Unverified Commit 12ce4238 authored by gbalduzz's avatar gbalduzz Committed by GitHub
Browse files

Merge branch 'master' into const_matrix_view

parents 7f93b249 99b884ac

include/dca/linalg/util/handle_functions.hpp

+5 −2
Original line number Diff line number Diff line
@@ -65,8 +65,11 @@ inline cublasHandle_t getHandle(const int thread_id, const int stream_id) { 


#else



// Implement SFINAE.

inline void resizeHandleContainer(int /*max_threads*/) {}



inline void resizeHandleContainer(const std::size_t max_threads) {

  if (getStreamContainer().get_max_threads() < max_threads)

    resizeStreamContainer(max_threads);

}



#endif  // DCA_HAVE_CUDA

include/dca/linalg/util/stream_container.hpp

+9 −8
Original line number Diff line number Diff line
@@ -13,6 +13,7 @@ 
#ifndef DCA_LINALG_UTIL_STREAM_CONTAINER_HPP

#define DCA_LINALG_UTIL_STREAM_CONTAINER_HPP



#include <array>

#include <cassert>

#include <functional>

#include <vector>
@@ -26,18 +27,18 @@ namespace util { 


class StreamContainer {

public:

  StreamContainer(int max_threads = 0) : streams_(max_threads * streams_per_thread_) {}

  StreamContainer(std::size_t max_threads = 0) : streams_(max_threads) {}



  int get_max_threads() const {

    return streams_.size() / streams_per_thread_;

  std::size_t get_max_threads() const {

    return streams_.size();

  }



  int get_streams_per_thread() const {

  std::size_t get_streams_per_thread() const {

    return streams_per_thread_;

  }



  void resize(const int max_threads) {

    streams_.resize(max_threads * streams_per_thread_);

    streams_.resize(max_threads);

  }



  StreamContainer(const StreamContainer&) = delete;
@@ -49,7 +50,7 @@ public: 
  CudaStream& operator()(int thread_id, int stream_id) {

    assert(thread_id >= 0 && thread_id < get_max_threads());

    assert(stream_id >= 0 && stream_id < streams_per_thread_);

    return streams_[stream_id + streams_per_thread_ * thread_id];

    return streams_[thread_id][stream_id];

  }



  // Synchronizes the 'stream_id'-th stream associated with thread 'thread_id'.
@@ -60,8 +61,8 @@ public: 
  }



private:

  constexpr static int streams_per_thread_ = 2;

  std::vector<CudaStream> streams_;

  constexpr static std::size_t streams_per_thread_ = 2;

  std::vector<std::array<CudaStream, streams_per_thread_>> streams_;

};



}  // namespace util

include/dca/phys/models/general_interaction.hpp

+4 −0
Original line number Diff line number Diff line
@@ -76,6 +76,9 @@ void general_interaction<parameters_type>::set_vertex( 


  // Get a random pair of correlated spin-orbitals.

  const int pos = rng() * correlated_orbitals.size();

  // This is instead of crashing on segv when module is unknown.

  // \todo catch earlier in release as well.

  assert( pos < correlated_orbitals.size() && "It is likely you have specified an unknown model" );

  const int lin_ind = correlated_orbitals[pos];



  std::array<int, 6> sub_ind;  // [0]=b1, [1]=s1, [2]=b2, [3]=s2, [4]=r1, [5]=r2
@@ -118,6 +121,7 @@ std::vector<int> general_interaction<parameters_type>::make_correlated_orbitals( 
        for (auto b_j : interacting_orbitals) {

          for (int s_i = 0; s_i < SpinDmn::dmn_size(); ++s_i) {

            for (auto b_i : interacting_orbitals) {

              // This 1.e-3 seems like a very magic number!

              if (std::abs(H_interaction(b_i, s_i, b_j, s_j, delta_r)) > 1.e-3) {

                int linear_index = get_spin_orbital_pair_domain<BandDmn, SpinDmn, RDmn>()(

                    b_i, s_i, b_j, s_j, r_i, r_j);