Solving ODE with numba

Numba Support: How do I do ...?
1

Hi, it seems like i need some help, cause i dont realy know to do.
I have some code(ode func and ode solver func):

import numpy as np
import matplotlib.pyplot as plt
from numba import njit
import numba as nb

@njit(cache = True)
def pend(t, y, b, c):
    theta, omega = y
    dydt = np.array([omega, -b*omega - c*np.sin(theta)])
    return dydt

@njit
def rungeStep(f, t, y0, tau, params):
    k1 = tau * f(t, y0, *params)
    k2 = tau * f(t + tau/2, y0 + k1 / 2, *params)
    k3 = tau * f(t + tau/2, y0 + k2 / 2, *params)
    k4 = tau * f(t + tau, y0 + k3, *params)
    return (k1 + 2 * k2 + 2 * k3 + k4) / 6

@njit(cache = True)
def integrate(f, t0, y0, tEnd, h, params):
    ys = [list(y0)]
    t = [t0]
    while t0 <= tEnd:
        y0 += rungeStep(f, t0, y0, h, params)
        t0 += h
        ys.append(list(y0))
        t.append(t0)
    return np.array(t), np.array(ys).T

args = (0.25, 5)
y0 = np.array([np.pi - 0.1, 0.0])

%timeit integrate(pend, 0, y0, 10, 1., args)

It works well, but i cant run it if i change args type from tuple to list for example. It works only for tuples.
Can you help me how can i solve this problem. And may be anyone can give me advance how to make it faster if it possible. Thank you in advance!

2

Why do you want to change the type of args from a tuple to a list? Would it be appropriate to use a NumPy array for args instead?

In general I think Numba will perform better using NumPy arrays than some of the list operations you’re using - if you can edit it so that you always use arrays instead of lists I think you should get better performance.