On Mon, Jan 3, 2011 at 10:52, Keith Goodman <kwgood...@gmail.com> wrote: > On Mon, Jan 3, 2011 at 7:41 AM, Erik Rigtorp <e...@rigtorp.com> wrote: >> On Mon, Jan 3, 2011 at 10:36, Keith Goodman <kwgood...@gmail.com> wrote: >>> On Mon, Jan 3, 2011 at 5:37 AM, Erik Rigtorp <e...@rigtorp.com> wrote: >>> >>>> It's only a view of the array, no copying is done. Though some >>>> operations like np.std() will copy the array, but that's more of a >>>> bug. In general It's hard to imagine any speedup gains by copying a >>>> 10GB array. >>> >>> I don't think that np.std makes a copy of the input data if the input >>> is an array. If the input is, for example, a list, then an array is >>> created. >> >> When I tried it on a big array, it tried to allocate a huge amount of >> memory. As I said it's probably a bug. > > Yes, that would be a big bug. > > np.std does have to initialize the output array. If the window size is > small compared to arr.shape[axis] then the memory taken by the output > array is of the same order as that of the input array. Could that be > what you are seeing? > >>> a = np.arange(10) > > Small window, output array shape (8,): > >>> rolling_window(a, 2) > array([[0, 1], > [1, 2], > [2, 3], > [3, 4], > [4, 5], > [5, 6], > [6, 7], > [7, 8], > [8, 9]]) > > Big window, output array shape (2,): > >>> rolling_window(a, 9) > array([[0, 1, 2, 3, 4, 5, 6, 7, 8], > [1, 2, 3, 4, 5, 6, 7, 8, 9]])
No the array was (500,2000) and i did np.std(rolling_window(a,252),-1) and it started to allocate > 2GB. _______________________________________________ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion