Measuring performance and using numpy UCL-RITS#185

week10/calc_pi_np.py

@@ -0,0 +1,56 @@
+import numpy as np
+import argparse
+import random
+from utils import format_time
+
+def point_in_circle(coord, radius=1):
+    """
+    Checks whether a point (x, y) is part of a circle with a set radius.
+
+    example
+    -------
+    >>> point_in_circle(0, 0)
+    True
+
+    """
+    r = np.sqrt(np.square(coord).sum(axis=1))
+    return r[r<radius]
+
+def calculate_pi_timeit(points):
+    """
+    Wrapper function to build calculate_pi with a particular number of points
+    and returns the function to be timed.
+    """
+    def calculate_pi():
+        """
+        Calculates an approximated value of pi by the Monte Carlo method.
+        """
+        coord = np.random.rand(points,2)
+        within_circle = point_in_circle(coord)
+        return 4 * sum(within_circle)/points
+
+    return calculate_pi
+
+
+def command():
+    """
+    entry point of the script to accept arguments
+    """
+
+    parser = argparse.ArgumentParser(description="Calculates an approximate value of PI and how long it takes",
+                                     formatter_class=argparse.ArgumentDefaultsHelpFormatter)
+    parser.add_argument('--npoints', '-np', default=10_000, type=int, help="Number of random points to use")
+    parser.add_argument('--number', '-n', default=100, type=int, help="Number of times to execute the calculations")
+    parser.add_argument('--repeat', '-r', default=5, type=int, help="How many times to repeat the timer")
+
+    arguments = parser.parse_args()
+
+    calc_pi = calculate_pi_timeit(arguments.npoints)
+    print(f"pi = {calc_pi()} (with {arguments.npoints})")
+    result = timeit.repeat(calc_pi, number=arguments.number, repeat=arguments.repeat)
+    best = min(result) / arguments.number
+    print(f"{arguments.number} loops, best of {arguments.repeat}: {format_time(best)} per loop")
+
+
+if __name__ == '__main__':
+    command()