rtcvb32: So they allow doubles in OpenMW? Certainly the editor isn't going to be compatible...
0Grapher: I don't know. Is
this what you meant?
I've never witnessed the janky-like effect first hand, so that's interesting.
Thinking about it, it could remain compatible by using doubles internally and storing floats to stay compatible...
Now to be sure we're on the same page, a float has 24 bits of precision, 1 sign flag, and 7 bits of the power of how much the precision is shifted. This means it can be incredibly precise when the number is less than 1, however once you go above 16 million, the next closest number starts jumping up more and more and more. I forget the exact amount, but I think 1 square cell was something like 64,000 (
and the average building blocks were 256[sup]3[/sup] ).
I forget the exact numbers, but for perspective I think 5 or so was about an inch. So to be more properly precise especially for walking, movement, placement of various objects, you needed a precision of 100 or 1000th. If we assume 1/100th, then the number of blocks you can go out is about 3 blocks for basically perfect precision (
10 if you include the center 0). After that getting 1/10th gets you out 26 blocks. In both positive and negative directions that's 52 blocks out (
2705 total blocks if you include the dead center 0).