Summary

In this chapter, we set the lower-level details of concurrency in Rust as a foundation. We discussed thread pools, which, it turns out, we had all the pieces in-hand from previous chapters, to understand a fairly sophisticated one. Then we looked into rayon and discovered that we could also understand an extremely sophisticated threadpool, hidden behind the type system to enable data parallelism in the programming model. We discussed architectural concerns with the thread-per-connection model and the challenges of splitting small datasets up into data parallel iterators. Finally, we did a walkthrough of a rayon and multi-processing-based genetics algorithm project. The std::process interface is lean compared to that exposed by some operating systems, but well-thought-out and quite useful, as demonstrated in the feruscore project that closed out the chapter. We'll pick up feruscore in the next chapter when we integrate C code into it, in lieu of calling out to a process.