多线程成员函数C ++

Question

So I am trying to multithread the dot product computation between two matrix objects in a "library" I'm writing. here is the code of interest

double mat_cont::dot_t( mat_cont & other, const int offset,  \
const int upper_lim)

{
    double local_sum = 0;   
    for (int i = offset; i < upper_lim+offset; ++i)
        local_sum+=(this->mat[i] + other.mat[i]);

    return local_sum;
}


double mat_cont::dot( mat_cont & other){

    future<double> threads[3];
    int part = (int)(dim * dim) / (int)4;

    double sum = 0;

    for (int i = 1; i < 4; ++i){
        threads[i-1] = 
            async (std::launch::async,&mat_cont::dot_t, 
                  other, part*i, (i+1)*part);
    }
    for(int i = 0; i < part; ++i)
        sum+=(this->mat[i] + other.mat[i]);

    for(int i = 1; i < 3; ++i)
        sum+=threads[i].get();
    return sum;
    }

and this compilation error is thrown

error: no matching function for call to 'async'
             threads[i-1] = async (std::launch::async,&mat_cont::dot_t, other, part*i, (i+1)*part);
                            ^~~~~ 
/Applications/Xcode.app/Contents/Developer/
Toolchains/XcodeDefault.xctoolchain/usr/include/c++/v1/future:2337:1: 
note: candidate template ignored: substitution failure [with _Fp = double (mat_cont::*)(mat_cont &, int, int), _Args =
  <mat_cont &, int, int>]: no type named 'type' in 'std::__1::__invoke_of<double (mat_cont::*)(mat_cont &, int, int), mat_cont, int, int>'async(launch __policy, _Fp&& __f, _Args&&... __args)

I'm wondering if I can multithread this part, or if I need to pass both of these objects into a friend function for multithreading. I've been trying to debug this for 2 hours now, any tips?

Answer 1

dot_t是dot_t的成员函数，它mat_cont 3个参数，因此async函数的调用中缺少一个元素- this指针应作为调用的第三个参数传递

threads[i-1] = async (std::launch::async,&mat_cont::dot_t, this, std::ref(other), part*i, (i+1)*part);

Answer 2

mat_cont::dot_t is a non-static member function, so it needs a this object to work on.

The easiest way to deal with that is to use a lambda that captures the this pointer. Then you can call dot_t exactly like you would in any other member function:

threads[i-1] = std::async(std::launch::async,
                          [this, i, part, &other]() {
                              return dot_t(other, part*i, (i+1)*part);
                          });

Live Example