[pypy-issue] [issue1140] numpypy: slower than CPython NumPy on the vectorized code

[Dmitrey]

New submission from Dmitrey <dmitrey15 at ukr.net>:

for some vectorized functions PyPy is about 5 times slower, see
http://pastebin.com/W2xy0cwE . DerApproximator.py and this test file can be
downloaded from https://www.box.com/s/5034296ac9cf838db382.
Setting 
from _numpypy.pypy import set_invalidation
set_invalidation(False)
decreases PyPy time from 2.4 sec to 1.5 sec, but it's still too far from CPython
NumPy result (0.5 sec).

----------
messages: 4305
nosy: Dmitrey, pypy-issue
priority: performance bug
status: unread
title: numpypy: slower than CPython NumPy on the vectorized code

________________________________________
PyPy bug tracker <tracker at bugs.pypy.org>
<https://bugs.pypy.org/issue1140>
________________________________________