diff options
| author | a.kudelin | 2020-02-08 19:51:01 +0300 |
|---|---|---|
| committer | a.kudelin | 2020-02-08 19:51:01 +0300 |
| commit | 3ead38d5073c2a6875c6d670a32e1fc6a8b6b940 (patch) | |
| tree | 19d13f018edcabfbd2b929f18b03aeb2e5a8c623 /cuda.psmp | |
| parent | e78d476dcf2a90448bae410091c34360f783ed7f (diff) | |
| download | aur-3ead38d5073c2a6875c6d670a32e1fc6a8b6b940.tar.gz | |
Updated to the current major release 7.1.0, added ELPA support
Diffstat (limited to 'cuda.psmp')
| -rw-r--r-- | cuda.psmp | 60 |
1 files changed, 30 insertions, 30 deletions
diff --git a/cuda.psmp b/cuda.psmp index 4491c74537b3..a68ce090532f 100644 --- a/cuda.psmp +++ b/cuda.psmp @@ -3,51 +3,51 @@ CXX = g++ CPP = FC = mpifort LD = mpifort -NVCC = nvcc # +NVCC = nvcc GPUVER = K20X # AR = ar -r # -DFLAGS = -D__FFTW3 \ - -D__LIBINT -D__LIBINT_MAX_AM=6 -D__LIBDERIV_MAX_AM1=5 -D__MAX_CONTR=4 \ - -D__parallel -D__SCALAPACK \ - -D__LIBXSMM \ - -D__LIBXC \ - -D__SPGLIB \ +DFLAGS = -D__FFTW3 \ + -D__LIBINT \ + -D__parallel -D__SCALAPACK \ + -D__ELPA \ + -D__LIBXSMM \ + -D__LIBXC \ + -D__SPGLIB \ -D__ACC -D__DBCSR_ACC -D__PW_CUDA # -INCS = -I/usr/include \ +INCS = -I/usr/include \ + -I/usr/include/elpa_openmp-$(_elpaver)/elpa \ + -I/usr/include/elpa_openmp-$(_elpaver)/modules \ -I/opt/cuda/include # -OPTFLAGS = -O3 -march=native -fopenmp +OPTFLAGS = -O3 -march=native -fopenmp -ftree-vectorize # CFLAGS = $(DFLAGS) $(INCS) $(OPTFLAGS) CXXFLAGS = $(DFLAGS) $(INCS) $(OPTFLAGS) -std=c++11 CPPFLAGS = -FCFLAGS = $(DFLAGS) $(INCS) $(OPTFLAGS) -ffree-form -ffree-line-length-none -NVFLAGS = $(DFLAGS) -O3 -Xcompiler="-fopenmp" \ - -gencode arch=compute_30,code=sm_30 \ - -gencode arch=compute_32,code=sm_32 \ - -gencode arch=compute_35,code=sm_35 \ - -gencode arch=compute_50,code=sm_50 \ - -gencode arch=compute_52,code=sm_52 \ - -gencode arch=compute_53,code=sm_53 \ - -gencode arch=compute_60,code=sm_60 \ - -gencode arch=compute_61,code=sm_61 \ - -gencode arch=compute_62,code=sm_62 \ - -gencode arch=compute_70,code=sm_70 \ - -gencode arch=compute_72,code=sm_72 \ +FCFLAGS = $(DFLAGS) $(INCS) $(OPTFLAGS) \ + -ffree-form -ffree-line-length-none +NVFLAGS = $(DFLAGS) -O3 -Xcompiler="-fopenmp" \ + -gencode arch=compute_30,code=sm_30 \ + -gencode arch=compute_35,code=sm_35 \ + -gencode arch=compute_37,code=sm_37 \ + -gencode arch=compute_50,code=sm_50 \ + -gencode arch=compute_60,code=sm_60 \ + -gencode arch=compute_61,code=sm_61 \ + -gencode arch=compute_70,code=sm_70 \ -gencode arch=compute_75,code=sm_75 # LDFLAGS = $(FCFLAGS) # -LIBS = -L/opt/cuda/lib64 -lcudart -lcublas -lcufft -lrt -lnvrtc \ - -lxsmmf -lxsmm \ - -lscalapack -lblas -llapack \ - -lfftw3 -lfftw3_threads \ - /usr/lib/libderiv.a \ - /usr/lib/libint.a \ - -lxcf03 -lxc \ - -lsymspg \ +LIBS = -L/opt/cuda/lib64 -lcudart -lcuda \ + -lcublas -lcufft -lrt -lnvrtc \ + -lxsmmf -lxsmm \ + -lelpa_openmp -lscalapack -lblas -llapack \ + -lfftw3 -lfftw3_threads \ + /usr/lib/libint2.a \ + -lxcf03 -lxc \ + -lsymspg \ -lstdc++ -lpthread -lm -ldl -lz |