Comment by danybittel
5 days ago
If you say 1 pass bucket sorting.. I assume you do sort the buckets as well?
I've implemented a radix sort on GPU to sort the splats (every frame).. and I'm not quite happy with performance yet. A radix sort (+ prefix scan) is quite involved with lot's of dedicated hierarchical compute shaders.. I might have to get back to tune it.
I might switch to float16s as well, I'm a bit hesitant, as 1 million+ splats, may exceed the precision of halfs.
We are purposefully trading off some sorting precision for speed with float16, and for scenes with large Z extents you'd probably get more Z-fighting, so I'm not sure if I'd recommend it for you if your goal is max reconstruction accuracy! But we'll likely add a 2-pass sort (i.e. radix sort with a large base / #buckets) in the future for higher precision (user selectable so you can decide what's more important for you). But I will say that implementing a sort on the CPU is much simpler than on the GPU, so it opens up possibilities if you're willing to do a readback from GPU to CPU and tolerate at least 1 frame of latency (usually not perceivable).
You might want to consider using words (16 bit integer) instead of halfs? Then you can use all the 65k value precision in a range you choose (by remapping 32bit floats to words), potentially adjust it every frame, or with a delay.
Yeah you're right, using float16 gets us 0x7C00 buckets of resolution only. We could explicitly turn it into a log encoding and spread it over 2^16 buckets and get 2x the range there! Other renderers do this dynamic per-frame range adjustment, we could do that too.