Duplicate detection looks solved: keep a hash set, skip what you have already seen. A benchmark suite of 4050 measurements across 480 workloads says the fastest strategy can be 94x faster than std::unordered_set, or 90,000x slower, depending on what you are deduplicating and what guarantees you need.
Dense integers are an array problem
When keys are dense, bounded 32-bit integers, a hash set wastes work: it hashes, probes buckets, and chases pointers. A bitset turns membership into one indexed bit. At one million uniform integers:
strategy
ns per insert
