[SciPy-User] [ANN] Bottleneck 0.5.0 released
Sat Aug 13 19:39:57 CDT 2011
On Sat, Aug 13, 2011 at 12:05 PM, Stef Mientki <firstname.lastname@example.org> wrote:
> thanks Cristoph,
> found the problem,
> I used a too old version of numpy,
> ( wouldn't it be an idea to replace line 17 in __init__.py with "print 'requires at least numpy 1.5.1"
Sounds like a good idea but unfortunately there are other reasons,
besides an old version of numpy, why the cython functions might fail
to load. For example, the compilation may have failed.
> now the fast routines are working ...
> ... at least sometimes
> ... at least on some computers
> What I've seen until now:
> Computer 1: numpy 1.4, so it uses slow routines: functional ok
> Computer 2: exactly the same python + libs: screen starts to "blink" to black a few times (for about
> half a second, with an interval about 2 seconds),
> after 10 times, the screen is filled with a repeating part of the screen and computer hangs totally.
Oh, my goodness! That is odd.
> Computer 2: numpy 1.6.1 : first program run, screen "blinks" black once, the fast bottleneck
> routines are use, and they function ok.
> Second run of the same program: screen blinks blank once, after a few seconds, the screen is again
> filled with a smaal repating part of the screen and the computer hangs totally.
> Any ideas ?
That is terrible. I have no clue as to the cause.
> Is the GPU used with these routines ?
> On 13-08-2011 17:09, Christoph Gohlke wrote:
>> On 8/13/2011 6:54 AM, Stef Mientki wrote:
>>> is it possible to create a windows executable
>>> (a lot of windows users can't compile C-code).
>>> I tried the prebuild versions from:
>>> but the fast routines are all missing there.
>> I don't see anything missing. Tests and benchmarks yield expected
>> results using numpy 1.6.1.
>> What's the output of `import bottleneck as bn;bn.test()` (requires nose
>>> On 13-06-2011 23:35, Keith Goodman wrote:
>>>> Bottleneck is a collection of fast NumPy array functions written in
>>>> Cython. It contains functions like median, nanmedian, nanargmax,
>>>> move_max, rankdata.
>>>> The fifth release of bottleneck adds four new functions, comes in a
>>>> single source distribution instead of separate 32 and 64 bit versions,
>>>> and contains bug fixes.
>>>> J. David Lee wrote the C-code implementation of the double heap moving
>>>> window median.
>>>> New functions:
>>>> - move_median(), moving window median
>>>> - partsort(), partial sort
>>>> - argpartsort()
>>>> - ss(), sum of squares, faster version of scipy.stats.ss
>>>> - Single source distribution instead of separate 32 and 64 bit versions
>>>> - nanmax and nanmin now follow Numpy 1.6 (not 1.5.1) when input is all NaN
>>>> Bug fixes:
>>>> - #14 Support python 2.5 by importing `with` statement
>>>> - #22 nanmedian wrong for particular ordering of NaN and non-NaN elements
>>>> - #26 argpartsort, nanargmin, nanargmax returned wrong dtype on 64-bit Windows
>>>> - #29 rankdata and nanrankdata crashed on 64-bit Windows
>>>> mailing list
>>>> mailing list 2
>>>> SciPy-User mailing list
>>> SciPy-User mailing list
>> SciPy-User mailing list
> SciPy-User mailing list
More information about the SciPy-User