Commit deab721b authored by Thomas Maier's avatar Thomas Maier
Browse files

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

Conflicts:
	src/phys/dca_step/cluster_solver/shared_tools/accumulation/tp/tp_accumulator_kernels.cu
parents 64c09232 1f18ae79

applications/cluster_solver_check/cluster_solver_check.cpp

+15 −10
Original line number Diff line number Diff line
@@ -35,7 +35,7 @@ 


int main(int argc, char** argv) {

  if (argc < 2) {

    std::cerr << "Usage: " << argv[0] << " input_file.json" << std::endl;

    std::cerr << "Usage: " << argv[0] << " input_file.json [skip ed]" << std::endl;

    return -1;

  }
@@ -43,8 +43,8 @@ int main(int argc, char** argv) { 


  try {

    std::string input_file(argv[1]);



    const bool perform_statistical_test = concurrency.number_of_processors() >= 8;

    const bool skip_ed = argc > 2 ? std::atoi(argv[2]) : false;

    const bool perform_statistical_test = concurrency.number_of_processors() >= 8 && !skip_ed;



    Profiler::start();
@@ -94,21 +94,26 @@ int main(int argc, char** argv) { 


    // ED solver

    EdSolver ed_solver(parameters, dca_data_imag, dca_data_real);

    if (!skip_ed) {

      ed_solver.initialize(0);

      ed_solver.execute();

      ed_solver.finalize(dca_loop_data);



    const auto Sigma_ed(dca_data_imag.Sigma);

    const int tested_frequencies = 10;

    const auto G_ed(dca::math::util::cutFrequency(dca_data_imag.G_k_w, tested_frequencies));



      if (concurrency.id() == concurrency.first()) {

        ed_solver.write(data_file_ed);

      }

    }



    const auto Sigma_ed(dca_data_imag.Sigma);

    const int tested_frequencies = 10;

    const auto G_ed(dca::math::util::cutFrequency(dca_data_imag.G_k_w, tested_frequencies));



    // QMC solver

    // The QMC solver uses the free Greens function G0 computed by the ED solver.

    // It is passed via the dca_data_imag object.

    if (skip_ed)

      dca_data_imag.initialize();



    ClusterSolver qmc_solver(parameters, dca_data_imag);

    qmc_solver.initialize(1);  // 1 = dummy iteration number

    qmc_solver.integrate();

cmake/dca_config.cmake

+22 −2
Original line number Diff line number Diff line
@@ -68,7 +68,8 @@ else() 
endif()



# Lattice type

set(DCA_LATTICE "square" CACHE STRING "Lattice type, options are: bilayer | square | triangular.")

set(DCA_LATTICE "square" CACHE STRING

    "Lattice type, options are: bilayer | square | triangular | twoband_chain | singleband_chain .")

set_property(CACHE DCA_LATTICE PROPERTY STRINGS bilayer square triangular)



if (DCA_LATTICE STREQUAL "bilayer")
@@ -86,8 +87,18 @@ elseif (DCA_LATTICE STREQUAL "triangular") 
  set(DCA_LATTICE_INCLUDE

    "dca/phys/models/analytic_hamiltonians/triangular_lattice.hpp")



elseif (DCA_LATTICE STREQUAL "twoband_chain")

  set(DCA_LATTICE_TYPE dca::phys::models::twoband_chain<dca::phys::domains::no_symmetry<1>>)

  set(DCA_LATTICE_INCLUDE

      "dca/phys/models/analytic_hamiltonians/twoband_chain.hpp")



elseif (DCA_LATTICE STREQUAL "singleband_chain")

  set(DCA_LATTICE_TYPE dca::phys::models::singleband_chain<dca::phys::domains::no_symmetry<1>>)

  set(DCA_LATTICE_INCLUDE

      "dca/phys/models/analytic_hamiltonians/singleband_chain.hpp")

else()

  message(FATAL_ERROR "Please set DCA_LATTICE to a valid option: bilayer | square | triangular.")

  message(FATAL_ERROR

          "Please set DCA_LATTICE to a valid option: bilayer | square | triangular | twoband_chain | singleband_chain .")

endif()



# Model type
@@ -259,6 +270,7 @@ endif() 
# Accumulation options.

option(DCA_WITH_MEMORY_SAVINGS "Save memory in the two particle accumulation at a slight performance

       cost." OFF)

mark_as_advanced(DCA_WITH_MEMORY_SAVINGS)

if (DCA_WITH_MEMORY_SAVINGS)

  set(MEMORY_SAVINGS true)

else()
@@ -271,6 +283,14 @@ else() 
  set(MC_ACCUMULATION_SCALAR double)

endif()



option(DCA_WITH_MANAGED_MEMORY "Use managed memory allocator." OFF)

mark_as_advanced(DCA_WITH_MANAGED_MEMORY)

if (DCA_WITH_MANAGED_MEMORY)

  set(TWO_PARTICLE_ALLOCATOR "dca::linalg::util::ManagedAllocator<T>")

else()

  set(TWO_PARTICLE_ALLOCATOR "dca::linalg::util::DeviceAllocator<T>")

endif()



configure_file("${PROJECT_SOURCE_DIR}/include/dca/config/accumulation_options.hpp.in"

        "${CMAKE_BINARY_DIR}/include/dca/config/accumulation_options.hpp" @ONLY)

cmake/dca_cuda.cmake

+1 −0
Original line number Diff line number Diff line
@@ -19,6 +19,7 @@ if (CUDA_FOUND) 
  # dca_add_haves_define(DCA_HAVE_CUDA)

  list(APPEND DCA_CUDA_LIBS ${CUDA_LIBRARIES} ${CUDA_cusparse_LIBRARY} ${CUDA_cublas_LIBRARY})

  CUDA_INCLUDE_DIRECTORIES(${CUDA_INCLUDE_DIRS})

  set(CUDA_SEPARABLE_COMPILATION ON)

endif()



# Find MAGMA.

include/dca/config/accumulation_options.hpp.in

+11 −0
Original line number Diff line number Diff line
@@ -12,6 +12,12 @@ 
#ifndef DCA_CONFIG_ACCUMULATON_OPTIONS_HPP

#define DCA_CONFIG_ACCUMULATON_OPTIONS_HPP



#ifdef DCA_HAVE_CUDA

#include "dca/linalg/util/allocators/device_allocator.hpp"

#include "dca/linalg/util/allocators/managed_allocator.hpp"

#endif  // DCA_HAVE_CUDA





namespace dca {

namespace config {

// dca::config::
@@ -20,6 +26,11 @@ struct AccumulationOptions { 
  using MCAccumulationScalar = @MC_ACCUMULATION_SCALAR@;



  static constexpr bool memory_savings = @MEMORY_SAVINGS@;



#ifdef DCA_HAVE_CUDA

  template <typename T>

  using TpAllocator = @TWO_PARTICLE_ALLOCATOR@;

#endif  // DCA_HAVE_CUDA

};



}  // config

include/dca/config/cmake_options.hpp

+1 −0
Original line number Diff line number Diff line
@@ -43,6 +43,7 @@ struct CMakeOptions { 
  static const std::string dca_with_reduced_vertex_function;

  static const std::string dca_with_single_precision_measurements;

  static const std::string dca_with_memory_savings;

  static const std::string dca_with_managed_memory;



  static const std::string dca_with_qmc_bit;