Function with_instance_map

include/pybind11/detail/internals.h:991–1012

template <typename F>
inline auto with_instance_map(const void *ptr, const F &cb)
    -> decltype(cb(std::declval<instance_map &>())) {
    auto &internals = get_internals();

#ifdef Py_GIL_DISABLED
    // Hash address to compute shard, but ignore low bits. We'd like allocations
    // from the same thread/core to map to the same shard and allocations from
    // other threads/cores to map to other shards. Using the high bits is a good
    // heuristic because memory allocators often have a per-thread
    // arena/superblock/segment from which smaller allocations are served.
    auto addr = reinterpret_cast<std::uintptr_t>(ptr);
    auto hash = mix64(static_cast<std::uint64_t>(addr >> 20));
    auto idx = static_cast<size_t>(hash & internals.instance_shards_mask);

    auto &shard = internals.instance_shards[idx];
    std::unique_lock<pymutex> lock(shard.mutex);
    return cb(shard.registered_instances);
#else
    (void) ptr;
    return cb(internals.registered_instances);
#endif
}

// Returns the number of registered instances for testing purposes.  The result may not be
// consistent if other threads are registering or unregistering instances concurrently.