diff --git a/gpu/test/Makefile b/gpu/test/Makefile
index 30204adaf2..6836314810 100644
--- a/gpu/test/Makefile
+++ b/gpu/test/Makefile
@@ -9,8 +9,10 @@ TESTS_SRC = TestGpuIndexFlat.cpp TestGpuIndexIVFPQ.cpp \
 TestGpuIndexBinaryFlat.cpp TestGpuIndexIVFFlat.cpp TestGpuMemoryException.cpp
 CUDA_TESTS_SRC = TestGpuSelect.cu
 
-TESTS_BIN = $(TESTS_SRC:.cpp=)
-CUDA_TESTS_BIN = $(CUDA_TESTS_SRC:.cu=)
+TESTS_OBJ = $(TESTS_SRC:.cpp=.o)
+CUDA_TESTS_OBJ = $(CUDA_TESTS_SRC:.cu=.o)
+
+TESTS_BIN = $(TESTS_OBJ:.o=) $(CUDA_TESTS_OBJ:.o=)
 
 
 # test_gpu_index.py test_pytorch_faiss.py
@@ -18,13 +20,13 @@ CUDA_TESTS_BIN = $(CUDA_TESTS_SRC:.cu=)
 run: $(TESTS_BIN) $(CUDA_TESTS_BIN)
 	for t in $(TESTS_BIN) $(CUDA_TESTS_BIN); do ./$$t || exit; done
 
-$(CUDA_TESTS_SRC:.cu=.o): %.o: %.cu gtest
+$(CUDA_TESTS_OBJ): %.o: %.cu gtest
 	$(NVCC) $(NVCCFLAGS) -g -O3 -o $@ -c $< -Igtest/include
 
-$(CUDA_TESTS_SRC:.cpp=.o): %.o: %.cpp gtest
+$(TESTS_OBJ): %.o: %.cpp gtest
 	$(CXX) $(CPPFLAGS) $(CXXFLAGS) $(CPUFLAGS) -o $@ -c $< -Igtest/include
 
-%: %.o TestUtils.o ../../libfaiss.a gtest/make/gtest.a
+$(TESTS_BIN): %: %.o TestUtils.o ../../libfaiss.a gtest/make/gtest.a
 	$(CXX) -o $@ $^ $(LDFLAGS) $(LIBS)
 
 demo_ivfpq_indexing_gpu: demo_ivfpq_indexing_gpu.o ../../libfaiss.a
@@ -46,4 +48,4 @@ clean:
 	rm -rf gtest
 	rm -f demo_ivfpq_indexing_gpu
 
-.PHONY: clean
+.PHONY: clean run