Yu-Hang "Maxin" Tang

Thanks for the prompt reply! I just tried Xodo and it works quite well. Since I don't really have any emotional attachments (:laughing:) to Adobe Reader, I guess the problem...

## Review checklist for @yhtang ### Conflict of interest - [x] I confirm that I have read the [JOSS conflict of interest (COI) policy](https://github.com/openjournals/joss/blob/master/COI.md) and that: I have no COIs...

I am a bit concerned regarding whether this submission meets the **Substantial scholarly effort** criterion. My concerns are based on the following facts: 1. Lines of code. The main package...

Regarding the documentation and the paper, I think it will really help if a definition can be given for geodesic transform, in both layman's terms and math formulas. Otherwise, I...

@masadcv Could you please also point me to the specific example where the 15x speedup over CPU and 60x speedup over GPU claim in the statement of need can be...

The CUDA kernel appears to be straightforward stencil code without much in-depth optimization. They are memory-bound because each CUDA thread repeatedly loads the pixel data for distance computation without sharing...

I believe there are more recent works that handle the dependencies in a more sophisticated way. For example, see this [GTC talk](https://on-demand.gputechconf.com/gtc/2013/presentations/S3090-GPUGeoS-Geodesic-Distance-Transforms-for-Editing.pdf).

> this method is protected by multiple patents Thanks for sharing this knowledge. This is sad and ridiculous. But since the NVIDIA guy already implemented and presented it, I suppose...

A possible approach to implement it would be to introduce a special value for the `indices` AST primitive, and then rewrite it using lhs&rhs information in the `__call__` of the...

Per discussion with @campsd, it might be better to collect more usage scenarios before deciding on to what extent should this feature be supported.