Is any of the dedicated hardware able to parallelize this process? That would decrease the computational delay, right?
Would there be any advantages to doing trilinear filtering for 2D even though it's meant for 3D, like we did with cross product?
To clarify, we're just assuming here that (u, v, w) are between 0 and 1 as an example, right? In an actual setting, are we assuming that we already have computed f at certain (integer) values nearby or are we assuming we have f at certain triangle vertices and have to use those vertices?
Is this the most efficient way to filter 3D data? Are there more specialized ways of doing thing?
What tradeoffs between basic MIPs mapping and trilinear filtering?