pv

~/pandas/asv_bench$ asv continuous -e 6c30cbecf8 upstream/master
     before     after       ratio
  [6c30cbec] [e5ed87b3]
+  390.54μs     3.23ms      8.27  stat_ops.stats_rolling_mean.time_rolling_sum
+  686.33μs     3.47ms      5.06  stat_ops.stats_rolling_mean.time_rolling_mean
+   22.32ms   110.21ms      4.94  gil.nogil_rolling_algos_fast.time_nogil_rolling_mean
+   69.84ms   305.07ms      4.37  groupby.groupby_multi.time_groupby_multi_python
+   14.66ms    57.83ms      3.95  timedelta.timedelta_convert_string.time_timedelta_convert_string
+    1.01ms     3.86ms      3.82  stat_ops.stats_rolling_mean.time_rolling_var
+   17.75ms    56.57ms      3.19  timedelta.timedelta_convert_string_seconds.time_timedelta_convert_string_seconds

pv

"""
Function approximation via adaptive sampling + using it for root finding.
"""
# Pauli Virtanen <pav@iki.fi>, 2014

from __future__ import division, print_function, absolute_import

import numpy as np
from scipy import interpolate

pv

import numpy as np

from math import sin, pi
from scipy.integrate import quad


def remainder_map(a, b, xs, xe):
    period = xe - xs
    interval = b - a
    n_periods, left = divmod(interval, period)

pv

pv

#include <stdlib.h>

#include <libqhull_r/libqhull_r.h>

static double points[] = {
           0.16577534087020684, -1.4005379279472099,
           1.3471447295999897, 0.1881148811227809,
           1.2882417449678667, -2.2184892610712645,
           105.76092199244057, -78.34866809894842,
           -1.4890379590829337, -0.19466063858976068,

pv

numpy/benchmarks$ asv --version
asv 0.3.dev1069+3f14e671
numpy/benchmarks$ time asv run -E existing --bench bench_core
· Discovering benchmarks
· Running 29 total benchmarks (1 commits * 1 environments * 29 benchmarks)
[  0.00%] ·· Building for existing-py_usr_bin_python
[  0.00%] ·· Benchmarking existing-py_usr_bin_python
[  3.45%] ··· Running bench_core.Core.time_arange_100                                                                                                                                   525.16ns
[  6.90%] ··· Running bench_core.Core.time_array_1                                                                                                                                      343.88ns
[ 10.34%] ··· Running bench_core.Core.time_array_empty                                                                                                                                  545.93ns

pv

! testdpotr_test_gh_2691.f90
!
! Direct Fortran translation of the Scipy test
! TestDpotr.test_gh_2691
!
subroutine garbage(okflag)
    implicit none
    integer, intent(out) :: okflag
    double precision, dimension(3,3) :: a, a2
    integer, parameter :: lwork = 100

pv

#
# Makefile for running LAPACK testsuite
#

FC = gfortran
FFLAGS = -O2
LDFLAGS =
LIBS = -llapack -lblas
LD = $(FC)
AR = ar

pv

$ pypy3
Python 3.5.3 (47bcad155e44, Mar 30 2018, 19:05:26)
[PyPy 5.11.0-alpha0 with GCC 7.3.1 20180303 (Red Hat 7.3.1-5)] on linux

$ OPT="-O0 -ggdb" pypy3 -mpip install numpy==1.14.2

$ cat runthis.py
import numpy as np
print(np.__file__, np.__version__)
x = np.zeros([1], dtype=int)

pv

$ make diff
...
$ make diff
make -s run "PYTHON=/home/pauli/prj/scipy/cutestrunpy/env/bin/python" "SCRIPT=cutest_slsqp.py" PYCUTEST_CACHE="/home/pauli/prj/scipy/cutestrunpy/cache/installed" 2>&1|tee run-installed.log
1.1.0
==============================================================
|  name  |  nit   |  nfev  |success |cons_ok |ok_trust_constr|
--------------------------------------------------------------
|  HS7   |   11   |   14   |   1    |   1    |   1    |
|  HS10  |   11   |   12   |   1    |   1    |   1    |