[Solved] NewtonCreateHeightFieldCollision: unsigned short?

[Dalon]

Hi,

i tried NewtonCreateHeightFieldCollision and I really like it

But I don't like the use of unsigned short* because I loose some details =/ For example... I work with etm (2.3) and it saves the height-values using float (0.0 for the lowest value and 1.0 for the highest) and simply has a scale factor for the height.

Is there any way to keep the terrain details?

[Aphex]

65 thousand discrete levels is probably enough for most landscapes... does yours cover a large height range?
Can you not just scale your float value by (2^16)-1 then add an offset?

[Dalon]

Possible heights are 0 - 1875 Ogreunits... But the details are a bit high... well just a difference of 0.1 units will be visible...

[JernejL]

If you need more precision, you can stack several heightmaps on top of each other.

[Dalon]

Wouldn't that be a bit too... ehm complicated?

Edit: Solved it... I now just multiplicate the float values by 8 and set the y-scale value to 0.125 instead of 1.0

[Leadwerks]

When working with a large set of data like this, you would each point to be stored in the fewest number of bytes possible. 8 bytes doesn't give you enough fidelity, but 2 bytes allows about 65,000 possible values. Mulitply your float values (assuming they are 0-1) by about 65,000 (8192*2*2*2, I forget the exact number) and you will have a short value.

I'm using terrain extensively, and the way Newton has it implemented is very smart.