mleleszi wrote: I agree that in theory this seems and like the best approach, I was just suggesting that maybe some benchmarks might be useful first on your POC before committing (we've only done benchmarks with std::unordered_map with an identity hash), to check how many collisions we get with a naive implementation, what's the memory overhead of that, whats the constant time overhead of the perfect hash function compared the identity hash (at runtime, not talking about building the table) etc
https://github.com/llvm/llvm-project/pull/174798 _______________________________________________ llvm-branch-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
