Maybe I misunderstood the speed complaints about HashMap in Rust. I thought it was the hash function that was the slow bit? What is the anticipated improvements from using SwissTable?
We do choose a hash function not designed for speed by default, but that doesn’t mean that the implementation of the table can’t be improved. This is effectively the third re-write of it.
I think the bottleneck depends on the size of your table and the size of your keys. A large table with small keys will bottleneck on memory IO, but a small table with large keys will bottleneck on hashing the keys. But I definitely haven't benchmarked any of this.