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

template<typename T>
struct Slices {
   Slices(T nof_threads, T problem_size) :
	 nof_threads((assert(nof_threads > 0), nof_threads)),
	 problem_size(problem_size),
	 remainder(problem_size % nof_threads),
	 slice_size(problem_size / nof_threads) {
   }
   T offset(T index) {
      assert(index < nof_threads);
      if (index < remainder) {
	 return index * (slice_size + 1);
      } else {
	 return remainder * (slice_size + 1) +
	        (index - remainder) * slice_size;
      }
   }
   T size(T index) {
      assert(index < nof_threads);
      if (index < remainder) {
	 return slice_size + 1;
      } else {
	 return slice_size;
      }
   }
   T nof_threads; T problem_size;
   T remainder; T slice_size;
};

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

   std::random_device random;
   std::mt19937 mt(random());
   std::uniform_real_distribution<ElementType> uniform(-100,100);
   
   hpc::matvec::apply(A, [&](ElementType& val, Index i, Index j) -> void {
      val = uniform(mt);
   });
}

int main() {
   using namespace hpc::matvec;
   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([=]() mutable { randomInit(A_); });
   }
   for (int index = 0; index < nof_threads; ++index) {
      threads[index].join();
   }
   /* print a small block of each of the initialized matrices */
   print(A, "A");
}