Here at Heiko we are working a lot with hashes in Rust. We have come up with a way to aggregate hashes together, treating them similar to integers and adding them together. Any overflowing bits are discarded.
For that we have created a function called aggregate_hashes which takes an array of hashes and adds them to each other one by one using the append_hash function. This one uses 64 bit addition of parts of the hashes and tracks any carry bits to be added to higher parts.
We have a suspicion that performance of this function could be improved if it added hashes together slightly differently. Specifically, if a single part from all hashes is added together, tracking any carry bits, and then the algorithm moves on to the higher parts.
We want to investigate if there is indeed any merit to these suspicions. There is already a sample benchmark for the aggregate_hashes function available. We would like you to create a new function in the aggregator that sums hashes in the proposed manner and create benchmarks to compare its performance to the old version. Feel free to modify or add any code necessary, but try to keep the current version of the aggregate_hashes function unchanged. Any additional suggestions for where performance could be improved is welcome, as well as any additional relevant tests.
Build project:
cargo build
Run tests:
cargo test
Run benchmarks:
cargo bench