🔒 Locked 🎮 Unity

Voxel rendering optimizations

Started by khyperia Nov 18, 2011 at 5:06 PM 4 replies 1.5k views

Original Post

Nov 18, 2011 at 5:06 PM

Ok, so. I head about a thing called "hardware instancing" and have a few questions. (I am using XNA)
First: Is it possible to store a transformation matrix, texture coordinate, and color in the "many times" part of hardware instancing? I was thinking I could simply start with a single square, and transform that everywhere there was a visible face.
Second: Would it be faster to transform the entire cube around, acknowledging that most faces will not be visible, or just transform a face?
Third: Is it possible to combine a transformation AND rotation matrix into the same matrix, then apply the final matrix to whatever you want to transform? (I just want to store a single matrix in the hardware instancing thing, transformation and rotation)
Answer any question(s) that you want, any help is appreciated.

Adaline

Nov 18, 2011 at 6:07 PM

Hello

Third: Is it possible to combine a transformation AND rotation matrix into the same matrix, then apply the final matrix to whatever you want to transform? (I just want to store a single matrix in the hardware instancing thing, transformation and rotation)

Yes of course that's possible : you can set each step of the complete transform with a matrix and multiply them together to get a single matrix that contains the whole transform
(see matrices multiplication as transform composition)
Hope that helps

hupsilardee

Nov 19, 2011 at 9:06 PM

Are the cubes axis aligned, like Minecraft?

Ben Bowen

Nov 22, 2011 at 3:42 AM

Are the cubes axis aligned, like Minecraft?

They're voxels, buddy.

hupsilardee

Nov 22, 2011 at 8:24 PM

[quote name='hupsilardee' timestamp='1321736791' post='4885706']
Are the cubes axis aligned, like Minecraft?

They're voxels, buddy.
[/quote]

If you read the original post, you might notice the mention of rotation matrices, "buddy".

OP: If they are axis aligned, it would be faster to transform the entire cube around several times, because then you need only apply a translation which is much faster than a matrix multiply. If you are, as I suspect, thinking of rotating a single 2-tri face with combinations of 90 degree rotations to create a cube, this probably will be much slower.

Ben Bowen

Nov 23, 2011 at 4:21 AM

If you read the original post, you might notice the mention of rotation matrices, "buddy".[/quote]

So there's no point to your question? And, I did read it. Thanks for the -1

EDIT:
Oh, sorry if I sounded like a jerk. I'll keep my head down now.

Topic Locked

This topic has been locked by a moderator. New replies are not allowed.