Passing shared_ptr as parameter in async call

I am converting my code to multi thread to performance enhancement.

I have vector of shared_ptr and an object of another class and I am passing a shared_ptr from vector and an object as parameter to function. I am calling it using std::async but it is giving me following error :

line from where I am making async call : required from here 
/usr/include/c++/4.8.2/functional1697.61: error: no type named 'type' 
in 'class std::result_of<void (*(std::shared_ptr<A>, B))
(const  std::shared_ptr<A>&, B&)>'typedef typename 
result_of<_Callable(_Args...)>::type result_type;

Here is code snippet :

void foo(std::vector<std::shared_ptr<A>>& a, B b){
    std::vector<std::future<void>> tasks;
    for(auto& sptr : a ){
        tasks.push_back(std::async(std::launch::async, foo1, a, b))
    }

 void foo1(const std::shared_ptr<A>& a, B& b ){
    //do some stuff
 }

Can you please help me. Thanks

Solution

I am converting my code to multi thread to performance enhancement.

Here we go ... I predict difficulties.

The error is telling you that the result of calling foo1 with the arguments that std::async will pass to it is not defined, i.e. you can't call the function with those arguments.

The reason is that the function foo1 takes a parameter of type B& but std::async copies its arguments and forwards the copies to the target function, so it is going to copy b and then call foo1 with that copy forwarded as an rvalue, which cannot be bound to an lvalue reference of type B&.

If you really want to pass b by reference then you need to wrap it:

std::async(std::launch::async, foo1, a, std::ref(b))

However you should be careful, it looks like every thread is going to have a non-const reference to the same B object, implying they might be modifying that object concurrently, which will lead to data races (and undefined behaviour) unless B is already thread-safe or you modify the function foo1 to synchronize access to B.

Just sprinkling multithreaded pixie dust on your code is not going to make it faster if the code is not safe to use in multiple threads.