Commit b97afaff authored by gbalduzz's avatar gbalduzz
Browse files

Merge remote-tracking branch 'origin/master' into gpu_trunk2

parents 7ae71330 d510f6f1

CITATION.md

+1 −1
Original line number Diff line number Diff line
@@ -4,5 +4,5 @@ If the DCA++ code contributes to your research, we kindly ask to acknowledge thi 


Urs R. Hähner, Gonzalo Alvarez, Thomas A. Maier, Raffaele Solcà, Peter Staar, Michael S. Summers, and Thomas C. Schulthess,

*DCA++: A software framework to solve correlated electron problems with modern quantum cluster methods,*

Comput. Phys. Commun (2019),

Comput. Phys. Commun. (2019),

10.1016/j.cpc.2019.01.006.

include/dca/linalg/matrix.hpp

+2 −0
Original line number Diff line number Diff line
@@ -25,6 +25,7 @@ 
#include "dca/linalg/util/allocators/allocators.hpp"

#include "dca/linalg/device_type.hpp"

#include "dca/linalg/util/copy.hpp"

#include "dca/linalg/util/memory.hpp"

#include "dca/linalg/util/stream_functions.hpp"



namespace dca {
@@ -290,6 +291,7 @@ Matrix<ScalarType, device_name>::Matrix(const std::string& name, std::pair<int, 
  assert(capacity_.first >= capacity.first && capacity_.second >= capacity.second);



  data_ = Allocator::allocate(nrElements(capacity_));

  util::Memory<device_name>::setToZero(data_, nrElements(capacity_));

}



template <typename ScalarType, DeviceType device_name>

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

+17 −15
Original line number Diff line number Diff line
@@ -396,10 +396,12 @@ double TpAccumulator<Parameters, linalg::CPU>::updateG4() { 


  switch (mode_) {

    case PARTICLE_HOLE_MAGNETIC:

      // Note: sums over spin indices are implied.

      //

      // G4(k1, k2, k_ex) = 1/2 (s1 * s2) <c^+(k1 + k_ex, s1) c(k1, s1)

      //                    c^+(k2, s2) c(k2 + k_ex, s2)>

      //                     = 1/2 (s1 * s2) <G(k1, k1 + k_ex, s1) G(k2 + k_ex, k2, s2) - (s1 ==s2)

      //                       G(k2 + k_ex, k1 + k_ex, s1) G(k1, k2, s1)>

      //                  = 1/2 (s1 * s2) <G(k1, k1 + k_ex, s1) G(k2 + k_ex, k2, s2) -

      //                    (s1 == s2) G(k2 + k_ex, k1 + k_ex, s1) G(k1, k2, s1)>.

      for (int w_ex_idx = 0; w_ex_idx < exchange_frq.size(); ++w_ex_idx) {

        const int w_ex = exchange_frq[w_ex_idx];

        for (int w2 = 0; w2 < WTpDmn::dmn_size(); ++w2)
@@ -425,7 +427,7 @@ double TpAccumulator<Parameters, linalg::CPU>::updateG4() { 
    case PARTICLE_HOLE_CHARGE:

      // G4(k1, k2, k_ex) += 1/2  <c^+(k1 + k_ex, s1) c(k1, s1) c^+(k2, s2) c(k2 + k_ex, s2)> =

      //                  = 1/2 <G(k1, k1 + k_ex, s1) G(k2 + k_ex, k2, s2) -

      //                  (s1 ==s2) G(k2 + k_ex, k1 + k_ex, s1) G(k1, k2, s1)>

      //                    (s1 == s2) G(k2 + k_ex, k1 + k_ex, s1) G(k1, k2, s1)>.

      for (int w_ex_idx = 0; w_ex_idx < exchange_frq.size(); ++w_ex_idx) {

        const int w_ex = exchange_frq[w_ex_idx];

        for (int w2 = 0; w2 < WTpDmn::dmn_size(); ++w2)
@@ -450,8 +452,8 @@ double TpAccumulator<Parameters, linalg::CPU>::updateG4() { 
      break;



    case PARTICLE_HOLE_TRANSVERSE:

      // G4 = 1/2 \sum_s <c^+(k1 + k_ex, s) c(k1, -s) c^+(k2, -s) c(k2 + k_ex, s)>

      //    = -1/2 \sum_s G( k2 + k_ex, k1 + k_ex, s) G4(k1, k2, -s)

      // G4 = 1/2 <c^+(k1 + k_ex, s) c(k1, -s) c^+(k2, -s) c(k2 + k_ex, s)>

      //    = -1/2 G(k2 + k_ex, k1 + k_ex, s) G(k1, k2, -s).

      for (int w_ex_idx = 0; w_ex_idx < exchange_frq.size(); ++w_ex_idx) {

        const int w_ex = exchange_frq[w_ex_idx];

        for (int w2 = 0; w2 < WTpDmn::dmn_size(); ++w2)
@@ -473,8 +475,8 @@ double TpAccumulator<Parameters, linalg::CPU>::updateG4() { 
      break;



    case PARTICLE_PARTICLE_UP_DOWN:

      // G4 = 1/2 \sum_s <c^+(k_ex-k1, s) c^+(k1, -s) c(k2, -s) c(k_ex-k2, s)>

      //    = 1/2 \sum_s G(k_ex-k2, k_ex-k1, s) G4(k2, k1, -s)

      // G4 = 1/2 <c^+(k_ex-k1, s) c^+(k1, -s) c(k2, -s) c(k_ex-k2, s)>

      //    = 1/2 G(k_ex-k2, k_ex-k1, s) G(k2, k1, -s).

      for (int w_ex_idx = 0; w_ex_idx < exchange_frq.size(); ++w_ex_idx) {

        const int w_ex = exchange_frq[w_ex_idx];

        for (int w2 = 0; w2 < WTpDmn::dmn_size(); ++w2)
@@ -544,9 +546,9 @@ void TpAccumulator<Parameters, linalg::CPU>::updateG4SpinDifference( 
    const bool cross_legs) {

  // This function performs the following update for each band:

  //

  // G4(k1, k2, w1, w2) += alpha * (G(up, k1_a, k2_a, w1_a, w2_a) * G(up, k1_b, k2_b, w1_b, w2_b)

  //                       + sign * G(down, k1_a, k2_a, w1_a, w2_a) * G(down, k1_b, k2_b, w1_b,

  //                       w2_b)

  // G4(k1, k2, w1, w2) += alpha * (G(up, k1_a, k2_a, w1_a, w2_a)

  //                       + sign * G(down,k1_a, k2_a, w1_a, w2_a)) *

  //                          (G(up,k1_b,k2_b,w1_b,w2_b) + sign * G(down,k1_b,k2_b,w1_b,w2_b))

  if (n_bands_ == 1) {

      *G4_ptr += alpha * (getGSingleband(0, k1_a, k2_a, w1_a, w2_a) +

			  Complex(sign) * getGSingleband(1, k1_a, k2_a, w1_a, w2_a)) *

src/config/cmake_options.cpp.in

+1 −1
Original line number Diff line number Diff line
@@ -64,7 +64,7 @@ void CMakeOptions::print() { 
            << "DCA_WITH_SINGLE_PRECISION_MEASUREMENTS = " << dca_with_single_precision_measurements

            << "\n"

            << "DCA_WITH_MEMORY_SAVINGS = " << CMakeOptions::dca_with_memory_savings << "\n"

            << "DCA_WITH_MANAGED_MEMORY = " << CMakeOptions::dca_with_memory_savings << "\n"

            << "DCA_WITH_MANAGED_MEMORY = " << CMakeOptions::dca_with_managed_memory << "\n"

            << "\n"

            << "DCA_WITH_QMC_BIT = " << CMakeOptions::dca_with_qmc_bit << "\n"

            << std::endl;

src/linalg/lapack/multiply_diagonal_gpu.cu

+2 −2
Original line number Diff line number Diff line
@@ -5,8 +5,8 @@ 
// See LICENSE for terms of usage.

// See CITATION.md for citation guidelines, if DCA++ is used for scientific publications.

//

// Author: Raffaele Solca' (rasolca@itp.phys.ethz.ch)

//         Giovanni Balduzzi (gbalduzz@itp.phys.ethz.ch)

// Author: Giovanni Balduzzi (gbalduzz@itp.phys.ethz.ch)

//         Raffaele Solca' (rasolca@itp.phys.ethz.ch)

//

// This file implements laset_gpu.hpp.