template <class K, class V, size_t N, class Mutex>
using Map = phmap::parallel_flat_hash_map<K, V, phmap::priv::hash_default_hash<K>, phmap::priv::hash_default_eq<K>,
 phmap::priv::Allocator<phmap::priv::Pair<const K, V>>, N, Mutex>;
using OutputTable = Map<Bitset, double, 8, std::mutex>;
OutputTable test(const Table& table1, const Table& table2) {
 OutputTable table;
 #pragma omp parallel for // schedule(static, 1)
 for (size_t i = 0; i < table1._table.subcnt(); ++i) {
 const auto& set1 = table1._table.get_inner_m(i).set_;
 const auto& set2 = table2._table.get_inner_m(i).set_;
 {
 for (const auto& [index, vec1] : set1) {
 auto it = set2.find(index);
 if (it != set2.end()) {
 const auto& vec2 = it->second;
 // create a new hashmap to contain all results sharing the common pattern
 for (const auto& e1 : vec1) {
 for (const auto& e2 : vec2) {
 Bitset b(e1.first);
 b |= e2.first;
 table.lazy_emplace_l(
 b, [&](OutputTable::value_type& entry) { entry.second += e1.second * e2.second; },
 [&](const OutputTable::constructor& c) { c(std::move(b), e1.second * e2.second); });
 }
 }
 }
 }
 }
 }
 return table;
}