias works.


Edited by cephalo, 23 June 2014 - 09:23 AM.


Sponsor:

#2 MJP   Moderators   -  Reputation: 10243

Like
4Likes
Like

Posted 23 June 2014 - 02:57 PM

Depth bias basically "nudges" every pixel forward or backwards in view-space, based on the bias value provided. So it doesn't just affect comparison with the depth buffer, it also affects the resulting depth value written to the depth buffer.

If you always want whatever you draw to end up on top, why not just disable depth testing? Or am I misunderstanding what you're attempting to accomplish?



#3 cephalo   Members   -  Reputation: 528

Like
0Likes
Like

Posted 24 June 2014 - 09:43 AM

Well, I only want things to end up on top if they are co-planar or close enough to co-planar to cause z fighting. Thanks for the answer. That tells me what I need to do. If one wants to benefit from the depth bias setting, one must change the rasterizer state.


Edited by cephalo, 24 June 2014 - 09:46 AM.


#4 C0lumbo   Crossbones+   -  Reputation: 2129

Like
1Likes
Like

Posted 24 June 2014 - 12:40 PM

I usually end up organising my rendering code so that it happens like this:

 

Render Opaque Stuff (environments, characters, etc)

Turn z-write off, z-bias on

Render depth-biased decal stuff (blob shadows/bullet holes, etc)

Turn z-write back on, z-bias off

Render 1-bit transparent stuff (foliage, etc)

Turn z-write off

Render 8-bit transparent stuff (particles, etc)

 

By turning off the z-write while you render your decal stuff, you can stack lots of decals without having to increment the depth bias between each layer.

 

Edit: Also, I almost always z-bias by creating a new projection matrix that has a minutely narrower field of view. I find it tends to succeed better across different platforms.


Edited by C0lumbo, 24 June 2014 - 12:42 PM.


#5 cephalo   Members   -  Reputation: 528

Like
0Likes
Like

Posted 25 June 2014 - 08:08 AM

Good info C0lumbo, thanks.



#6 Hodgman   Moderators   -  Reputation: 27837

Like
2Likes
Like

Posted 25 June 2014 - 09:47 AM

If your co-planar stuff uses the exact same vertices (and the same vertex shader), then there's no need for depth biasing -- just use the depth-test mode of less-than-or-equal and the latest pass will always win, or use equal for every pass after the first one. In this situation, there will never be any z-fighting.

 

If your decal's vertices don't line up perfectly, then yeah, you'll get z-fighting, which you can fix by z-biasing the decal towards the viewer (which modifies the decal's z values that go into the depth-test).

Alternatively you can depth-bias the entire first pass of the scene in the opposite direction, but that's a much less common choice wink.png

 

I prefer the first option because the right depth-bias value to use depends on the distance of the object from the viewer. The z-buffer units aren't evenly distributed (there's more precision near the camera), so a value of "-10" might mean -0.1mm when close to the camera and -1m when far from the camera wacko.png







PARTNERS