I don't know Python but this benchmark caught my eye. >>def D4_Transform(x, s1=None, d1=None, d2=None): >> """ >> D4 Wavelet transform in NumPy >> (C) Sturla Molden >> """ >> C1 = 1.7320508075688772 >> C2 = 0.4330127018922193 >> C3 = -0.066987298107780702 >> C4 = 0.51763809020504137 >> C5 = 1.9318516525781364 >> if d1 == None: >> d1 = numpy.zeros(x.size/2) >> s1 = numpy.zeros(x.size/2) >> d2 = numpy.zeros(x.size/2)
Are these definitions ever used? It looks like s1, d1 and d2 are all redefined below without reference to these previous values. >> odd = x[1::2] >> even = x[:-1:2] >> d1[:] = odd[:] - C1*even[:] >> s1[0] = even[0] + C2*d1[0] + C3*d1[-1] >> s1[1:] = even[1:] + C2*d1[1:] + C3*d1[:-1] >> d2[0] = d1[0] + s1[-1] >> d2[1:] = d1[1:] + s1[:-1] >> even[:] = C4 * s1[:] >> odd[:] = C5 * d2[:] Does that line create an array that is never used? If so, is C5 also never used? >> if x.size > 2: >> >>D4_Transform(even,s1[0:even.size/2],d1[0:even.size/2],d2[0:even.size/2]) What is the result of this function? I'm interested in translating this function into other languages, like OCaml, to see how good Python's performance is (I am amazed it can beat Matlab, not that I've used Matlab). In particular, I think you are eagerly allocating arrays when, in a functional language, you could just as easily compose closures. For example, this program is 3x faster than the Python on my machine: let rec d4 s1 d1 d2 x = let c1 = 1.7320508075688772 in let c2 = 0.4330127018922193 in let c3 = -0.066987298107780702 in let c4 = 0.51763809020504137 in let c5 = 1.9318516525781364 in let n = Array.length x in let odd i = x.(2*i) and even i = x.(2*i + 1) in let d1 i = odd i -. c1 *. even i in let f = function -1 -> n/2 - 1 | i -> i in let s1 i = even i +. c2 *. d1 i +. c3 *. d1 (f(i-1)) in let d2 i = d1 i +. s1 (f(i-1)) in let even = Array.init (n/2) (fun i -> c4 *. s1 i) in if n > 2 then d4 s1 d1 d2 even else s1, d1, d2 but I'm not sure it is correct! -- Dr Jon D Harrop, Flying Frog Consultancy Objective CAML for Scientists http://www.ffconsultancy.com/products/ocaml_for_scientists -- http://mail.python.org/mailman/listinfo/python-list
