[cfe-dev] [libcxx] Extending the behavior of std::hash in libc++

Justin Lebar via cfe-dev cfe-dev at lists.llvm.org
Tue Dec 8 17:07:43 PST 2015


> The thing I would like is a more explicit discussion of hash function families, and the ability to select from a family.

How would this be different from the hashtable / bloom filter hashing
in a random number into the beginning or end of the bitstream?  Is the
idea that we could be more efficient than this, or that we could get
better hashes if we had explicit support for seeding the hash?

On Tue, Dec 8, 2015 at 3:45 PM, Geoff Pike <gpike at google.com> wrote:
> The thing I would like (also... a pony...) is a more explicit
> discussion of hash function families, and the ability to select from a
> family. There are relatively important hash-based data structures that
> require 2+ hash functions, notably Bloom filters and Cuckoo hash
> tables. Given the big push to revamp std::hash, this seems like a
> great opportunity to address the needs of library authors who want to
> be able to select randomly from a family of hash functions. Ideally
> selecting new hash functions could happen "at will," e.g., at table
> rehash in Cuckoo.
>
> Apologies if this is already covered by your proposed changes to the
> C++ standard.
>
> thanks,
>
> Geoff


More information about the cfe-dev mailing list