diff --git a/CHANGELOG.md b/CHANGELOG.md
index 1c17bcb16955bef021402290af06bd44677e9799..dc9e3d003f016c03694a83f0766517a18708aaa4 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -1,6 +1,12 @@
# Change Log
All notable changes to this project will be documented in this file.
+## [4.1.0] (Codename Thesis)
+- On a general base the code should not use CUDA_ON_CPU but if it does CUDA_ON_CPU macro now cover both SEQUENTIAL and OpenMP backend. The macros CUDIFY_USE_CUDA,CUDIFY_USE_HIP,CUDIFY_USE_OPENMP,CUDIFY_USE_SEQUENTIAL,CUDIFY_USE_NONE can be checked to control which CUDA backend is used
+
+### Fixed
+- Minors bug
+
## [4.0.0] September 2021 (Codename Heisenberg)
- Adding DCPSE, Level-set based numerics (Closest-point)
diff --git a/CMakeLists.txt b/CMakeLists.txt
index e24e7b719f27c4421ab4ab179aa074cca6227ea2..b84176689a16400607dab98cace5868d796a854a 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -39,22 +39,17 @@ set(METIS_DIR ${METIS_ROOT})
set(PARMETIS_DIR ${PARMETIS_ROOT})
set(OPENBLAS_ROOT CACHE PATH "Root path for blas library")
set(SuiteSparse_ROOT ${SUITESPARSE_ROOT})
-set(CUDA_ON_CPU CACHE BOOL "Make Cuda work on heap")
set(CPACK_RUN_INSTALL_DEPENDENCIES CACHE BOOL "Set to true if we are creating deb or RPM packages")
set(ENABLE_GARBAGE_INJECTOR CACHE BOOL "Enable the injector of garbage in the memory allocator")
set(ENABLE_VCLUSTER_GARBAGE_INJECTOR CACHE BOOL "Enable the injector of garbage in the vcluster memory buffers")
set(HIP_ENABLE CACHE BOOL "Enable HIP compiler")
set(AMD_ARCH_COMPILE "gfx900" CACHE STRING "AMD gpu architecture used to compile kernels")
-# Enabling real GPU is stronger than using CUDA_ON_CPU
-if (ENABLE_GPU)
- set(CUDA_ON_CPU OFF)
- # Test CLang
- if (NOT HIP_ENABLE)
- find_package(CUDA)
- set(CMAKE_CUDA_COMPILER_LIBRARY_ROOT ${CUDA_TOOLKIT_ROOT_DIR})
- enable_language(CUDA)
- endif()
+add_subdirectory (openfpm_io)
+
+if (CUDA_ON_BACKEND STREQUAL "CUDA")
+ enable_language(CUDA)
+ find_package(CUDA)
endif()
set (CMAKE_CXX_STANDARD 14)
@@ -168,6 +163,10 @@ else()
message( FATAL_ERROR "ParMetis is required in order to install OpenFPM")
endif()
+if (OPENMP_FOUND)
+ set(DEFINE_HAVE_OPENMP "#define HAVE_OPENMP")
+endif()
+
if(MPI_FOUND)
get_filename_component(OPENFPM_MPI_DEP "${MPI_C_INCLUDE_DIRS}" DIRECTORY)
@@ -180,32 +179,17 @@ else()
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")
- if (Boost_CONTEXT_FOUND)
- set(DEFINE_HAVE_BOOST_CONTEXT "#define HAVE_BOOST_CONTEXT")
- set(OPTIONAL_BOOST_LIBS "-lboost_context")
- else()
- #if context is not there CUDA_ON_CPU cannot be activated
- set(CUDA_ON_CPU OFF)
- endif()
- if (Boost_FIBER_FOUND)
- set(DEFINE_HAVE_BOOST_FIBER "#define HAVE_BOOST_FIBER")
- string(CONCAT OPTIONAL_BOOST_LIBS ${OPTIONAL_BOOST_LIBS} " -lboost_fiber")
- endif()
file(WRITE optional_boost_libs "${OPTIONAL_BOOST_LIBS}")
else()
file(WRITE error_code "202")
message( FATAL_ERROR "BOOST is required in order to install OpenFPM" )
endif()
-if(ENABLE_GPU AND CUDA_FOUND)
- set(DEFINE_CUDA_GPU "#define CUDA_GPU")
+if (OPENMP_FOUND)
+ file(WRITE openmp_flags ${OpenMP_CXX_FLAGS})
endif()
-if(CUDA_ON_CPU)
+if(ENABLE_GPU AND CUDA_FOUND)
set(DEFINE_CUDA_GPU "#define CUDA_GPU")
endif()
@@ -264,8 +248,8 @@ endif()
file(WRITE error_code "0")
file(WRITE cuda_lib "${CUDA_cudart_static_LIBRARY} ${CUDA_cudadevrt_LIBRARY}")
-if(CUDA_ON_CPU)
- file(WRITE cuda_include "-I${CUDA_INCLUDE_DIRS} -D__NVCC__ -DCUDART_VERSION=11000 -DCUDA_ON_CPU")
+if(CUDA_ON_BACKEND STREQUAL "SEQUENTIAL" OR CUDA_ON_BACKEND STREQUAL "OpenMP")
+ file(WRITE cuda_include "-I${CUDA_INCLUDE_DIRS} -D__NVCC__ -DCUDART_VERSION=11000")
file(WRITE cuda_on_cpu "YES")
else()
file(WRITE cuda_include "-I${CUDA_INCLUDE_DIRS}")
@@ -276,7 +260,6 @@ file(WRITE mpi_libs "${MPI_C_LINK_FLAGS} ${MPI_C_LIBRARIES}")
configure_file(${CMAKE_CURRENT_SOURCE_DIR}/src/config/config_cmake.h.in ${CMAKE_CURRENT_SOURCE_DIR}/src/config/config.h)
-add_subdirectory (openfpm_io)
add_subdirectory (openfpm_numerics)
file(WRITE cuda_options "${WARNING_SUPPRESSION_AND_OPTION_NVCC_TEXT}")
diff --git a/build.sh b/build.sh
index 1ded29927353de9bb87bf4073f7c7f5b034782d2..3b7ff26e8b0435a2d68bf31f4dd7876f2456f042 100755
--- a/build.sh
+++ b/build.sh
@@ -26,6 +26,7 @@ if [ x"$hostname" == x"cifarm-centos-node.mpi-cbg.de" ]; then
source /opt/rh/devtoolset-8/enable
./install_CMAKE_on_CI.sh $HOME/openfpm_dependencies/openfpm_pdata/$branch/
export PATH="$HOME/openfpm_dependencies/openfpm_pdata/$branch/CMAKE/bin:$PATH"
+ foward_options="--with-cuda-on-backend=OpenMP"
fi
if [ x"$hostname" == x"cifarm-ubuntu-node" ]; then
@@ -33,12 +34,13 @@ if [ x"$hostname" == x"cifarm-ubuntu-node" ]; then
echo "Ubuntu node"
./install_MPI_mpich.sh $HOME/openfpm_dependencies/openfpm_pdata/$branch/ 4
export PATH="/opt/bin:$PATH"
- foward_options="--enable-cuda-on-cpu"
+ foward_options="--with-cuda-on-backend=OpenMP"
fi
if [ x"$hostname" == x"cifarm-mac-node.mpi-cbg.de" ]; then
echo "Mac node"
export PATH="/usr/local/bin:$PATH"
+ foward_options="--with-cuda-on-backend=SEQUENTIAL"
# rm -rf /Users/admin/openfpm_dependencies/openfpm_pdata/$branch/
# rm -rf $HOME/openfpm_dependencies/openfpm_pdata/$branch/PETSC
#./install_CMAKE_on_CI.sh $HOME/openfpm_dependencies/openfpm_pdata/$branch/
diff --git a/configure b/configure
index 9fe5ac8164d0789b63283fe3e942c1626aa6f0fe..29474677fd8291e1521bd90ee5887a112929ccfe 100755
--- a/configure
+++ b/configure
@@ -100,6 +100,7 @@ enable_debug
with_metis
with_hdf5
with_libhilbert
+with_cuda_on_backend
enable_cuda_on_cpu
enable_scan_coverty
enable_test_performance
@@ -108,8 +109,6 @@ with_parmetis
enable_se_class1
enable_se_class2
enable_se_class3
-with_blitz
-with_algoim
with_alpaka
with_action_on_error
with_boost
@@ -124,7 +123,6 @@ with_petsc
with_eigen
with_vcdevel
enable_gpu
-enable_hip
enable_asan
enable_garbageinj
enable_garbageinjv
@@ -245,13 +243,17 @@ do
conf_options="$conf_options -DSCAN_COVERTY=ON"
;;
cuda_on_cpu)
- conf_options="$conf_options -DCUDA_ON_CPU=ON"
+ conf_options="$conf_options -DCUDA_ON_BACKEND=SEQUENTIAL"
;;
test_performance)
conf_options="$conf_options -DTEST_PERFORMANCE=ON"
;;
gpu)
- conf_options="$conf_options -DCMAKE_CUDA_HOST_COMPILER=$(which mpic++) "
+ if [ x"$CXX" == x"" ]; then
+ conf_options="$conf_options"
+ else
+ conf_options="$conf_options -DCMAKE_CUDA_HOST_COMPILER=$(which $CXX)"
+ fi
if [ x"$CXXCUDA" == x"" ]; then
conf_options="$conf_options"
else
@@ -259,10 +261,6 @@ do
fi
conf_options="$conf_options -DENABLE_GPU=ON"
;;
- hip)
- conf_options="$conf_options -DHIP_ENABLE=ON -DENABLE_GPU=ON"
- enable_hip_conf=1
- ;;
asan)
conf_options="$conf_options -DENABLE_ASAN=ON"
;;
@@ -477,6 +475,9 @@ do
ac_useropt_orig=$ac_useropt
ac_useropt=`$as_echo "$ac_useropt" | sed 's/[-+.]/_/g'`
case $ac_useropt in
+ cuda_on_backend)
+ conf_options="$conf_options -DCUDA_ON_BACKEND=$ac_optarg"
+ ;;
libhilbert)
conf_options="$conf_options -DLIBHILBERT_ROOT=$ac_optarg"
;;
@@ -514,12 +515,6 @@ do
vcdevel)
conf_options="$conf_options -DVc_ROOT=$ac_optarg"
;;
- blitz)
- conf_options="$conf_options -DBLITZ_ROOT=$ac_optarg"
- ;;
- algoim)
- conf_options="$conf_options -DALGOIM_ROOT=$ac_optarg"
- ;;
alpaka)
conf_options="$conf_options -DALPAKA_ROOT=$ac_optarg"
;;
@@ -594,10 +589,6 @@ fi
cd build
## remove enerything
-if [ x"$enable_hip_conf" != x"1" ]; then
- conf_options="$conf_options -DCMAKE_C_COMPILER=mpicc -DCMAKE_CXX_COMPILER=mpic++"
-fi
-
echo "Calling cmake ../. $conf_options"
printf "cmake ../. $conf_options" > cmake_build_options
rm ../error_code
diff --git a/example/Numerics/OdeInt/Makefile b/example/Numerics/OdeInt/Makefile
index 49784aa93f9eb62a4f0fa3fb77ccffb40999513e..6f7eb526c4fba5204d2a60a76d6fe0e6cb20c5be 100644
--- a/example/Numerics/OdeInt/Makefile
+++ b/example/Numerics/OdeInt/Makefile
@@ -8,7 +8,7 @@ OBJ = main.o
OBJ2 = main2.o
%.o: %.cpp
- $(CC) -O3 -c --std=c++11 -o $@ $< $(INCLUDE_PATH)
+ $(CC) -O3 -c --std=c++14 -o $@ $< $(INCLUDE_PATH)
example_odeint: $(OBJ)
$(CC) -o $@ $^ $(CFLAGS) $(LIBS_PATH) $(LIBS)
diff --git a/example/SparseGrid/1_gray_scott_3d_sparse_gpu/Makefile b/example/SparseGrid/1_gray_scott_3d_sparse_gpu/Makefile
index d815f8d56145b1bbe9bb064359549e4943477684..584d57a653262a160ca02024efb45045d9fbce5c 100644
--- a/example/SparseGrid/1_gray_scott_3d_sparse_gpu/Makefile
+++ b/example/SparseGrid/1_gray_scott_3d_sparse_gpu/Makefile
@@ -8,7 +8,7 @@ ifdef CUDA_ON_CPU
INCLUDE_PATH_NVCC=
CUDA_CC_LINK=mpic++
CUDA_OPTIONS=-DCUDA_ON_CPU -D__NVCC__ -DCUDART_VERSION=11000
- LIBS_SELECT=$(LIBS_CUDA_ON_CPU)
+ LIBS_SELECT=$(LIBS)
else
ifeq (, $(shell which nvcc))
CUDA_CC=mpic++ -x c++ $(INCLUDE_PATH)
diff --git a/example/SparseGrid/2_gray_scott_3d_sparse_gpu_opt/Makefile b/example/SparseGrid/2_gray_scott_3d_sparse_gpu_opt/Makefile
index 7c80e2810f9bc30ccf4df3c880e5d1a58876fca8..7e018d4c11fc04934f50cc6208ebd60e6aaed81f 100644
--- a/example/SparseGrid/2_gray_scott_3d_sparse_gpu_opt/Makefile
+++ b/example/SparseGrid/2_gray_scott_3d_sparse_gpu_opt/Makefile
@@ -8,7 +8,7 @@ ifdef CUDA_ON_CPU
INCLUDE_PATH_NVCC=
CUDA_CC_LINK=mpic++
CUDA_OPTIONS=-DCUDA_ON_CPU -D__NVCC__ -DCUDART_VERSION=11000
- LIBS_SELECT=$(LIBS_CUDA_ON_CPU)
+ LIBS_SELECT=$(LIBS)
else
ifeq (, $(shell which nvcc))
CUDA_CC=mpic++ -x c++ $(INCLUDE_PATH)
diff --git a/example/SparseGrid/3_gray_scott_3d_sparse_gpu_cs_opt/Makefile b/example/SparseGrid/3_gray_scott_3d_sparse_gpu_cs_opt/Makefile
index 15133370060405ef2c77e59de99de4d9ecd6eac5..8cb8a8b165b74e1c2813267715a7dfbc9869e8b7 100644
--- a/example/SparseGrid/3_gray_scott_3d_sparse_gpu_cs_opt/Makefile
+++ b/example/SparseGrid/3_gray_scott_3d_sparse_gpu_cs_opt/Makefile
@@ -9,7 +9,7 @@ ifdef CUDA_ON_CPU
INCLUDE_PATH_NVCC=
CUDA_CC_LINK=mpic++
CUDA_OPTIONS=-DCUDA_ON_CPU -D__NVCC__ -DCUDART_VERSION=11000
- LIBS_SELECT=$(LIBS_CUDA_ON_CPU)
+ LIBS_SELECT=$(LIBS)
else
ifeq (, $(shell which nvcc))
CUDA_CC=mpic++ -x c++ $(INCLUDE_PATH)
diff --git a/example/SparseGrid/4_gray_scott_3d_sparse_surface_cs/Makefile b/example/SparseGrid/4_gray_scott_3d_sparse_surface_cs/Makefile
index 6239057a8dd94a2061672be96d85a91cb426ea7d..6f7669c147c66ceefe7dfd51d05a37133d333873 100644
--- a/example/SparseGrid/4_gray_scott_3d_sparse_surface_cs/Makefile
+++ b/example/SparseGrid/4_gray_scott_3d_sparse_surface_cs/Makefile
@@ -9,7 +9,7 @@ gray_scott_sparse_cs_surface: OPT += -DTEST_RUN
gray_scott_sparse_cs_surface: gray_scott_sparse_surface_cs
%.o: %.cpp
- $(CC) -O3 -g $(OPT) -c --std=c++11 -o $@ $< $(INCLUDE_PATH)
+ $(CC) -O3 -g $(OPT) -c --std=c++14 -o $@ $< $(INCLUDE_PATH)
gray_scott_sparse_surface_cs: $(OBJ)
$(CC) -o $@ $^ $(CFLAGS) $(LIBS_PATH) $(LIBS)
diff --git a/example/SparseGrid/7_gray_scott_3d_sparse_gpu_sphere_expanding/Makefile b/example/SparseGrid/7_gray_scott_3d_sparse_gpu_sphere_expanding/Makefile
index 15133370060405ef2c77e59de99de4d9ecd6eac5..8cb8a8b165b74e1c2813267715a7dfbc9869e8b7 100644
--- a/example/SparseGrid/7_gray_scott_3d_sparse_gpu_sphere_expanding/Makefile
+++ b/example/SparseGrid/7_gray_scott_3d_sparse_gpu_sphere_expanding/Makefile
@@ -9,7 +9,7 @@ ifdef CUDA_ON_CPU
INCLUDE_PATH_NVCC=
CUDA_CC_LINK=mpic++
CUDA_OPTIONS=-DCUDA_ON_CPU -D__NVCC__ -DCUDART_VERSION=11000
- LIBS_SELECT=$(LIBS_CUDA_ON_CPU)
+ LIBS_SELECT=$(LIBS)
else
ifeq (, $(shell which nvcc))
CUDA_CC=mpic++ -x c++ $(INCLUDE_PATH)
diff --git a/example/SparseGrid/8_filling_benchmark/Makefile b/example/SparseGrid/8_filling_benchmark/Makefile
index 94629ac16d26f46dee525a56e3efa9159628326c..26f3b8401d9f68bcddc697905a274dc62904461c 100644
--- a/example/SparseGrid/8_filling_benchmark/Makefile
+++ b/example/SparseGrid/8_filling_benchmark/Makefile
@@ -9,7 +9,7 @@ gray_scott_sparse_opt_test: OPT += -DTEST_RUN
gray_scott_sparse_opt_test: gray_scott_sparse_opt
%.o: %.cpp
- $(CC) -mavx -O3 -g $(OPT) -c --std=c++11 -o $@ $< $(INCLUDE_PATH)
+ $(CC) -mavx -O3 -g $(OPT) -c --std=c++14 -o $@ $< $(INCLUDE_PATH)
gray_scott_sparse_opt: $(OBJ)
$(CC) -o $@ $^ $(CFLAGS) $(LIBS_PATH) $(LIBS)
diff --git a/example/SparseGrid/8_filling_benchmark_gpu/Makefile b/example/SparseGrid/8_filling_benchmark_gpu/Makefile
index c6a4b43bbd6b2846da27e6f3b042aa403d726697..872e123da5369f47a8643425bb40ad2c5936a6b3 100644
--- a/example/SparseGrid/8_filling_benchmark_gpu/Makefile
+++ b/example/SparseGrid/8_filling_benchmark_gpu/Makefile
@@ -8,7 +8,7 @@ ifdef CUDA_ON_CPU
INCLUDE_PATH_NVCC=
CUDA_CC_LINK=mpic++
CUDA_OPTIONS=-DCUDA_ON_CPU -D__NVCC__ -DCUDART_VERSION=11000
- LIBS_SELECT=$(LIBS_CUDA_ON_CPU)
+ LIBS_SELECT=$(LIBS)
else
ifeq (, $(shell which nvcc))
CUDA_CC=mpic++ -x c++ $(INCLUDE_PATH)
diff --git a/gdbgui b/gdbgui
index d39f0e88c48e33fc3fc5aa9f98171d4eae4914cb..c8048e7898e3a20d1ad1749d82d750e743247d8a 160000
--- a/gdbgui
+++ b/gdbgui
@@ -1 +1 @@
-Subproject commit d39f0e88c48e33fc3fc5aa9f98171d4eae4914cb
+Subproject commit c8048e7898e3a20d1ad1749d82d750e743247d8a
diff --git a/install b/install
index cc97a218fbeeccf2c5a19601554dc9a4f8c8dfb6..4e078ee175e20359a51a9c2003d551259e272023 100755
--- a/install
+++ b/install
@@ -297,7 +297,7 @@ configure_options=" $configure_options --with-parmetis=$i_dir/PARMETIS "
configure_options=" $configure_options --with-metis=$i_dir/METIS "
if [ x"$gpu_support" == x"1" ]; then
- configure_options=" $configure_options --enable-gpu "
+ configure_options=" $configure_options --with-cuda-on-backend=CUDA"
fi
MPI_installed=0
diff --git a/openfpm_data b/openfpm_data
index 98ea2a2f1d55516f99189ba4c7d10ebfd3186665..e63938e7f2c2a91c62039aed9bee60084df97290 160000
--- a/openfpm_data
+++ b/openfpm_data
@@ -1 +1 @@
-Subproject commit 98ea2a2f1d55516f99189ba4c7d10ebfd3186665
+Subproject commit e63938e7f2c2a91c62039aed9bee60084df97290
diff --git a/openfpm_devices b/openfpm_devices
index 162b8cabf2c0e91a7cb9a4d19e9d6784b26c4661..853f020c131a8ada56a88439b4b1345648dd65db 160000
--- a/openfpm_devices
+++ b/openfpm_devices
@@ -1 +1 @@
-Subproject commit 162b8cabf2c0e91a7cb9a4d19e9d6784b26c4661
+Subproject commit 853f020c131a8ada56a88439b4b1345648dd65db
diff --git a/openfpm_io b/openfpm_io
index 49900f22654dff41bd2698a2a1f37ba6b2d05c49..6142d9418c872c2a487e23fcbc57aade3e44fb1e 160000
--- a/openfpm_io
+++ b/openfpm_io
@@ -1 +1 @@
-Subproject commit 49900f22654dff41bd2698a2a1f37ba6b2d05c49
+Subproject commit 6142d9418c872c2a487e23fcbc57aade3e44fb1e
diff --git a/openfpm_numerics b/openfpm_numerics
index a6d5855baa64f55bc90d95922f74c88927c141d9..a0073269fa164fba2ca246540d5b9ee29d3659d7 160000
--- a/openfpm_numerics
+++ b/openfpm_numerics
@@ -1 +1 @@
-Subproject commit a6d5855baa64f55bc90d95922f74c88927c141d9
+Subproject commit a0073269fa164fba2ca246540d5b9ee29d3659d7
diff --git a/openfpm_vcluster b/openfpm_vcluster
index 49fe0597f9521146ae907c7f016f20f5326fc1bf..2102684c8d2cbfd71be5bf797ac36954984c35c2 160000
--- a/openfpm_vcluster
+++ b/openfpm_vcluster
@@ -1 +1 @@
-Subproject commit 49fe0597f9521146ae907c7f016f20f5326fc1bf
+Subproject commit 2102684c8d2cbfd71be5bf797ac36954984c35c2
diff --git a/script/help b/script/help
index 8ef53c443a0d832020cc14c62d1e0b50b92c0739..71339d236856bca21e07cb0f2755c173fdb1d19b 100755
--- a/script/help
+++ b/script/help
@@ -7,11 +7,13 @@ function HELP {
echo -e " \033[1mOptions:\033[0m"
echo -e " \033[1;34m-n\033[0m upgrade openfpm, it upgrades openfpm to the new version"
echo -e " \033[1;34m-u\033[0m update openfpm, it updates the current openfpm release with the latest bug fixes"
- echo -e " \033[1;34m-g\033[0m enable GPU"
+ echo -e " \033[1;34m-g\033[0m enable GPU with CUDA backend"
echo -e " \033[1;34m-i\033[0m directory Set the directory where to install the required dependencies"
echo -e " \033[1;34m-d\033[0m disable installation of requirements"
echo -e " \033[1;34m-s\033[0m skip user input"
- echo -e " \033[1;34m-c\033[0m foward this options to configure"
+ echo -e " \033[1;34m-c\033[0m foward this options to configure, possible options:"
+ echo -e " --prefix=/where/to/install"
+ echo -e " --with-cuda-on-backend=CUDA,SEQUENTIAL,OpenMP,HIP (equivalent CUDA is equivalent to -g)"
echo -e " \033[1;34m-h\033[0m Displays this help message"\\n
echo -e " \033[1;34m-m\033[0m Skip compilation"\\n
echo -e " \033[1;34m-l\033[0m do not install linear algebra"\\n
diff --git a/script/install_BOOST.sh b/script/install_BOOST.sh
index 0deb5de824058a09eb16661732b084b36848bdde..d080b347e8710c4caa5098d820ec6b15796827d0 100755
--- a/script/install_BOOST.sh
+++ b/script/install_BOOST.sh
@@ -1,6 +1,6 @@
#!/bin/bash
-source script/discover_os
+source discover_os
discover_os
# check if the directory $1/BOOST exist
@@ -35,7 +35,7 @@ if [ x"$platform" == x"osx" ]; then
./b2 -j $2 install --prefix=$1/BOOST address-model=64 architecture=arm abi=aapcs binary-format=mach-o toolset=$3
fi
else
- ./b2 -j $2 install --prefix=$1/BOOST
+ ./b2 -j $2 install --prefix=$1/BOOST address-model=64 architecture=x86 abi=sysv binary-format=mach-o toolset=clang
fi
else
./b2 -j $2 install --prefix=$1/BOOST
diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
index eee5aed20acdc3480b6526182051075b789c9e3e..07b37e322e0f3f027fd625fee14669e9f27c5a9f 100644
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@@ -5,7 +5,7 @@ add_definitions(-DSCAN_WITH_CUB)
########################### Executables
-if(CUDA_FOUND OR CUDA_ON_CPU OR HIP_FOUND)
+if(NOT CUDA_ON_BACKEND STREQUAL "None")
set(CUDA_SOURCES
Grid/tests/sgrid_dist_id_gpu_unit_tests.cu
Vector/cuda/vector_dist_gpu_MP_tests.cu
@@ -21,10 +21,10 @@ if(CMAKE_CXX_COMPILER_ID STREQUAL "Intel")
add_definitions("-DBOOST_MPL_CFG_HAS_TYPEOF")
endif()
-if (CUDA_ON_CPU)
- add_definitions(-DCUDA_ON_CPU)
+
+if (CUDA_ON_BACKEND STREQUAL "OpenMP" OR CUDA_ON_BACKEND STREQUAL "SEQUENTIAL")
+ set_source_files_properties(${CUDA_SOURCES} PROPERTIES COMPILE_FLAGS "-D__NVCC__ -DCUDART_VERSION=11000")
set_source_files_properties(${CUDA_SOURCES} PROPERTIES LANGUAGE CXX)
- set_source_files_properties(${CUDA_SOURCES} PROPERTIES COMPILE_FLAGS "-D__NVCC__ -DCUDART_VERSION=11000")
if(CMAKE_CXX_COMPILER_ID STREQUAL "GNU" OR CMAKE_CXX_COMPILER_ID STREQUAL "AppleClang" OR CMAKE_CXX_COMPILER_ID STREQUAL "Clang" OR CMAKE_CXX_COMPILER_ID STREQUAL "Intel")
add_definitions("-x c++")
endif()
@@ -196,6 +196,9 @@ target_link_libraries(pdata ${MPI_C_LIBRARIES})
target_link_libraries(pdata ${MPI_CXX_LIBRARIES})
target_link_libraries(pdata vcluster)
target_link_libraries(pdata ofpmmemory)
+if (OpenMP::OpenMP_CXX)
+ target_link_libraries(pdata OpenMP::OpenMP_CXX)
+endif()
#target_link_libraries(isolation_pdata ${Boost_LIBRARIES})
#target_link_libraries(isolation_pdata ${PARMETIS_LIBRARIES})
diff --git a/src/Vector/cuda/vector_dist_gpu_unit_tests.cu b/src/Vector/cuda/vector_dist_gpu_unit_tests.cu
index 9380e1e72cc5c1ca6b32525735cc374f0d623af7..0dcaed75000e430846baca3b394ee344803ee2c9 100644
--- a/src/Vector/cuda/vector_dist_gpu_unit_tests.cu
+++ b/src/Vector/cuda/vector_dist_gpu_unit_tests.cu
@@ -1857,6 +1857,10 @@ BOOST_AUTO_TEST_CASE(vector_dist_overflow_se_class1)
BOOST_AUTO_TEST_CASE( vector_dist_ghost_put_gpu )
{
+
+// This example require float atomic, until C++20 we are out of luck
+#ifndef CUDIFY_USE_OPENMP
+
Vcluster<> & v_cl = create_vcluster();
long int k = 25*25*25*create_vcluster().getProcessingUnits();
@@ -2069,6 +2073,8 @@ BOOST_AUTO_TEST_CASE( vector_dist_ghost_put_gpu )
BOOST_REQUIRE_EQUAL(ret,true);
}
}
+
+#endif
}
BOOST_AUTO_TEST_SUITE_END()
diff --git a/src/config/config_cmake.h.in b/src/config/config_cmake.h.in
index ef3eb90ec3ed96273f9a864c41a8a404171b104e..2637a83154496169f06d34f2864921938c87ac68 100644
--- a/src/config/config_cmake.h.in
+++ b/src/config/config_cmake.h.in
@@ -4,13 +4,18 @@ ${DEFINE_COVERTY_SCAN}
/* GPU support */
${DEFINE_CUDA_GPU}
+/* Define CUDIFY backend */
+${DEFINE_CUDIFY_BACKEND}
+
+/* OpenMP support */
+${DEFINE_HAVE_OPENMP}
+
/* HIP GPU support */
${DEFINE_HIP_GPU}
/* HIP Cudify GPU support */
${DEFINE_CUDIFY_USE_HIP}
-
/* Debug */
${DEFINE_DEBUG} /**/
@@ -184,9 +189,6 @@ ${DEFINE_STOP_ON_ERROR}
/* Garbage injector*/
${DEFINE_GARBAGE_INJECTOR}
-/* VCluster Garbage injector*/
-${DEFINE_VCLUSTER_GARBAGE_INJECTOR}
-
/* Test coverage mode */
${DEFINE_TEST_COVERAGE_MODE}