Parallel Algorithms Library(PAL) is a modern c++(1x) library. It will be designed to take advantage of CPU cores. NOTE: it is not designed to take advantage of GPU. The main goal is to create a modern C++ parallel algorithms lib with no dependencies.

It will feature standard parallel algorithms such as reduce and parallel pipeline. It uses a work-stealing queue pattern. Visit the Wiki for more info.

• Currently features work stealing parallel_for_each, paralle_for_each_range and parallel_reduce algorithms
• Tested on Mac OS X and Linux

Building is optional, it simply builds the unit tests. Pal is currently headers only. This project uses cmake. It also uses googlemock.

1. Download googletest and copy it into the Root of Pal.
2. Run cmake cmake .. -G Xcode // for xcode cmake .. -G "Unix Makefiles" // for Linux

Here is a quick demo on how to use parallel_for_each on a vector

std::vector<int,Pal::aligned_allocator<int>> intVector(257, 1);
using Iterator = std::vector<int,Pal::aligned_allocator<int> >::iterator;

// assign 2 to each element in the vector
Pal::parallel_for_each(intVector.begin(), intVector.end(), [](int& val)
{
    val = 2;
});

Working with arrays;

Pal::aligned_allocator<int> allocator;
int* array  = allocator.allocate(256);

Pal::parallel_for_each_range(0, 256, [&array](Pal::chunk_range<int> range)
{
    for(auto it = range.begin; it != range.end; ++it)
    {
        array[it] = 2;
    }
});

allocator.deallocate(array);

parallel_reduce will take a list and reduce it to one value using a binary function. Example below features parallel sum implementation using stl::plus. The binary functor can be any binary function.

// expected value = n(n+1)/2
int n = 257;
int expected = (n*(n+1))/2;

using IntVector = std::vector<int, Pal::aligned_allocator<int>>;
IntVector v(n);
int counter = 1;
std::generate(v.begin(), v.end(), [&counter]()
{
    return counter++;
});
int value = Pal::parallel_reduce(v.begin(), v.end(), std::plus<int>());
std::cout << value << std::endl; // prints out 33153