Return algorithm diagrams to CUB docs
- Dominant language
- C++
- Stars
- 2.5k
- Forks
- 486
- Avg merge
- 2d 6h
- Merged PRs (30d)
- 295
Description
We used to have algorithm diagrams in CUB docs. For instance, reduction used to have the following picture:

We haven't carried these visualizations into the new sphinx-based docs but users found these very useful. There's an issue with adopting these images as is. First of all, they are at very low resolution. Apart from that, I'd like to avoid storing binary blobs in version control system. These concerns can be addressed if we change the approach to drawing these diagrams and use something like [sphinxcontrib-tikz](https://pypi.org/project/sphinxcontrib-tikz). That'd let us store images as text inline. This would further help when we adopt a different docs theme that had dark mode.
### Tasks
- [ ] Integrate sphinxcontrib-tikz into CCCL docs
- [ ] Implement tikz version of [block loading](https://github.com/NVlabs/cub/blob/gh-pages/docs/images/block_load_logo.png) logo
- [ ] Implement tikz version of [block reduction](https://github.com/NVlabs/cub/blob/gh-pages/docs/images/block_reduce.png) logo
- [ ] Implement tikz version of block scan raking logo
- [ ] Implement tikz version of block scan warp scans logo
- [ ] Implement tikz version of store logo
- [ ] Implement tikz version of blocked arrangement diagram
- [ ] Implement tikz version of CNP abstraction diagram
- [ ] Implement tikz version of CUB overview diagram
- [ ] Implement tikz version of generic abstraction diagram
- [ ] Implement tikz version of histogram logo
- [ ] Implement tikz version of partition logo
- [ ] Implement tikz version of reduce logo
- [ ] Implement tikz version of run length encode logo
- [ ] Implement tikz version of scan logo
- [ ] Implement tikz version of segmented sorting logo
- [ ] Implement tikz version of select flags logo
- [ ] Implement tikz version of select logo
- [ ] Implement tikz version of striped arrangement diagram
- [ ] Implement tikz version of unique logo
- [ ] Implement tikz version of transpose logo
Legacy renders can be found here: https://github.com/NVlabs/cub/tree/gh-pages/docs/images.
Contributor guide
Assessment
This issue has not been assessed yet.