Hi Sutts, it isn't good in general to have your models at a scale like that where the coordinate values are in the hundreds of thousands. The way floating point math works is that the precision of calculations is reduced when numeric values are far away from the origin like that.
One of the differences between 32-bit and 64-bit code is that floating point operations in 32-bit code using the x87 floating point instructions would actually have an extended 80 bit register value so some operations could be carried out at a higher precision before being stored back into memory. That has been lost in the transition to 64-bit, the 64-bit instruction set does not have that kind of extended precision.
So a side effect of that is that it is possible for cases that are difficult to calculate numerically like you have set up here that the 32-bit version could behave better than the 64-bit version.
However, what you're trying to do here is really not a great fit for NURBS in general. NURBS are not well suited for terrain, polygons are a better fit for this.
- Michael
|