Makefile

CXX = g++
CXXFLAGS = -Wall -g -O0 -std=c++11
# CXXFLAGS = -Wall -std=c++11
CU = nvcc
CUFLAGS = -g -G -O0 -std=c++11
# CUFLAGS = -std=c++11

CPU_OUT = cpu.out
CUDA_OUT = cuda.out
TEST_OUT = tests.out
INFO_OUT = info.out

all: cpu cuda

clean: 
	$(RM) *.out

cpu: mlp.cc matrix.cc
	$(CXX) $(CXXFLAGS) -o $(CPU_OUT) mlp.cc matrix.cc

cuda: mlp.cc matrix.cu
	$(CU) $(CUFLAGS) -o $(CUDA_OUT) mlp.cc matrix.cu

cpu_tests: tests.cc
	$(CXX) $(CXXFLAGS) -o $(TEST_OUT) tests.cc matrix.cc

cuda_tests: tests.cc
	$(CU) $(CUFLAGS) -o $(TEST_OUT) tests.cc matrix.cu

run_all:
	$(MAKE) run_cpu
	$(MAKE) run_cuda

run_cpu:
	$(MAKE) clean
	$(MAKE) cpu
	$(PWD)/$(CPU_OUT)

run_cuda:
	$(MAKE) clean
	$(MAKE) cuda
	$(PWD)/$(CUDA_OUT)

run_cpu_tests:
	$(MAKE) clean
	$(MAKE) cpu_tests
	$(PWD)/${TEST_OUT}

run_cuda_tests:
	$(MAKE) clean
	$(MAKE) cuda_tests
	$(PWD)/${TEST_OUT}

run_cpu_profile:
	$(MAKE) clean
	$(CXX) $(CXXFLAGS) -pg -o $(CPU_OUT) mlp.cc matrix.cc
	$(PWD)/$(CPU_OUT)
	gprof -p -l -b -a $(PWD)/$(CPU_OUT)

run_cuda_profile:
	$(MAKE) clean
	$(CU) $(CUFLAGS) -o $(CUDA_OUT) mlp.cc matrix.cu
	nvprof \
	--cpu-profiling on \
	--cpu-profiling-mode flat \
	--cpu-profiling-show-library on \
	$(PWD)/$(CUDA_OUT)

run_info:
	$(CU) $(CUFLAGS) -o $(INFO_OUT) info.cu
	$(PWD)/$(INFO_OUT)

.SILENT: