# Tested with: GFortran 4.8
NVCC     = /usr/local/cuda/bin/nvcc
CC         = gcc
CPP        =
FC         = mpif90 -fopenmp
LD         = mpif90 -fopenmp
AR         = ar -r
LIBBASE    = /usr/local/lib
# OBlas_INC   = $(LIBBASE)/OpenBLAS/include
# OBlas_LIB   = $(LIBBASE)/OpenBLAS/lib
FFTW_INC   = /usr/local/fftw3/include
FFTW_LIB   = /usr/local/fftw3/lib
LIBINT_INC = $(LIBBASE)/libint/include
LIBINT_LIB = $(LIBBASE)/libint/lib
LIBXC_INC  = $(LIBBASE)/libxc/include
LIBXC_LIB  = $(LIBBASE)/libxc/lib
CUDA_LIB   = /usr/local/cuda/lib64
CUDA_INC   = /usr/local/cuda/include
DFLAGS     = -D__GFORTRAN -D__parallel -D__BLACS -D__SCALAPACK -D__FFTSG -D__FFTW3 -D__LIBINT -D__LIBXC2\
             -D__ACC -D__DBCSR_ACC -D__PW_CUDA \
             -I$(FFTW_INC) -I$(LIBINT_INC) -I$(LIBXC_INC)
CPPFLAGS   =
NVFLAGS    = $(DFLAGS) -g -arch sm_20
FCFLAGS    = $(DFLAGS) -g -O3 -march=native -finline -ffree-form
LDFLAGS    = $(FCFLAGS)
LIBS       = $(LIBBASE)/libscalapack.a \
             $(LIBBASE)/liblapack.a \
             $(LIBBASE)/libfblas.a \
             $(FFTW_LIB)/libfftw3_threads.a\
             $(FFTW_LIB)/libfftw3.a\
             $(LIBXC_LIB)/libxcf90.a\
             $(LIBXC_LIB)/libxc.a\
             $(LIBINT_LIB)/libderiv.a\
             $(LIBINT_LIB)/libint.a \
             $(CUDA_LIB)/libcudart.so \
             $(CUDA_LIB)/libcublas.so \
             $(CUDA_LIB)/libcufft.so \
             -lstdc++ -lpthread