From a quick glance it looks like the heavy computational work is done in OpenCL kernels. Typically, such projects are either written in Python / some other flexible high level language for the easy metaprogramming or C++ if a high performance CPU fallback or tight CUDA integration is desired. I've tried out Rust in the past, but decided against using it for my own computational projects because it didn't seem to offer any significant advantages for this type of work and I didn't enjoy the extra hurdles it imposed. It'd be interesting to hear an opposing view from other computational scientists.
I'm pining for integer generics so we can get a library as versatile as C++'s Eigen for heavy numerical work in Rust. :(
It's funny how as soon as that gets added it gets more attention ;)
I'm certainly not out to disparage Rust in any way, merely looking to compare my experiences with other computational researchers. While language choice is in many ways subjective and we should all pick the ones we like, I've learned a lot in the past from discussions about the subject.
This, along with really simple testing, benchmarking and the ease of adding dependencies (through crates) made it a good choice. There are almost no downsides from using rust, provided you have experience with it!
Disclaimer: I've played with Rust a tiny little bit.