adding convolution benchmarking dir

benchmark/convolution/conv2d.py

+import sys, timeit, time
+import numpy
+import theano, theano.tensor.signal.conv
+
+try:
+    img_shape =  int(sys.argv[1]), int(sys.argv[2])
+    ker_shape =  int(sys.argv[3]), int(sys.argv[4])
+    dtype = sys.argv[5]
+except:
+    print >> sys.stderr, "Usage: %s <img rows> <img cols> <ker rows> <ker cols> <dtype>" % sys.argv[0]
+    sys.exit(-1)
+
+img = theano.shared(numpy.ones(img_shape, dtype=dtype))
+ker = theano.shared(numpy.ones(ker_shape, dtype=dtype))
+out = theano.shared(numpy.ones((2,2,2), dtype=dtype))
+
+f = theano.function([], theano.tensor.signal.conv.conv2d(img, ker))
+T = timeit.Timer(f)
+print min(T.repeat(repeat=3, number=1)), 'without shape'
+
+f = theano.function([], [], updates={out:theano.tensor.signal.conv.conv2d(img,
+    ker,image_shape=img_shape,filter_shape=ker_shape)})
+T = timeit.Timer(f)
+print min(T.repeat(repeat=3, number=1)), 'with shape'

benchmark/convolution/opencv.py

+import sys, timeit, time
+import numpy
+import scikits.image.opencv
+
+try:
+    img_shape =  int(sys.argv[1]), int(sys.argv[2])
+    ker_shape =  int(sys.argv[3]), int(sys.argv[4])
+    dtype = sys.argv[5]
+except:
+    print >> sys.stderr, "Usage: %s <img rows> <img cols> <ker rows> <ker cols> <dtype>" % sys.argv[0]
+    sys.exit(-1)
+
+img = numpy.ones(img_shape, dtype=dtype)
+ker = numpy.ones(ker_shape, dtype=dtype)
+
+def f():
+    scikits.image.opencv.cvFilter2D(img, ker)
+
+T = timeit.Timer(f)
+print min(T.repeat(repeat=3, number=1))
+