Commit 1575944c authored by Doak, Peter W.'s avatar Doak, Peter W.
Browse files

Merge branch 'gpu_trunk2' into master

parents a8196ec3 c685f954

cmake/dca_config.cmake

+9 −0
Original line number Diff line number Diff line
@@ -270,6 +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()
@@ -282,6 +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

+14 −3
Original line number Diff line number Diff line
@@ -9,8 +9,14 @@ 
//

// This class stores compile time options for the MC accumulation.



#ifndef DCA_CONFIG_ACCUMULATON_OPTIONS_HPP

#define DCA_CONFIG_ACCUMULATON_OPTIONS_HPP

#ifndef DCA_CONFIG_ACCUMULATION_OPTIONS_HPP

#define DCA_CONFIG_ACCUMULATION_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 {
@@ -20,9 +26,14 @@ 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

}  // dca



#endif  // DCA_CONFIG_ACCUMULATON_OPTIONS_HPP
 
#endif  // DCA_CONFIG_ACCUMULATION_OPTIONS_HPP

include/dca/config/cmake_options.hpp

+3 −2
Original line number Diff line number Diff line
@@ -43,13 +43,14 @@ 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;



  static void print();

};



}  // config

}  // dca

}  // namespace config

}  // namespace dca



#endif  // DCA_CONFIG_CMAKE_OPTIONS_HPP

include/dca/linalg/vector.hpp

+3 −2
Original line number Diff line number Diff line
@@ -38,6 +38,7 @@ class Vector : public Allocator { 
public:

  using ThisType = Vector<ScalarType, device_name, Allocator>;

  using ValueType = ScalarType;

  using AllocatorType = Allocator;



  Vector();

  Vector(const std::string& name);
@@ -405,7 +406,7 @@ std::size_t Vector<ScalarType, device_name, Allocator>::deviceFingerprint() cons 
  return device_name == GPU ? capacity_ * sizeof(ScalarType) : 0;

}



}  // linalg

}  // dca

}  // namespace linalg

}  // namespace dca



#endif  // DCA_LINALG_VECTOR_HPP