diff --git a/OpenMP/linear-algebra/kernels/atax/Makefile b/OpenMP/linear-algebra/kernels/atax/Makefile
index c82dd7b..a047a1d 100644
--- a/OpenMP/linear-algebra/kernels/atax/Makefile
+++ b/OpenMP/linear-algebra/kernels/atax/Makefile
@@ -1,3 +1,6 @@
+# Remove all make implicit rules because this is already complex enough as it is
+MAKEFLAGS+= -r
+
 # -DPOLYBENCH_TIME makes Polybench output the execution time of the program
 CXXFLAGS+= -DPOLYBENCH_TIME
 # -O3 applies all compiler optimization, improving from 800ms to 300ms
@@ -13,19 +16,12 @@ NVCC:=$(CUDA_HOME)/bin/nvcc
 NVCFLAGS:=$(CXXFLAGS) $(NVOPT)
 
 
-INCPATHS:=-I../../../utilities
+%.elf: %.cu.o polybench.cu.o
+	$(NVCC) $(NVCFLAGS) $^ -o $@ $(LDFLAGS)
 
-
-# Create an automake definition for .cu files
-%.cu.o:
+%.cu.o: %.cu
 	$(NVCC) $(NVCFLAGS) -c $< -o $@
 
-%.exe: %.cu.o
-	$(NVCC) $(NVCFLAGS) $(INCPATHS) $^ -o $@ $(LDFLAGS)
-
-%.elf: %.cu.o
-	$(NVCC) $(NVCFLAGS) $(INCPATHS) $^ -o $@ $(LDFLAGS)
-
 
 .PHONY: bench