basic benchmark shows ~order of magnitude speedup

Tests/bench_cffi_access.py

@@ -0,0 +1,43 @@
+from tester import *
+
+# not running this test by default. No DOS against travis.
+
+from PIL import PyAccess
+
+import time
+
+def iterate_get(size, access):
+    (w,h) = size
+    for x in range(w):
+        for y in range(h):
+            access[(x,y)]
+
+def iterate_set(size, access):
+    (w,h) = size
+    for x in range(w):
+        for y in range(h):
+            access[(x,y)] = access[(x,y)]
+
+def timer(func, label, *args):
+    starttime = time.time()
+    func(*args)
+    endtime = time.time()
+    print ("%s: %.4f s" %(label, endtime-starttime))
+
+def test_direct():
+    im = lena()
+    im.load()
+    caccess = im.im.pixel_access(False)
+    access = PyAccess.new(im, False)
+
+    assert_equal(caccess[(0,0)], access[(0,0)])
+
+    print ("Size: %sx%s" % im.size)
+    timer(iterate_get, 'PyAccess - get', im.size, access)
+    timer(iterate_set, 'PyAccess - set', im.size, access)
+    timer(iterate_get, 'C-api - get', im.size, caccess)
+    timer(iterate_set, 'C-api - set', im.size, caccess)
+    
+    
+
+