PyPy - FD - nsp - 150

from numpy import arange
from FuncDesigner import *
from openopt import NSP

y = oovar('y')

N = 150
x = [oovar() for i in range(N)]
koeffs = 1.07 ** arange(1, N+1) #

objective = sum([abs(x[i]) * koeffs[i] for i in range(N)]) + abs(y-15) + abs(y+15) + y**2

startPoint = {y:80}
for i in range(N):
    startPoint[x[i]] = cos(1+i)

p = NSP(objective, startPoint, maxIter = 1e5)

r = p.solve('ralg')

'''
PyPy 1.8 time: 118 sec
CPython 2.7 time: 267 sec
for nVars ~1-40 performance is almost same,
for nVars 50-150 PyPy performance is 1.5-2 times better
for vectorized problems
x = oovar()
objective = sum(abs(x) * koeffs) + abs(y-15) + abs(y+15) + y**2
startPoint = {x: cos(1+arange(N)), y:80}
PyPy performance is some percents slower than CPython
'''