Chili is a Rust port of Spice, a low-overhead parallelization library similar to rayon::join. It allows for running two closures in parallel at fork points during computation, ideal for scenarios with many small computations where estimating remaining work on a branch is costly. An example is provided for summing nodes in a binary tree in parallel efficiently. Benchmarks on AMD Ryzen 7 4800HS and Apple M1 show impressive performance improvements with increasing node count, highlighting the minimal overhead of chili. The overhead on AMD Ryzen 7 4800HS remains consistent across different thread counts in the 1K node case.
https://github.com/dragostis/chili