performance on low spec processors

[HankB]

Performance on low spec processors (Intel Atom D410 or Celeron J1900) is poor. Of course performance bottleneck depends on the balance between processor and disk speed. With these processors they tend to show one processor core saturated while wait time (reported by top) remains low, even with HDD (vs. SSD) based systems. Profiling will likely reveal that the hash calculation is probably the bottleneck. If that is the case, one of the following strategies might improve performance.

• More efficient hash calculation.
• Drop the hash calculation and perform only byte by byte comparison.
• Use multiple cores for calculating hashes, allowing these to be performed in parallel. (Note: many low spec processors such as the D410 only have one core with hyperthreading for a virtual second.)

[widget-]

FastCopy (a utility similar to an rsync GUI on Windows that takes advantage of some faster APIs than e.g. MSYS would use) uses the xxHash3 algorithm to compare file hashes. The xxHash3 Github Pages page has some performance comparisons with other hashes. It also has links to various Go implementations. You may want to consider it or a similar algorithm.

[HankB]

That looks interesting. Care to submit a PR with an implementation? :D