diff --git a/README.md b/README.md
index 378ffe6..3d9b878 100644
--- a/README.md
+++ b/README.md
@@ -133,15 +133,15 @@ N = 1024*1024*32
 M = 4
 x = np.arange(N, dtype='f8')
 
-def run():
+def some_numpy_computation():
     total = 0
     for i in range(M):
         total += x.sum()
     return total
 
 
-thread1 = threading.Thread(target=run)
-thread2 = threading.Thread(target=run)
+thread1 = threading.Thread(target=some_numpy_computation)
+thread2 = threading.Thread(target=some_numpy_computation)
 
 def main(args=None):
     thread1.start()
diff --git a/per4m/example2.py b/per4m/example2.py
index 33c3bcd..8c88378 100644
--- a/per4m/example2.py
+++ b/per4m/example2.py
@@ -1,21 +1,35 @@
 # same as example1, but without explicit viztracer calls
 import threading
 import time
-import time
 
+# if we don't run with gil_load, we just skip it
+import gil_load
+try:
+    gil_load.init()
+    gil_load.start()
+    use_gil_load = True
+except RuntimeError:
+    use_gil_load = False
 
-def run():
+def some_computation():
     total = 0
     for i in range(1_000_000):
         total += i
     return total
 
 
-thread1 = threading.Thread(target=run)
-thread2 = threading.Thread(target=run)
+thread1 = threading.Thread(target=some_computation)
+thread2 = threading.Thread(target=some_computation)
 def main(args=None):
     thread1.start()
     thread2.start()
     time.sleep(0.2)
     for thread in [thread1, thread2]:
         thread.join()
+    if use_gil_load:
+        gil_load.stop()
+        stats = gil_load.get()
+        print(gil_load.format(stats))
+
+if __name__ == "__main__":
+    main()
diff --git a/per4m/example3.py b/per4m/example3.py
index 0975979..c206650 100644
--- a/per4m/example3.py
+++ b/per4m/example3.py
@@ -7,6 +7,7 @@ import numpy as np
 import gil_load
 try:
     gil_load.init()
+    gil_load.start()
     use_gil_load = True
 except RuntimeError:
     use_gil_load = False
@@ -16,24 +17,23 @@ N = 1024*1024*32
 M = 4
 x = np.arange(N, dtype='f8')
 
-def run():
+
+def some_numpy_computation():
     total = 0
     for i in range(M):
         total += x.sum()
     return total
 
 
-if use_gil_load:
-    gil_load.start()
+thread1 = threading.Thread(target=some_numpy_computation)
+thread2 = threading.Thread(target=some_numpy_computation)
 
-thread1 = threading.Thread(target=run)
-thread2 = threading.Thread(target=run)
 
 def main(args=None):
     thread1.start()
     thread2.start()
     total = 0
-    for i in range(1_000_000):
+    for i in range(2_000_000):
         total += i
     for thread in [thread1, thread2]:
         thread.join()
@@ -42,3 +42,5 @@ def main(args=None):
         gil_load.stop()
         stats = gil_load.get()
         print(gil_load.format(stats))
+if __name__ == "__main__":
+    main()