使用 PPL 查找数组中的最大元素

Question

我需要实现一个函数，使用 ppl.h.

找到浮点数组中的最大元素

我有这个代码，基于 this answer:

float find_largest_element_in_matrix_PPL(float* m, size_t dims)
{
    float max_element;
    int row, col;
    concurrency::combinable<float> locals([] { return INT_MIN + 0.f; });
    concurrency::parallel_for_each(size_t(0), dims * dims, [&locals](int curr)
    {
        float & localMax = locals.local();
        localMax = max<float>(localMax, curr);
    });

    max_element = locals.combine([](float left, float right) { return max<float>(left, right); });
    cout << max_element << endl;
    return max_element;
}

但是，此代码存在问题：

执行前抛出如下异常：

Error C2780 'void Concurrency::_Parallel_for_each_impl(const _Random_iterator &,const _Random_iterator &,const _Function &,_Partitioner &&,std::random_access_iterator_tag)': expects 5 arguments - 4 provided parp D:\Microsoft Visual Studio 14.0\VC\include\ppl.h 2987

Error C2780 'void Concurrency::_Parallel_for_each_impl(_Forward_iterator,const _Forward_iterator &,const _Function &,const Concurrency::auto_partitioner &,std::forward_iterator_tag)': expects 5 arguments - 4 provided parp D:\Microsoft Visual Studio 14.0\VC\include\ppl.h 2987

Error C2893 Failed to specialize function template 'iterator_traits<_Iter>::iterator_category std::_Iter_cat(const _Iter &)' parp D:\Microsoft Visual Studio 14.0\VC\include\ppl.h 2987

你能帮我解决这个问题吗？
如何重写代码以利用 parallel_for？（我无法在 parallel_for 块中引用传递给函数的数组参数）

Answer 1

float SimpleTest::SimpleTestfind_largest_element_in_matrix_PPL(float* m, size_t dims)
{
    float max_element;
    concurrency::combinable<float> locals([&]{ return INT_MIN + 0.f; });
    int last= dims*dims;
    concurrency::parallel_for(0, last, [&](int curr)
    {
        float & localMax = locals.local();
        localMax = max<float>(localMax, curr);
    });

    max_element = locals.combine([](float left, float right) { return max<float>(left, right); });
    std::cout << max_element << endl;
    return max_element;
}

也许会奏效

使用 PPL 查找数组中的最大元素

Find max element in array using PPL

c++

parallel-processing

ppl