Makefile

# Compile host code (no GPU)  with these flags if
# you want to debug. Use GDB on Linux or HPC, or Mac.
# Use Windows tools if you compile C++ under windows.
# NOTE: Most of what you have in this file is not relevant.
CXXFLAGS=-g -ggdb3 -Wall -Wextra -pedantic

# Uncomment this if you want optimized host C++ executable.
# No good for debugging.
#CXXFLAGS=-O3 -Wall -Wextra -pedantic

# Linker flags to enable OpenMP on some older systems
# Unused in this file
LDFLAGS=-lgmp

# NOTE: Below you have option "-ccbin /usr/bin/g++-13"
# This points to the binary of the G++ compiler to be used
# to compile host code under NVCC (NVIDIA C/C++ Compiler)
# It is very important to match it with your
# OS and available compiler.

# Uncomment this if you want optimized code
#NVCCFLAGS=-O3 -std=c++17 -ccbin /usr/bin/g++-13

# Uncomment this if you want to debug the code
# NOTE: Use cuda-gdb as the debugger on Linux or HPC
NVCCFLAGS=-g -G -O0 -std=c++17 -ccbin /usr/bin/g++-13

PROGS = rk4 rk4_no_cuda rk4_no_cuda_omp rk4_mpi

all: $(PROGS)


# Add switched -g -G for debugging with cuda-gdb
rk4: rk4.cu
	nvcc $(NVCCFLAGS) $< -o $@

rk4_no_cuda: rk4_no_cuda.cpp
	$(CXX) $(CXXFLAGS) $< -o $@

rk4_no_cuda_omp: rk4_no_cuda_omp.cpp
	$(CXX) -fopenmp $(CXXFLAGS) $< -o $@

# Run MPI version like this:
# mpirun -np 4 ./rk4_mpi
rk4_mpi: rk4_mpi.cpp
	mpic++ -O3 -std=c++17 rk4_mpi.cpp -o rk4_mpi

run: $(PROGS)
	@(echo '==== Pure Python ===='; time python ./rk4.py >/dev/null)
	@(echo '==== C++ code / single thread ==== '; time ./rk4_no_cuda >/dev/null)
	@(echo '==== C++ code / Open MP ===='; time ./rk4_no_cuda_omp >/dev/null)
	@(echo '==== C++ code / Open MPI ===='; time mpirun -np 10 ./rk4_mpi >/dev/null)
	@(echo '==== CUDA C++ code / GPU ===='; time ./rk4 >/dev/null)

clean:
	rm $(PROGS)