what is double precision useful for?

[Throckmorton]

32 bit is single because it could be written as one word on a 32-bit processor.

Double performance doesn't impact games because the number format is specified in the code and pretty much all games choose single precision floats because of the performance difference. I think this is a good idea because the difference in quality of the calculations is very minimal but single precision allows the programmers to add a greater quantity of effects.

Also keep in mind that single has 24 bits of precision while monitors have 8 bits per channel or sometimes 10 bits. So the "errors" as you call them, or the rounding, still rarely impacts the final result.

I thought this was about position of 3D elements, rather than color

 

[Wall Street]

I thought this was about position of 3D elements, rather than color

Shaders are usually concerned with computing the color of pixels. The rasterizer and z-buffers calculate the positions of the 3D objects but are not what is ring discussed here.

 

[BFG10K]

except only the voodoo1/2 cards had 16 bit w-buffer support as well and that is what unreal 1's glide renderer used.

Yes, I'd forgotten about the extended z buffer format.

they also used power VR's proprietary API iirc.

They used a lot of "metal APIs" but none had any visible depth advantages compared to running 32 bit Direct3D/OpenGL when it was later patched into the game.

and matrix path of neo and serious sam 2.

I take this is your way of retracting your statement "DX doesn't support wide open areas"?

games no longer use 16 bit formats for a reason, so what makes you think 24 bit formats will be used forever? just wondering

You don't need DP to represent >24bit Z buffers.

except if someone wants to replicate a 32 bit int z-buffer in software/compute, then 32 bit float precision wont be enough.

You appear to be confusing terms. Also the Voodoo's "FP16" depth doesn't need DP to represent it.

 

[Exophase]

games no longer use 16 bit formats for a reason, so what makes you think 24 bit formats will be used forever? just wondering

Did not say 24-bit, merely said there isn't a demand for 32-bit integer depth. In other words, 32-bit float depth is sufficient.

 

[Anarchist420]

You don't need DP to represent >24bit Z buffers.

You appear to be confusing terms. Also the Voodoo's "FP16" depth doesn't need DP to represent it.

I never said that the Voodoo's FP16 depth needed DP to represent it (whatever that means). But I dont see how one can replicate (i.e., simulate with 100% accuracy) a 32 bit fixed point z-buffer with 32 bit float precision.

I take this is your way of retracting your statement "DX doesn't support wide open areas"?

maybe, but 32 bit integer z-buffers are the best w-buffer substitute.

 

[cbn]

EDIT: I posted to the wrong thread.

 

[William Gaatjes]

If you like an in depth article about floating point values en IEEE754 format :

http://www.mathpath.org/concepts/Num/frac.htm
http://www.oxfordmathcenter.com/drupal7/node/43

What everyone should know about using floating point arithmetic
http://docs.oracle.com/cd/E19957-01/806-3568/ncg_goldberg.html