diff options
| author | Julian Taylor <jtaylor.debian@googlemail.com> | 2016-12-12 23:15:21 +0100 |
|---|---|---|
| committer | Julian Taylor <jtaylor.debian@googlemail.com> | 2016-12-12 23:19:06 +0100 |
| commit | 3b31fa130959bbc5544e7b5fa5226076466d6d88 (patch) | |
| tree | 316a7047ebba9686fcd66d1861036b8feffdc801 /numpy/lib/nanfunctions.py | |
| parent | 82a1d81fc63fc74d97a5ef949790f91f39f41f04 (diff) | |
| download | numpy-3b31fa130959bbc5544e7b5fa5226076466d6d88.tar.gz | |
ENH: update the small nanmedian threshold
The apply_along_axis path is significantly more expensive than currently
accounted for in the check. Increase the minimum axis size from 400 to
1000 elements.
Either apply_along_axis got more expensive over time or the original
benchmarking was flawed.
Diffstat (limited to 'numpy/lib/nanfunctions.py')
| -rw-r--r-- | numpy/lib/nanfunctions.py | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/numpy/lib/nanfunctions.py b/numpy/lib/nanfunctions.py index c024055ba..08358a030 100644 --- a/numpy/lib/nanfunctions.py +++ b/numpy/lib/nanfunctions.py @@ -869,7 +869,7 @@ def _nanmedian(a, axis=None, out=None, overwrite_input=False): else: # for small medians use sort + indexing which is still faster than # apply_along_axis - if a.shape[axis] < 400: + if a.shape[axis] < 1000: return _nanmedian_small(a, axis, out, overwrite_input) result = np.apply_along_axis(_nanmedian1d, axis, a, overwrite_input) if out is not None: |