Fast, arbitrary floating points in Python

Question

So I'm trying to compute the mandelbrot set at very large magnifications, and have run into a bit of a brick wall.

Using pythons own float paired with a jit gave me plenty of speed.

But after having switched over to use decimal instead, my function just started returning 0 for all fields. Turning off Numba's jit solved the problem, but the performance of my software has plummetted to unusable(~30 seconds for 100 iterations over a 64x64 field with a floating point precision of 6).

My code:

def mandelbrot_point(creal, cimag, maxiter):
    real=creal
    imag=cimag
    for n in range(maxiter):
        real2=real*real
        imag2=imag*imag
        if real2+imag2>4.0:
            return n
        imag=2*real*imag+cimag
        real=real2-imag2+creal
    return 0