Comparison of Open3d 0.13 and 0.17

### Checklist

- [X] I have searched for [similar issues](https://github.com/isl-org/Open3D/issues).
- [X] For Python issues, I have tested with the [latest development wheel](http://www.open3d.org/docs/latest/getting_started.html#development-version-pip).
- [X] I have checked the [release documentation](http://www.open3d.org/docs/release/) and the [latest documentation](http://www.open3d.org/docs/latest/) (for `master` branch).

### My Question

I have been exploring and comparing the performance of various versions of your library. I noticed substantial differences in runtime when executing the same program (https://github.com/Llavrih/IHS_AMR/blob/main/src/multi_camera.py) across different versions. Notably, I observed that version 0.13 was significantly faster compared to 0.17. Could you shed some light on the reasons behind this performance enhancement? Are there any specific optimizations or changes in version 0.13 that could have led to such an improvement in comparison to 0.17?

Additionally, I tried the 0.17 functions that uses CUDA cores, and I achieved even better results. Given my positive experience with CUDA, I am interested in further optimizing my code (https://github.com/Llavrih/IHS_AMR/blob/main/src/cuda_test.py) to leverage this feature more effectively. Could you please provide some guidance or recommend
ations on which parts of the code I can focus on to maximize the utilization of CUDA?

The graph presents a comparison of the programs linked above. It illustrates the frequency for each iteration within a single program cycle.