diff --git a/CMakeLists.txt b/CMakeLists.txt
new file mode 100644
index 0000000000000000000000000000000000000000..87a205e64b631f60178a47fbc45df1ec7a19467b
--- /dev/null
+++ b/CMakeLists.txt
@@ -0,0 +1,62 @@
+cmake_minimum_required(VERSION 3.8 FATAL_ERROR)
+project(openfpm_pdata LANGUAGES C CXX CUDA)
+
+
+list(APPEND CMAKE_MODULE_PATH ${CMAKE_CURRENT_LIST_DIR}/CMakeFiles/)
+
+set(BOOST_INCLUDE ${Boost_INCLUDE_DIR} CACHE PATH "Include directory for BOOST")
+set(SE_CLASS1 CACHE BOOL "Activate compilation with SE_CLASS1")
+set(SE_CLASS2 CACHE BOOL "Activate compilation with SE_CLASS2")
+set(SE_CLASS3 CACHE BOOL "Activate compilation with SE_CLASS3")
+
+
+find_package(Boost 1.52.0 REQUIRED unit_test_framework iostreams program_options)
+find_package(CUDA)
+find_package(MPI REQUIRED)
+
+if(CUDA_FOUND)
+ set(OPENFPM_INIT_FILE "initialize/initialize_wrapper_cuda.cu")
+else()
+ set(OPENFPM_INIT_FILE "initialize/initialize_wrapper_cuda.cpp")
+endif()
+
+###### CONFIG.h FILE ######
+
+if(SE_CLASS1)
+ set(DEFINE_SE_CLASS1 "#define SE_CLASS1")
+endif()
+
+if(SE_CLASS2)
+ set(DEFINE_SE_CLASS2 "#define SE_CLASS2")
+endif()
+
+if(SE_CLASS3)
+ set(DEFINE_SE_CLASS3 "#define SE_CLASS3")
+endif()
+
+if(CUDA_FOUND)
+ set(DEFINE_CUDA_GPU "#define CUDA_GPU")
+endif()
+
+if(MPI_FOUND)
+ set(DEFINE_HAVE_MPI "#define HAVE_MPI")
+else()
+ message( FATAL_ERROR "MPI is required in order to install OpenFPM" )
+endif()
+
+if (Boost_FOUND)
+ set(DEFINE_HAVE_BOOST "#define HAVE_BOOST")
+ set(DEFINE_HAVE_BOOST_IOSTREAMS "#define HAVE_BOOST_IOSTREAMS")
+ set(DEFINE_HAVE_BOOST_PROGRAM_OPTIONS "#define HAVE_BOOST_PROGRAM_OPTIONS")
+ set(DEFINE_HAVE_BOOST_UNIT_TEST_FRAMEWORK "#define HAVE_BOOST_UNIT_TEST_FRAMEWORK")
+else()
+ message( FATAL_ERROR "BOOST is required in order to install OpenFPM" )
+endif()
+
+
+configure_file(${CMAKE_CURRENT_SOURCE_DIR}/src/config/config_cmake.h.in ${CMAKE_CURRENT_SOURCE_DIR}/src/config/config.h)
+
+include_directories(SYSTEM ${MPI_INCLUDE_PATH})
+
+add_subdirectory (src)
+
diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
new file mode 100644
index 0000000000000000000000000000000000000000..9cb41bb2b10dfce34eae65c346ceff64057c3334
--- /dev/null
+++ b/src/CMakeLists.txt
@@ -0,0 +1,66 @@
+cmake_minimum_required(VERSION 3.8 FATAL_ERROR)
+
+########################### Executables
+
+add_executable(vcluster_test main.cpp VCluster/VCluster.cpp ../../openfpm_devices/src/memory/HeapMemory.cpp ../../openfpm_devices/src/memory/PtrMemory.cpp ../../openfpm_devices/src/Memleak_check.cpp ../../openfpm_devices/src/memory/CudaMemory.cu VCluster/cuda/VCluster_semantic_unit_cuda_tests.cu )
+
+add_library(vcluster STATIC VCluster/VCluster.cpp)
+
+###########################
+
+target_compile_options(vcluster_test PUBLIC $<$<COMPILE_LANGUAGE:CUDA>:-Xcudafe "--display_error_number --diag_suppress=2885 --diag_suppress=2887 --diag_suppress=2888 --diag_suppress=186 --diag_suppress=111" --expt-extended-lambda>)
+
+target_include_directories (vcluster_test PUBLIC ${CUDA_INCLUDE_DIRS})
+target_include_directories (vcluster_test PUBLIC ${CMAKE_CURRENT_SOURCE_DIR})
+target_include_directories (vcluster_test PUBLIC ${CMAKE_CURRENT_SOURCE_DIR}/../../openfpm_devices/src/)
+target_include_directories (vcluster_test PUBLIC ${CMAKE_CURRENT_SOURCE_DIR}/../../openfpm_vcluster/src/)
+target_include_directories (vcluster_test PUBLIC ${CMAKE_CURRENT_SOURCE_DIR}/../../openfpm_data/src/)
+target_include_directories (vcluster_test PUBLIC ${CMAKE_CURRENT_SOURCE_DIR}/config)
+
+target_include_directories (vcluster PUBLIC ${CMAKE_CURRENT_SOURCE_DIR})
+target_include_directories (vcluster PUBLIC ${CMAKE_CURRENT_SOURCE_DIR}/config)
+target_include_directories (vcluster PUBLIC ${CMAKE_CURRENT_SOURCE_DIR}/../../openfpm_data/src/)
+target_include_directories (vcluster PUBLIC ${CMAKE_CURRENT_SOURCE_DIR}/../../openfpm_devices/src/)
+
+target_link_libraries(vcluster_test ${Boost_LIBRARIES})
+
+# Request that particles be built with -std=c++11
+# As this is a public compile feature anything that links to particles
+# will also build with -std=c++11
+target_compile_features(vcluster_test PUBLIC cxx_std_11)
+target_link_libraries(vcluster_test ${MPI_C_LIBRARIES})
+
+install(TARGETS vcluster DESTINATION openfpm_vcluster/lib)
+install(FILES MPI_wrapper/MPI_IallreduceW.hpp
+ MPI_wrapper/MPI_IrecvW.hpp
+ MPI_wrapper/MPI_IBcastW.hpp
+ MPI_wrapper/MPI_IsendW.hpp
+ MPI_wrapper/MPI_util.hpp
+ MPI_wrapper/MPI_IAllGather.hpp
+ DESTINATION openfpm_vcluster/include/MPI_wrapper)
+
+install(FILES VCluster/VCluster_base.hpp
+ VCluster/VCluster.hpp
+ VCluster/VCluster_meta_function.hpp
+ DESTINATION openfpm_vcluster/include/VCluster )
+
+install (FILES util/Vcluster_log.hpp
+ DESTINATION openfpm_vcluster/include/util)
+
+#if(BUILD_TESTING)
+
+# add_executable(particle_test test.cu)
+
+# set_target_properties(particle_test PROPERTIES CUDA_SEPARABLE_COMPILATION ON)
+# target_link_libraries(particle_test PRIVATE particles)
+
+# add_test(NAME particles_10k COMMAND particle_test 10000 )
+# add_test(NAME particles_256k COMMAND particle_test 256000 )
+
+# if(APPLE)
+ # We need to add the default path to the driver (libcuda.dylib) as an rpath,
+ # so that the static cuda runtime can find it at runtime.
+ # set_property(TARGET particle_test PROPERTY BUILD_RPATH ${CMAKE_CUDA_IMPLICIT_LINK_DIRECTORIES})
+ # endif()
+ #endif()
+
diff --git a/src/VCluster/VCluster.hpp b/src/VCluster/VCluster.hpp
index b5a74b435d6de7c67e161d69227d8d54377ce12a..aeb9aea2a13b7af9411e0803ae35677190270bb2 100644
--- a/src/VCluster/VCluster.hpp
+++ b/src/VCluster/VCluster.hpp
@@ -136,8 +136,6 @@ class Vcluster: public Vcluster_base<InternalMemory>
pack_unpack_cond_with_prp<has_max_prop<T, has_value_type<T>::value>::value, op, T, S, layout_base>::packing(mem, send.get(i), sts, send_buf,opt);
}
- self_base::tags.clear();
-
// receive information
base_info<InternalMemory> bi(&this->recv_buf,prc_recv,sz_recv_byte,this->tags,opt);
@@ -168,6 +166,7 @@ class Vcluster: public Vcluster_base<InternalMemory>
}
else
{
+ self_base::tags.clear();
prc_recv.clear();
self_base::sendrecvMultipleMessagesNBX(prc_send_.size(),(size_t *)send_sz_byte.getPointer(),(size_t *)prc_send_.getPointer(),(void **)send_buf.getPointer(),msg_alloc,(void *)&bi);
}
@@ -203,7 +202,7 @@ class Vcluster: public Vcluster_base<InternalMemory>
struct base_info
{
//! Receive buffer
- openfpm::vector<BMemory<Memory>> * recv_buf;
+ openfpm::vector_fr<BMemory<Memory>> * recv_buf;
//! receiving processor list
openfpm::vector<size_t> & prc;
//! size of each message
@@ -215,7 +214,7 @@ class Vcluster: public Vcluster_base<InternalMemory>
size_t opt;
//! constructor
- base_info(openfpm::vector<BMemory<Memory>> * recv_buf, openfpm::vector<size_t> & prc, openfpm::vector<size_t> & sz, openfpm::vector<size_t> & tags,size_t opt)
+ base_info(openfpm::vector_fr<BMemory<Memory>> * recv_buf, openfpm::vector<size_t> & prc, openfpm::vector<size_t> & sz, openfpm::vector<size_t> & tags,size_t opt)
:recv_buf(recv_buf),prc(prc),sz(sz),tags(tags),opt(opt)
{}
};
@@ -258,7 +257,7 @@ class Vcluster: public Vcluster_base<InternalMemory>
if (rinfo.opt & MPI_GPU_DIRECT)
{
#if defined(MPIX_CUDA_AWARE_SUPPORT) && MPIX_CUDA_AWARE_SUPPORT
- return rinfo.recv_buf->last().getDevicePointerNoCopy();
+ return rinfo.recv_buf->last().getDevicePointer();
#else
return rinfo.recv_buf->last().getPointer();
#endif
@@ -639,7 +638,7 @@ class Vcluster: public Vcluster_base<InternalMemory>
// we sort based on processor
rcv.sort();
- openfpm::vector<BMemory<InternalMemory>> recv_ord;
+ openfpm::vector_fr<BMemory<InternalMemory>> recv_ord;
recv_ord.resize(rcv.size());
openfpm::vector<size_t> prc_ord;
@@ -657,7 +656,12 @@ class Vcluster: public Vcluster_base<InternalMemory>
}
// move rcv into recv
- self_base::recv_buf.swap(recv_ord);
+ // Now we swap back to recv_buf in an ordered way
+ for (size_t i = 0 ; i < rcv.size() ; i++)
+ {
+ self_base::recv_buf.get(i).swap(recv_ord.get(i));
+ }
+
prc.swap(prc_ord);
sz_recv.swap(sz_recv_ord);
diff --git a/src/VCluster/VCluster_base.hpp b/src/VCluster/VCluster_base.hpp
index b9f54cc5f9d1abe473a9b1a8e39df646a23d8231..48a3b70ba99d61e8be6f66849a64ffbcc5d5fc71 100644
--- a/src/VCluster/VCluster_base.hpp
+++ b/src/VCluster/VCluster_base.hpp
@@ -142,16 +142,9 @@ class Vcluster_base
//! vector of functions to execute after all the request has been performed
std::vector<int> post_exe;
-#if defined(CUDA_GPU) && defined(__NVCC__)
-
- //! standard context for mgpu
+ //! standard context for mgpu (if cuda is detected otherwise is unused)
mgpu::standard_context_t * context;
-#else
-
- void * context = NULL;
-
-#endif
// Object array
@@ -198,7 +191,7 @@ class Vcluster_base
protected:
//! Receive buffers
- openfpm::vector<BMemory<InternalMemory>> recv_buf;
+ openfpm::vector_fr<BMemory<InternalMemory>> recv_buf;
//! tags receiving
openfpm::vector<size_t> tags;
@@ -287,7 +280,7 @@ public:
#if defined(CUDA_GPU) && defined(__NVCC__)
- context = new mgpu::standard_context_t();
+ context = new mgpu::standard_context_t(false);
#endif
}
@@ -345,25 +338,22 @@ public:
#endif
-#if defined(CUDA_GPU) && defined(__NVCC__)
-
/*! \brief If nvidia cuda is activated return a mgpu context
*
+ * \param iw ignore warning
*
*/
- mgpu::standard_context_t & getmgpuContext()
+ mgpu::standard_context_t & getmgpuContext(bool iw = true)
{
- if (context == NULL)
+ if (context == NULL && iw == true)
{
- std::cout << __FILE__ << ":" << __LINE__ << " error: it seem that modern gpu context is not initialized."
+ std::cout << __FILE__ << ":" << __LINE__ << " Warning: it seem that modern gpu context is not initialized."
"Either a compatible working cuda device has not been found, either openfpm_init has been called in a file that not compiled with NVCC" << std::endl;
}
return *context;
}
-#endif
-
/*! \brief Get the MPI_Communicator (or processor group) this VCluster is using
*
* \return MPI comunicator
diff --git a/src/VCluster/VCluster_meta_function.hpp b/src/VCluster/VCluster_meta_function.hpp
index 80baa15cdd1f60a47eb5d988bc19d49bbca990c5..55037ea4ef15b0ca428c04f2fc26257bc40e8d9c 100644
--- a/src/VCluster/VCluster_meta_function.hpp
+++ b/src/VCluster/VCluster_meta_function.hpp
@@ -17,7 +17,7 @@ struct unpack_selector_with_prp
template<typename op,
int ... prp>
static void call_unpack(S & recv,
- openfpm::vector<BMemory<Memory>> & recv_buf,
+ openfpm::vector_fr<BMemory<Memory>> & recv_buf,
openfpm::vector<size_t> * sz,
openfpm::vector<size_t> * sz_byte,
op & op_param,
@@ -76,7 +76,7 @@ struct unpack_each_prop_buffer
*
*/
inline unpack_each_prop_buffer(S & recv,
- openfpm::vector<BMemory<HeapMemory>> & recv_buf,
+ openfpm::vector_fr<BMemory<HeapMemory>> & recv_buf,
op & op_param,
size_t i,
openfpm::vector<size_t> * sz,
@@ -141,7 +141,7 @@ struct process_receive_mem_traits_inte
size_t i;
//! Receive buffer
- openfpm::vector<BMemory<Memory>> & recv_buf;
+ openfpm::vector_fr<BMemory<Memory>> & recv_buf;
//! Fake vector that map over received memory
openfpm::vector<typename sT::value_type,PtrMemory,typename layout_base<typename sT::value_type>::type,layout_base,openfpm::grow_policy_identity> & v2;
@@ -157,7 +157,7 @@ struct process_receive_mem_traits_inte
*
*/
inline process_receive_mem_traits_inte(openfpm::vector<typename sT::value_type,PtrMemory,typename layout_base<typename sT::value_type>::type,layout_base,openfpm::grow_policy_identity> & v2,
- openfpm::vector<BMemory<Memory>> & recv_buf,
+ openfpm::vector_fr<BMemory<Memory>> & recv_buf,
size_t i,
size_t opt)
:i(i),recv_buf(recv_buf),v2(v2),opt(opt)
@@ -174,11 +174,11 @@ struct process_receive_mem_traits_inte
PtrMemory * ptr1;
- if (opt == MPI_GPU_DIRECT)
+ if (opt & MPI_GPU_DIRECT)
{
#if defined(MPIX_CUDA_AWARE_SUPPORT) && MPIX_CUDA_AWARE_SUPPORT
// add the received particles to the vector
- ptr1 = new PtrMemory(recv_buf.get(i).getDevicePointerNoCopy(),recv_buf.get(i).size());
+ ptr1 = new PtrMemory(recv_buf.get(i).getDevicePointer(),recv_buf.get(i).size());
#else
// add the received particles to the vector
ptr1 = new PtrMemory(recv_buf.get(i).getPointer(),recv_buf.get(i).size());
@@ -200,7 +200,7 @@ template<bool inte_or_lin,typename T, typename S, template<typename> class layou
struct unpack_selector_with_prp_lin
{
template<typename op, unsigned int ... prp> static int call_unpack_impl(S & recv,
- openfpm::vector<BMemory<Memory>> & recv_buf,
+ openfpm::vector_fr<BMemory<Memory>> & recv_buf,
openfpm::vector<size_t> * sz,
openfpm::vector<size_t> * sz_byte,
op & op_param,
@@ -237,7 +237,7 @@ template<typename T, typename S, template<typename> class layout_base, typename
struct unpack_selector_with_prp_lin<true,T,S,layout_base,Memory>
{
template<typename op, unsigned int ... prp> static int call_unpack_impl(S & recv,
- openfpm::vector<BMemory<Memory>> & recv_buf,
+ openfpm::vector_fr<BMemory<Memory>> & recv_buf,
openfpm::vector<size_t> * sz,
openfpm::vector<size_t> * sz_byte,
op & op_param,
@@ -282,7 +282,7 @@ template<typename T, typename S, template<typename> class layout_base, typename
struct unpack_selector_with_prp<true,T,S,layout_base,Memory>
{
template<typename op, unsigned int ... prp> static void call_unpack(S & recv,
- openfpm::vector<BMemory<Memory>> & recv_buf,
+ openfpm::vector_fr<BMemory<Memory>> & recv_buf,
openfpm::vector<size_t> * sz,
openfpm::vector<size_t> * sz_byte,
op & op_param,
@@ -317,7 +317,7 @@ struct call_serialize_variadic<index_tuple<prp...>>
template<typename op, typename T, typename S, template<typename> class layout_base, typename Memory>
inline static void call_unpack(S & recv,
- openfpm::vector<BMemory<Memory>> & recv_buf,
+ openfpm::vector_fr<BMemory<Memory>> & recv_buf,
openfpm::vector<size_t> * sz,
openfpm::vector<size_t> * sz_byte,
op & op_param,
@@ -507,7 +507,7 @@ struct pack_unpack_cond_with_prp
template<typename Memory>
static void unpacking(S & recv,
- openfpm::vector<BMemory<Memory>> & recv_buf,
+ openfpm::vector_fr<BMemory<Memory>> & recv_buf,
openfpm::vector<size_t> * sz,
openfpm::vector<size_t> * sz_byte,
op & op_param,
@@ -553,7 +553,7 @@ struct op_ssend_recv_add_sr
layout_base,
prp...>(v2);
- recv.template hostToDevice<prp...>();
+ recv.template hostToDevice<prp...>(recv.size(),recv.size()+v2.size());
#endif
@@ -719,7 +719,7 @@ struct op_ssend_gg_recv_merge_impl<true>
recv.template merge_prp_v<replace_,
typename T::value_type,
HeapMemory,
- openfpm::grow_policy_double,
+ typename S::grow_policy,
layout_base,
prp...>(v2,start);
diff --git a/src/config/config_cmake.h.in b/src/config/config_cmake.h.in
new file mode 100644
index 0000000000000000000000000000000000000000..a0d62c9edc325fad8f80ce5ca11c06c068f57baf
--- /dev/null
+++ b/src/config/config_cmake.h.in
@@ -0,0 +1,159 @@
+/* Coverty scan */
+${DEFINE_COVERTY_SCAN}
+
+/* GPU support */
+${DEFINE_CUDA_GPU}
+
+/* Debug */
+${DEFINE_DEBUG} /**/
+
+/* Debug */
+${DEFINE_DEBUG_MODE} /**/
+
+/* Define to dummy `main' function (if any) required to link to the Fortran
+ libraries. */
+${DEFINE_F77_DUMMY_MAIN}
+
+/* Define if F77 and FC dummy `main' functions are identical. */
+${DEFINE_FC_DUMMY_MAIN_EQ_F77}
+
+/* Define if you have a BLAS library. */
+${DEFINE_HAVE_BLAS}
+
+/* define if the Boost library is available */
+${DEFINE_HAVE_BOOST}
+
+/* define if the Boost::IOStreams library is available */
+${DEFINE_HAVE_BOOST_IOSTREAMS} /**/
+
+/* define if the Boost::PROGRAM_OPTIONS library is available */
+${DEFINE_HAVE_BOOST_PROGRAM_OPTIONS} /**/
+
+/* define if the Boost::Unit_Test_Framework library is available */
+${DEFINE_HAVE_BOOST_UNIT_TEST_FRAMEWORK} /**/
+
+/* Have clock time */
+${DEFINE_HAVE_CLOCK_GETTIME} /**/
+
+/* Define to 1 if you have the <dlfcn.h> header file. */
+${DEFINE_HAVE_DLFCN_H}
+
+/* Define if you have EIGEN library. */
+${DEFINE_HAVE_EIGEN}
+
+/* Define to 1 if you have the <Eigen/Dense> header file. */
+${DEFINE_HAVE_EIGEN_DENSE}
+
+/* Define to 1 if you have the <Eigen/LU> header file. */
+${DEFINE_HAVE_EIGEN_LU}
+
+/* Defined if you have HDF5 support */
+${DEFINE_HAVE_HDF5}
+
+/* Define to 1 if you have the <inttypes.h> header file. */
+${DEFINE_HAVE_INTTYPES_H}
+
+/* Define if you have LAPACK library */
+${DEFINE_HAVE_LAPACK}
+
+/* Define if you have LIBHILBERT library */
+${DEFINE_HAVE_LIBHILBERT}
+
+/* Have quad math lib */
+${DEFINE_HAVE_LIBQUADMATH}
+
+/* Define to 1 if you have the <memory.h> header file. */
+${DEFINE_HAVE_MEMORY_H}
+
+/* Define if you have METIS library */
+${DEFINE_HAVE_METIS}
+
+/* MPI Enabled */
+${DEFINE_HAVE_MPI}
+
+/* We have OSX */
+${DEFINE_HAVE_OSX}
+
+/* Define if you have PARMETIS library */
+${DEFINE_HAVE_PARMETIS}
+
+/* Define if you have PETSC library */
+${DEFINE_HAVE_PETSC}
+
+/* Define to 1 if you have the <stdint.h> header file. */
+${DEFINE_HAVE_STDINT_H}
+
+/* Define to 1 if you have the <stdlib.h> header file. */
+${DEFINE_HAVE_STDLIB_H}
+
+/* Define to 1 if you have the <strings.h> header file. */
+${DEFINE_HAVE_STRINGS_H}
+
+/* Define to 1 if you have the <string.h> header file. */
+${DEFINE_HAVE_STRING_H}
+
+/* Define if you have SUITESPARSE library. */
+${DEFINE_HAVE_SUITESPARSE}
+
+/* Define to 1 if you have the <sys/stat.h> header file. */
+${DEFINE_HAVE_SYS_STAT_H}
+
+/* Define to 1 if you have the <sys/types.h> header file. */
+${DEFINE_HAVE_SYS_TYPES_H}
+
+/* Define to 1 if you have the <unistd.h> header file. */
+${DEFINE_HAVE_UNISTD_H}
+
+/* Define to the sub-directory where libtool stores uninstalled libraries. */
+#define LT_OBJDIR ".libs/"
+
+/* NVCC compiling */
+${DEFINE_NVCC} /**/
+
+/* Name of package */
+#define PACKAGE "openfpm_pdata"
+
+/* Define to the address where bug reports for this package should be sent. */
+#define PACKAGE_BUGREPORT "BUG-REPORT-ADDRESS"
+
+/* Define to the full name of this package. */
+#define PACKAGE_NAME "OpenFPM_pdata"
+
+/* Define to the full name and version of this package. */
+#define PACKAGE_STRING "OpenFPM_pdata 1.0.0"
+
+/* Define to the one symbol short name of this package. */
+#define PACKAGE_TARNAME "openfpm_pdata"
+
+/* Define to the home page for this package. */
+#define PACKAGE_URL ""
+
+/* Define to the version of this package. */
+#define PACKAGE_VERSION "1.0.0"
+
+/* Test performance mode */
+${DEFINE_PERFORMANCE_TEST}
+
+/* Security enhancement class 1 */
+${DEFINE_SE_CLASS1}
+
+/* Security enhancement class 2 */
+${DEFINE_SE_CLASS2}
+
+/* Security enhancement class 3 */
+${DEFINE_SE_CLASS3}
+
+/* Define to 1 if you have the ANSI C header files. */
+${DEFINE_STDC_HEADERS}
+
+/* If an error occur stop the program */
+${DEFINE_STOP_ON_ERROR}
+
+/* Test coverage mode */
+${DEFINE_TEST_COVERAGE_MODE}
+
+/* when an error accur continue but avoid unsafe operation */
+/* #undef THROW_ON_ERROR */
+
+/* Version number of package */
+#define VERSION "1.0.0"