Open StasJ opened 3 years ago
Note this issue may be particularly relevant for unstructured grids. As a specific use case it takes about 2 minutes for the first rendering of a ~150M tetrahedra cell mesh on a 18-core 2.3-GHz Intel Xeon 6140 (Skylake) processor. But subsequent renderings are on the order of a second. Presumably the initial cost is for computing acceleration structures that may only need to change when a new mesh is loaded?
Understood, this is a very reasonable request, we like to have that as well (already on the wish list) :-). We started on the infrastructure to facilitate such more fine-granular, smarter updates. However, still some pieces missing, so will take some time.
When committing a volume, OSPRay does some calculations which can be quite expensive. I am assuming these are related to acceleration structures pertaining to the mesh. We have a common use case of changing the data values of a mesh without changing the mesh itself. Committing the volume after making this change results in rerunning the expensive calculations. Is there any way this can be avoided?