In C++, why is this custom hash function slow?

Question

In C++, I often would like to created an unordered std::unordered_set or std::unordered_map of some type like std::tuple<int,int,int> or std::pair<int,int>. But I haven't found a way to do so that is fast. My custom hash functions are all slow. By slow I mean that programs that use them run faster if use the ordered version (set or map).

Here's an example of a custom hash function I created. I'm following examples I found on the internet. This type of example seems to be common. (An exclusive-or between primitive fields in a large data type).

using namespace std;
typedef pair<int,int> pi;
struct pair_hash
{
    size_t operator()(const pi& p) const
    {
        auto h1 = hash<int>{}(p.first);
        auto h2 = hash<int>{}(p.second);
        return h1 ^ h2;
    }
};

Here's another one:

using namespace std;
typedef tuple<int,int,int,int> t4;

struct t4_hash
{
    size_t operator()(const t4 t) const
    {
        auto h1 = hash<int>{}(get<0>(t));
        auto h2 = hash<int>{}(get<1>(t));
        auto h3 = hash<int>{}(get<2>(t));
        auto h4 = hash<int>{}(get<3>(t));
        return h1 ^ h2 ^ h3 ^ h4;
    }
};