Update benchmark: run each bench separetely to avoid bad malloc implementation bug

(Running list benchmark before array fragment the memory too much and
 generate a noticable slow down).

Add Klib bench.
o

bench/Makefile

@@ -3,15 +3,29 @@ CXX=c++ -std=c++11
 CFLAGS=-O2 -march=native -Wall
 CPPFLAGS=-I..
 
-all: bench-mlib bench-stl
+all: bench-mlib bench-stl bench-klib
 
 bench-mlib:
 	$(CC) $(CFLAGS) $(CPPFLAGS) bench-mlib.c -o bench-mlib.exe
-	./bench-mlib.exe
+	@./bench-mlib.exe 1
+	@./bench-mlib.exe 2
+	@./bench-mlib.exe 3
+	@./bench-mlib.exe 4
 
 bench-stl:
 	$(CXX) $(CFLAGS) $(CPPFLAGS) bench-stl.cpp -o bench-stl.exe
-	./bench-stl.exe
+	@./bench-stl.exe 1
+	@./bench-stl.exe 2
+	@./bench-stl.exe 3
+	@./bench-stl.exe 4
+	@./bench-stl.exe 5
+
+bench-klib:
+	$(CC) $(CFLAGS) $(CPPFLAGS) $${KLIB_CPPFLAGS} bench-klib.c -o bench-klib.exe
+	@./bench-klib.exe 1
+	@./bench-klib.exe 2
+	@./bench-klib.exe 3
+	@./bench-klib.exe 4
 
 clean:
-	rm -f *.o bench-mlib.exe bench-stl.exe *~ 
+	rm -f *.o bench-*.exe *~