Commit 83d87224 authored by Weile's avatar Weile
Browse files

passing meas_id to accumulator related function

parent 67a1763e

include/dca/phys/dca_step/cluster_solver/ctaux/ctaux_accumulator.hpp

+12 −10
Original line number Diff line number Diff line
@@ -54,7 +54,8 @@ namespace solver { 
namespace ctaux {

// dca::phys::solver::ctaux::



  template <dca::linalg::DeviceType device_t, class Parameters, class Data, DistType DIST = dca::DistType::NONE, typename Real = double>

template <dca::linalg::DeviceType device_t, class Parameters, class Data,

          DistType DIST = dca::DistType::NONE, typename Real = double>

class CtauxAccumulator : public MC_accumulator_data {

public:

  using this_type = CtauxAccumulator<device_t, Parameters, Data, DIST, Real>;
@@ -95,7 +96,7 @@ public: 
  template <typename walker_type>

  void updateFrom(walker_type& walker);



  void measure();

  void measure(const int meas_id = -1);



  // Sums all accumulated objects of this accumulator to the equivalent objects

  // of the 'other' accumulator.
@@ -173,7 +174,7 @@ private: 
  void accumulate_equal_time_quantities(const std::array<linalg::Matrix<Real, linalg::GPU>, 2>& M);

  void accumulate_equal_time_quantities(const std::array<linalg::Matrix<Real, linalg::CPU>, 2>& M);



  void accumulate_two_particle_quantities();

  void accumulate_two_particle_quantities(const int meas_id = -1);



protected:

  const Parameters& parameters_;
@@ -212,8 +213,8 @@ protected: 
};



template <dca::linalg::DeviceType device_t, class Parameters, class Data, DistType DIST, typename Real>

CtauxAccumulator<device_t, Parameters, Data, DIST, Real>::CtauxAccumulator(const Parameters& parameters_ref,

                                                                     Data& data_ref, int id)

CtauxAccumulator<device_t, Parameters, Data, DIST, Real>::CtauxAccumulator(

    const Parameters& parameters_ref, Data& data_ref, int id)

    : MC_accumulator_data(),



      parameters_(parameters_ref),
@@ -365,12 +366,12 @@ void CtauxAccumulator<device_t, Parameters, Data, DIST, Real>::updateFrom(walker 
}



template <dca::linalg::DeviceType device_t, class Parameters, class Data, DistType DIST, typename Real>

void CtauxAccumulator<device_t, Parameters, Data, DIST, Real>::measure() {

void CtauxAccumulator<device_t, Parameters, Data, DIST, Real>::measure(const int meas_id) {

  number_of_measurements += 1;

  accumulated_sign += current_sign;



  if (perform_tp_accumulation_)

    accumulate_two_particle_quantities();

    accumulate_two_particle_quantities(meas_id);



  accumulate_single_particle_quantities();
@@ -467,9 +468,10 @@ void CtauxAccumulator<device_t, Parameters, Data, DIST, Real>::accumulate_equal_ 
 *************************************************************/



template <dca::linalg::DeviceType device_t, class Parameters, class Data, DistType DIST, typename Real>

void CtauxAccumulator<device_t, Parameters, Data, DIST, Real>::accumulate_two_particle_quantities() {

void CtauxAccumulator<device_t, Parameters, Data, DIST, Real>::accumulate_two_particle_quantities(

    const int meas_id) {

  profiler_type profiler("tp-accumulation", "CT-AUX accumulator", __LINE__, thread_id);

  GFLOP += 1e-9 * two_particle_accumulator_.accumulate(M_, hs_configuration_, current_sign);

  GFLOP += 1e-9 * two_particle_accumulator_.accumulate(M_, hs_configuration_, current_sign, meas_id);

}



template <dca::linalg::DeviceType device_t, class Parameters, class Data, DistType DIST, typename Real>

include/dca/phys/dca_step/cluster_solver/shared_tools/accumulation/tp/tp_accumulator_gpu.hpp

+4 −4
Original line number Diff line number Diff line
@@ -73,12 +73,12 @@ public: 
  // Returns: number of flop.

  template <class Configuration, typename RealIn>

  float accumulate(const std::array<linalg::Matrix<RealIn, linalg::GPU>, 2>& M,

                   const std::array<Configuration, 2>& configs, int sign);

                   const std::array<Configuration, 2>& configs, int sign, const int meas_id = -1);



  // CPU input. For testing purposes.

  template <class Configuration>

  float accumulate(const std::array<linalg::Matrix<double, linalg::CPU>, 2>& M,

                   const std::array<Configuration, 2>& configs, int sign);

                   const std::array<Configuration, 2>& configs, int sign, const int meas_id = -1);



  // Downloads the accumulation result to the host.

  void finalize();
@@ -296,7 +296,7 @@ template <class Parameters> 
template <class Configuration, typename RealIn>

float TpAccumulator<Parameters, linalg::GPU, DistType::NONE>::accumulate(

    const std::array<linalg::Matrix<RealIn, linalg::GPU>, 2>& M,

    const std::array<Configuration, 2>& configs, const int sign) {

    const std::array<Configuration, 2>& configs, const int sign, const int meas_id) {

  Profiler profiler("accumulate", "tp-accumulation", __LINE__, thread_id_);

  float flop = 0;
@@ -321,7 +321,7 @@ template <class Parameters> 
template <class Configuration>

float TpAccumulator<Parameters, linalg::GPU>::accumulate(

    const std::array<linalg::Matrix<double, linalg::CPU>, 2>& M,

    const std::array<Configuration, 2>& configs, const int sign) {

    const std::array<Configuration, 2>& configs, const int sign, const int meas_id) {

  std::array<linalg::Matrix<double, linalg::GPU>, 2> M_dev;

  for (int s = 0; s < 2; ++s)

    M_dev[s].setAsync(M[s], queues_[0]);

include/dca/phys/dca_step/cluster_solver/shared_tools/accumulation/tp/tp_accumulator_mpi_gpu.hpp

+10 −10
Original line number Diff line number Diff line
@@ -77,12 +77,12 @@ private: 
public:

  template <class Configuration, typename RealIn>

  float accumulate(const std::array<linalg::Matrix<RealIn, linalg::GPU>, 2>& M,

                   const std::array<Configuration, 2>& configs, int sign);

                   const std::array<Configuration, 2>& configs, int sign, const int meas_id = -1);



  // CPU input. For testing purposes.

  template <class Configuration>

  float accumulate(const std::array<linalg::Matrix<double, linalg::CPU>, 2>& M,

                   const std::array<Configuration, 2>& configs, int sign);

                   const std::array<Configuration, 2>& configs, int sign, const int meas_id = -1);



  // Downloads the accumulation result to the host.

  void finalize();
@@ -123,10 +123,10 @@ private: 
  std::array<MPI_Request, 2> recv_requests_{MPI_REQUEST_NULL, MPI_REQUEST_NULL};

  std::array<MPI_Request, 2> send_requests_{MPI_REQUEST_NULL, MPI_REQUEST_NULL};



#ifndef DCA_HAVE_CUDA_AWARE_MPI

#ifndef DCA_WITH_CUDA_AWARE_MPI

  std::array<std::vector<Complex>, 2> sendbuffer_;

  std::array<std::vector<Complex>, 2> recvbuffer_;

#endif  // DCA_HAVE_CUDA_AWARE_MPI

#endif  // DCA_WITH_CUDA_AWARE_MPI

};



template <class Parameters>
@@ -162,7 +162,7 @@ template <class Parameters> 
template <class Configuration, typename RealIn>

float TpAccumulator<Parameters, linalg::GPU, DistType::MPI>::accumulate(

    const std::array<linalg::Matrix<RealIn, linalg::GPU>, 2>& M,

    const std::array<Configuration, 2>& configs, const int sign) {

    const std::array<Configuration, 2>& configs, const int sign, const int meas_id) {

  // typename BaseClass::Profiler profiler("accumulate", "tp-accumulation", __LINE__, BaseClass::thread_id_);

  float flop = 0;
@@ -189,7 +189,7 @@ template <class Parameters> 
template <class Configuration>

float TpAccumulator<Parameters, linalg::GPU, DistType::MPI>::accumulate(

    const std::array<linalg::Matrix<double, linalg::CPU>, 2>& M,

    const std::array<Configuration, 2>& configs, const int sign) {

    const std::array<Configuration, 2>& configs, const int sign, const int meas_id) {

  std::array<linalg::Matrix<double, linalg::GPU>, 2> M_dev;

  for (int s = 0; s < 2; ++s)

    M_dev[s].setAsync(M[s], queues_[0]);
@@ -366,7 +366,7 @@ void TpAccumulator<Parameters, linalg::GPU, DistType::MPI>::send(const std::arra 
  using dca::parallel::MPITypeMap;

  const auto g_size = data[0].size().first * data[0].size().second;



#ifdef DCA_HAVE_CUDA_AWARE_MPI

#ifdef DCA_WITH_CUDA_AWARE_MPI

  for (int s = 0; s < 2; ++s) {

    MPI_Isend(data[s].ptr(), g_size, MPITypeMap<Complex>::value(), target, thread_id_ + 1,

              MPI_COMM_WORLD, &request[s]);
@@ -381,7 +381,7 @@ void TpAccumulator<Parameters, linalg::GPU, DistType::MPI>::send(const std::arra 
    MPI_Isend(sendbuffer_[s].data(), g_size, MPITypeMap<Complex>::value(), target, thread_id_ + 1,

              MPI_COMM_WORLD, &request[s]);

  }

#endif  // DCA_HAVE_CUDA_AWARE_MPI

#endif  // DCA_WITH_CUDA_AWARE_MPI

}



template <class Parameters>
@@ -390,7 +390,7 @@ void TpAccumulator<Parameters, linalg::GPU, DistType::MPI>::receive( 
  using dca::parallel::MPITypeMap;

  const auto g_size = data[0].size().first * data[0].size().second;



#ifdef DCA_HAVE_CUDA_AWARE_MPI

#ifdef DCA_WITH_CUDA_AWARE_MPI

  for (int s = 0; s < 2; ++s) {

    MPI_Irecv(data[s].ptr(), g_size, MPITypeMap<Complex>::value(), source, thread_id_ + 1,

              MPI_COMM_WORLD, &request[s]);
@@ -408,7 +408,7 @@ void TpAccumulator<Parameters, linalg::GPU, DistType::MPI>::receive( 
    cudaMemcpy(data[s].ptr(), recvbuffer_[s].data(), g_size * sizeof(Complex),

               cudaMemcpyHostToDevice);

  }

#endif  // DCA_HAVE_CUDA_AWARE_MPI

#endif  // DCA_WITH_CUDA_AWARE_MPI

}



}  // namespace accumulator

include/dca/phys/dca_step/cluster_solver/stdthread_qmci/stdthread_qmci_cluster_solver.hpp

+1 −1
Original line number Diff line number Diff line
@@ -425,7 +425,7 @@ void StdThreadQmciClusterSolver<QmciSolver>::startWalkerAndAccumulator(int id) { 
      {

        Profiler profiler("Accumulator measuring", "stdthread-MC", __LINE__, id);

        accumulator_obj.updateFrom(walker);

        accumulator_obj.measure();

        accumulator_obj.measure(meas_id);

      }

      if (print)

        walker.updateShell(meas_id, n_meas);