johguenther
commented
1 year ago An additional workaround to what you already described could be resampling to VDB or AMR volumes, which is would be accurate and less wasteful with memory than resampling to a uniform grid. Yet I agree, the best way is that we natively support rectilinear grids in OSPRay.
I've had a few cases where I wanted to visualize volume data that's on a cartesian grid, but with non-uniform spacing per axis. I.e. the spacing is specified using an array per axis, giving the location of the coordinate planes for that axis.
In order to render in OSPRay I've so far either resampled such a volume to a uniform grid (not very attractive, due to extra preprocessing and storage requirements, plus loss of volume precision), or by setting up an unstructured hexahedral volume in OSPRay matching the spacing (which vastly grows the memory requirements). I was wondering if support for rendering such a non-uniform grid could be added, without doing either of those work-arounds?