SipHash is considerably slower than an unkeyed hash function, isn’t it? One idea I had to avoid its overhead is to use double hashing, with a fast primary hash

SipHash is considerably slower than an unkeyed hash function, isn’t it? One idea I had to avoid its overhead is to use double hashing, with a fast primary hash and SipHash as the secondary hash. Naïvely this feels as though it should provide good performance in the common case and also good security - even if there are many primary hash collisions, they won’t also be secondary hash collisions. Does anyone know if there is an analysis of this approach anywhere?