Switch on/off the USE_OMP define for using OpenMP with CPUs (473bb207) · Commits · Claudio Gheller / HPC_Imaging

Build/Makefile.leo

+5 −3

Original line number	Original line	Diff line number	Diff line
	@@ -5,10 +5,12 @@ MPICC = mpicc
	MPIC++ = mpic++		MPIC++ = mpic++

	OPTIMIZE = -O4 -fopenmp -march=native -mavx -mavx2		OPTIMIZE = -O4 -fopenmp -march=native -mavx -mavx2
			OPT_PURE_MPI = -O4 -march=native -mavx -mavx2

	OMP_GPU = -mp=multicore,gpu -gpu=cuda11.8 -gpu=cc80		OMP_GPU = -mp=multicore,gpu -gpu=cuda11.8 -gpu=cc80

	CUDA_INC = -I/leonardo/prod/spack/03/install/0.19/linux-rhel8-icelake/gcc-8.5.0/nvhpc-23.1-x5lw6edfmfuot2ipna3wseallzl4oolm/Linux_x86_64/23.1/cuda/11.8/include		CUDA_INC = -I/leonardo/prod/spack/03/install/0.19/linux-rhel8-icelake/gcc-8.5.0/nvhpc-23.1-x5lw6edfmfuot2ipna3wseallzl4oolm/Linux_x86_64/23.1/cuda/11.8/include
	CUDA_LIB = -L/leonardo/prod/spack/03/install/0.19/linux-rhel8-icelake/gcc-8.5.0/nvhpc-23.1-x5lw6edfmfuot2ipna3wseallzl4oolm/Linux_x86_64/23.1/cuda/11.8/lib64		CUDA_LIB = -L/leonardo/prod/spack/03/install/0.19/linux-rhel8-icelake/gcc-8.5.0/nvhpc-23.1-x5lw6edfmfuot2ipna3wseallzl4oolm/Linux_x86_64/23.1/cuda/11.8/lib64 -L/leonardo/prod/spack/03/install/0.19/linux-rhel8-icelake/gcc-8.5.0/nvhpc-23.1-x5lw6edfmfuot2ipna3wseallzl4oolm/Linux_x86_64/23.1/cuda/11.8/targets/x86_64-linux/lib/stubs

	FFTW_INCL=		FFTW_INCL=
	FFTW_LIB=		FFTW_LIB=
	@@ -34,10 +36,10 @@ NCCL_LIB = -L/leonardo/prod/spack/03/install/0.19/linux-rhel8-icelake/gcc-8.5.0/
	NVC = nvc		NVC = nvc
	NVC++ = nvc++		NVC++ = nvc++
	NVFLAGS = -O4 -fast -march=native $(OMP_GPU) -mavx -mavx2		NVFLAGS = -O4 -fast -march=native $(OMP_GPU) -mavx -mavx2
	NVLIB = $(CUDA_INC) $(CUDA_LIB) -lcudart		NVLIB = $(CUDA_INC) $(CUDA_LIB) -lcuda -lcudart


	NVLIB_2 = $(CUDA_INC) $(CUDA_LIB) $(MPI_INC) $(MPI_LIB) $(CUFFTMP_INC) $(CUFFTMP_LIB) $(NVSHMEM_INC) $(NVSHMEM_LIB) -lnvshmem_host -lnvshmem_device -lcudart -lcufftMp		NVLIB_2 = $(CUDA_INC) $(CUDA_LIB) $(MPI_INC) $(MPI_LIB) $(CUFFTMP_INC) $(CUFFTMP_LIB) $(NVSHMEM_INC) $(NVSHMEM_LIB) -lcuda -lnvidia-ml -lnvshmem_host -lnvshmem_device -lcudart -lcufftMp

	NVLIB_3 = $(CUDA_INC) $(CUDA_LIB) $(MPI_INC) $(MPI_LIB) $(NCCL_INC) $(NCCL_LIB) -lcudart -lnccl		NVLIB_3 = $(CUDA_INC) $(CUDA_LIB) $(MPI_INC) $(MPI_LIB) $(NCCL_INC) $(NCCL_LIB) -lcudart -lnccl

Makefile

+23 −6

Original line number	Original line	Diff line number	Diff line
	@@ -25,8 +25,6 @@ FFTW_MPI_LIB =

	CFLAGS += -I./		CFLAGS += -I./

	FLAGS=$(OPTIMIZE)

	FFTWLIBS =		FFTWLIBS =

	# ========================================================		# ========================================================
	@@ -37,7 +35,10 @@ FFTWLIBS =
	OPT += -DUSE_FFTW		OPT += -DUSE_FFTW

	# use omp-ized version of fftw routines		# use omp-ized version of fftw routines
	#OPT += -DHYBRID_FFTW		OPT += -DHYBRID_FFTW

			# switch on the OpenMP parallelization
			OPT += USE_OMP

	# write the full 3D cube of gridded visibilities and its FFT transform		# write the full 3D cube of gridded visibilities and its FFT transform
	#OPT += -DWRITE_DATA		#OPT += -DWRITE_DATA
	@@ -60,7 +61,7 @@ OPT += -DPHASE_ON
	# Gridding kernel: GAUSS, GAUSS_HI_PRECISION, KAISERBESSEL		# Gridding kernel: GAUSS, GAUSS_HI_PRECISION, KAISERBESSEL
	#OPT += -DGAUSS_HI_PRECISION		#OPT += -DGAUSS_HI_PRECISION

	#OPT += -DGAUSS		OPT += -DGAUSS

	#OPT += -DKAISERBESSEL		#OPT += -DKAISERBESSEL

	@@ -73,7 +74,7 @@ OPT += -DPHASE_ON
	#OPT += -DNVIDIA		#OPT += -DNVIDIA

	#use cuda for GPUs		#use cuda for GPUs
	OPT += -DCUDACC		#OPT += -DCUDACC

	# use GPU acceleration via OMP		# use GPU acceleration via OMP
	#OPT += -DACCOMP		#OPT += -DACCOMP
	@@ -85,7 +86,7 @@ OPT += -DCUDACC
	#OPT += -DRCCL_REDUCE		#OPT += -DRCCL_REDUCE

	# use GPU to perform FFT		# use GPU to perform FFT
	OPT += -DCUFFTMP		#OPT += -DCUFFTMP

	#support for AMD GPUs		#support for AMD GPUs
	#OPT += __HIP_PLATFORM_AMD__		#OPT += __HIP_PLATFORM_AMD__
	@@ -95,6 +96,13 @@ OPT += -DCUFFTMP

	# ========================================================		# ========================================================


			ifeq (USE_OMP,$(findstring USE_OMP,$(OPT)))
			FLAGS=$(OPTIMIZE)
			else
			FLAGS=$(OPT_PURE_MPI)
			endif

	ifeq (FITSIO,$(findstring FITSIO,$(OPT)))		ifeq (FITSIO,$(findstring FITSIO,$(OPT)))
	LIBS += -L$(FITSIO_LIB) -lcfitsio		LIBS += -L$(FITSIO_LIB) -lcfitsio
	endif		endif
	@@ -166,7 +174,11 @@ endif

	ifeq (USE_FFTW,$(findstring USE_FFTW,$(OPT)))		ifeq (USE_FFTW,$(findstring USE_FFTW,$(OPT)))
	CFLAGS += $(FFTW_MPI_INC)		CFLAGS += $(FFTW_MPI_INC)
			ifeq (HIBRYD_FFTW,$(findstring HYBRID_FFTW,$(OPT)))
	FFTWLIBS = $(FFTW_MPI_LIB) -lfftw3_omp -lfftw3_mpi -lfftw3 -lm		FFTWLIBS = $(FFTW_MPI_LIB) -lfftw3_omp -lfftw3_mpi -lfftw3 -lm
			else
			FFTWLIBS = $(FFTW_MPI_LIB) -lfftw3_mpi -lfftw3 -lm
			endif
	endif		endif

	# define rules for sources that contains GPU code		# define rules for sources that contains GPU code
	@@ -291,8 +303,13 @@ w-stacking: $(OBJ) $(DEPS) Makefile

	#$(OBJ): $(DEPS) Makefile		#$(OBJ): $(DEPS) Makefile

			ifeq (USE_OMP,$(findstring USE_OMP,$(OPT)))
	%.o: %.c $(DEPS)		%.o: %.c $(DEPS)
	$(MPICC) $(OPTIMIZE) $(OPT) -c -o $@ $< $(CFLAGS)		$(MPICC) $(OPTIMIZE) $(OPT) -c -o $@ $< $(CFLAGS)
			else
			%.o: %.c $(DEPS)
			$(MPICC) $(OPT_PURE_MPI) $(OPT) -c -o $@ $< $(CFLAGS)
			endif

	clean:		clean:
	rm -f *.o		rm -f *.o