Hi,

I'm trying to cast data stored in voxels into 64-bit doubles for simulation precision purposes. Is this viable? I couldn't seem to find an "attribute cast" sop for volumes. @jlait

Thanks,
Peter

Hi Peter,

I doubt regular volumes support 64bit precision, but VDB volumes do. So you can use that. But if you are using DOPs, I don't think you can retain 64bit precision without doing a trip back to 32bit volumetric representation.

Hi Yunus,

Thank you very much for the reply! I'll give that a try. The algorithm I'm trying to implement has a P2G/G2P transfer scheme for each timestep. 32-bit floats are seemingly prone to fp cancellation during this process.

Thanks,
Peter

Could probably do these calculations on points with a 64bit attribute and then transfer the result back to a regular volume at the end. That's probably what I would do.