bs_erf_numba_numpy.py

# Copyright (C) 2017-2018 Intel Corporation
#
# SPDX-License-Identifier: MIT


import base_bs_erf
import numba as nb
import numpy as np
from numpy import log, exp, sqrt
from math import erf

# Numba does know erf function from numpy or scipy
@nb.vectorize(nopython=True)
def nberf(x):
    return erf(x)

@nb.jit(nopython=True, parallel=True)
def black_scholes( nopt, price, strike, t, rate, vol, call, put ):
	mr = -rate
	sig_sig_two = vol * vol * 2

	P = price
	S = strike
	T = t

	a = log(P / S)
	b = T * mr

	z = T * sig_sig_two
	c = 0.25 * z
	y = 1./sqrt(z)

	w1 = (a - b + c) * y
	w2 = (a - b - c) * y

	d1 = 0.5 + 0.5 * nberf(w1)
	d2 = 0.5 + 0.5 * nberf(w2)

	Se = exp(b) * S

	r =  P * d1 - Se * d2
	call[:] = r  # temporary `r` is necessary for faster `put` computation
	put[:] = r - P + Se

base_bs_erf.run("Numba@jit-numpy", black_scholes, nparr=True, pass_args=True)