RandomTensor.hpp

Go to the documentation of this file.

#ifndef RANDOM_TENSOR_DEFINED
#define RANDOM_TENSOR_DEFINED
 
#include <math/Complex.hpp>
#include <Sisi4s.hpp>
#include <util/Tensor.hpp>
#include <complex>
#include <random>
 
namespace sisi4s {
template <typename Distribution, typename RandomEngine>
inline void setRandom(double &value,
                      Distribution &distribution,
                      RandomEngine &randomEngine) {
  // #ifdef INTEL_COMPILER
  //     value = distribution(randomEngine);
  //     value = -1.0 + 2.0*rand() / RAND_MAX; // distribution(randomEngine);
  // #else
  value = distribution(randomEngine);
  // #endif
}
 
template <typename Distribution, typename RandomEngine>
inline void setRandom(complex &value,
                      Distribution &distribution,
                      RandomEngine &randomEngine) {
  // #ifdef INTEL_COMPILER
  //     value.real() = -1.0 + 2.0*rand() / RAND_MAX; //
  //     distribution(randomEngine); value.imag() = -1.0 + 2.0*rand() /
  //     RAND_MAX;
  //     // distribution(randomEngine);
  // #else
  value.real(distribution(randomEngine));
  value.imag(distribution(randomEngine));
  // #endif
}
 
class DefaultRandomEngine : public std::mt19937 {
public:
  DefaultRandomEngine() { seed(Sisi4s::world->rank); }
};
 
/*
    std::normal_distribution<double> normalDistribution(0.0, 1.0);
*/
template <typename F, typename Distribution, typename RandomEngine>
void setRandomTensor(Tensor<F> &t,
                     Distribution &distribution,
                     RandomEngine &randomEngine) {
  int64_t indicesCount, *indices;
  F *values;
  t.read_local(&indicesCount, &indices, &values);
  for (int64_t i(0); i < indicesCount; ++i) {
    setRandom(values[i], distribution, randomEngine);
  }
  t.write(indicesCount, indices, values);
  free(indices);
  free(values);
}
} // namespace sisi4s
 
#endif