Separate benchmarch stuff into its own file.

Author: Jonathan Rocher

numexpr/compare_numexpr_np.py

@@ -0,0 +1,41 @@
+""" Comparing execution time between Numexpr and Numpy
+"""
+
+import numpy as np
+from numexpr import evaluate
+import time
+from matplotlib.pyplot import (figure, plot, show, legend, xlabel, ylabel, 
+                               yscale, title)
+
+exec_times_np = []
+exec_times_ne = []
+
+sizes = np.arange(0, 8, 0.5)
+for N in sizes:
+    a = np.arange(10**N)
+    b = np.arange(10**N)
+    start1 = time.time()
+    c1 = 2*a**3+b**2+16
+    start2 = time.time()
+    exec_times_np.append(start2-start1)
+    print "Numpy polynomial computation %s seconds" % (start2-start1)
+    start3 = time.time()
+    c2 = evaluate("2*a**3+b**2+16")
+    t4 = time.time()
+    print "Using numexpr, the polynomial computation took %s seconds" % (t4-start3)
+    exec_times_ne.append(t4-start3)
+    del a
+    del b
+    del c1
+    del c2
+
+figure()
+plot(sizes, exec_times_np, label = "Numpy")
+plot(sizes, exec_times_ne, label = "Numexpr")
+xlabel("log10(size(arrays))")
+ylabel("execution time in sec")
+yscale("log")
+title("Comparing execution time to compute 2*a**3+b**2+16")
+legend(loc="upper left")
+show()
+