#include <thread>
#include <random>
#include <cstdlib>
#include <list>
#include <mutex>
#include <vector>
#include <hpc/matvec/gematrix.h>
#include <hpc/matvec/apply.h>
#include <hpc/matvec/print.h>
#include <hpc/aux/slices.h>

template<typename T>
struct RandomEnginePool {
      using EngineType = T;
      T get() {
         /* check if we have a free engine in unused */
         {
            std::lock_guard<std::mutex> lock(mutex);
            if (unused.size() > 0) {
               T rg = std::move(unused.front());
               unused.pop_front();
               return rg;
            }
         }
         /* prepare new random generator */
         return T(r());
      }
      void free(T&& engine) {
         std::lock_guard<std::mutex> lock(mutex);
         unused.push_back(engine);
      }
   private:
      std::mutex mutex;
      std::random_device r;
      std::list<T> unused;
};

template<typename T>
struct RandomEngineGuard {
   using EngineType = T;
   RandomEngineGuard(RandomEnginePool<T>& pool) : pool(pool), engine(pool.get()) {
   }
   ~RandomEngineGuard() {
      pool.free(std::move(engine));
   }
   T& get() {
      return engine;
   }
   RandomEnginePool<T>& pool;
   T engine;
};

template<typename MA, typename Pool>
typename std::enable_if<hpc::matvec::IsRealGeMatrix<MA>::value, void>::type
randomInit(MA& A, Pool& pool) {
   using ElementType = typename MA::ElementType;
   using Index = typename MA::Index;
   using EngineType = typename Pool::EngineType;

   std::uniform_real_distribution<double> uniform(-100, 100);
   RandomEngineGuard<EngineType> guard(pool);
   auto& engine(guard.get());

   hpc::matvec::apply(A, [&](ElementType& val, Index i, Index j) -> void {
      val = uniform(engine);
   });
}

int main() {
   using namespace hpc::matvec;
   using namespace hpc::aux;
   RandomEnginePool<std::mt19937> pool;
   GeMatrix<double> A(51, 7);
   unsigned int nof_threads = std::thread::hardware_concurrency();

   std::vector<std::thread> threads(nof_threads);
   Slices<GeMatrix<double>::Index> slices(nof_threads, A.numRows);
   for (int index = 0; index < nof_threads; ++index) {
      auto firstRow = slices.offset(index);
      auto numRows = slices.size(index);
      auto A_ = A(firstRow, 0, numRows, A.numCols);
      threads[index] = std::thread([=,&pool]() mutable { randomInit(A_, pool); });
   }
   for (int index = 0; index < nof_threads; ++index) {
      threads[index].join();
   }
   print(A, "A");
}