In the impl with LOCKS it will be similar i guess to spsc? Wonder if any obvious optimisations can be made since lots of contention could be present due to loads of Ps/Cs

#include <mutex>
#include <condition_variable>
#include <optional>
#include <vector>

template<typename T>
class LockedQueue {
    std::vector<T> buf;
    size_t head = 0, tail = 0, sz = 0;
    std::mutex mtx;
    std::condition_variable cv;

public:
    explicit LockedQueue(size_t cap) : buf(cap) {}

    bool push(T val) {
        std::unique_lock lock(mtx);
        if (sz == buf.size()) return false;
        buf[tail] = std::move(val);
        tail = (tail + 1) % buf.size();
        ++sz;
        cv.notify_one();
        return true;
    }

    std::optional<T> pop() {
        std::unique_lock lock(mtx);
        cv.wait(lock, [this] { return sz > 0; });
        T val = std::move(buf[head]);
        head = (head + 1) % buf.size();
        --sz;
        return val;
    }
};