Why my python function is faster than gmp multiplication?

Question

My question is on the title. I don't explain the result of this python script. I think that there is no error on the script as I have made in my C script. I don't explain why it is not reproductable in C. I made a function that's seems to be faster than gmp multiplication...

here is the python script

import gmpy2
from gmpy2 import mpz

import sys
import time

import numpy as np
from math import ceil, floor

n_tests = 200
g=4

bzero=1
czero=1
dzer=1




indice=np.array([1, 8, 11, 14, 20, 23, 26, 29, 35, 38, 41, 48, 50, 53, 60, 63, 68, 71, 74, 77, 83, 86, 89, 96, 98, 101, 108, 111, 113, 120, 123, 126, 131, 134, 137, 144, 146, 149, 156, 159, 161, 168, 171, 174, 180, 183, 186, 189, 194, 197, 204, 207, 209, 216, 219, 222, 228, 231, 234, 237, 243, 246, 249, 256])

v0=np.ones(64)
v1=indice
v2=indice**2
v3=indice**3
t2=mpz('0')
for i in range(64):
    h=v1[i]
    h2=h**2
    h3=h**3
    h=mpz(str(h))
    h2=mpz(str(h2))
    h3=mpz(str(h3))
    t2=gmpy2.add(gmpy2.add(gmpy2.add(t2,h),h2),h3)
#   t2=gmpy2.add(t2,h3)

a=256
b=256*257/2
c=np.sum((np.arange(1, (g ** g+1), 1))**2)
d=np.sum((np.arange(1, (g ** g+1), 1))**3)
t=b+c+d
X1=mpz('6864797660130609714981900799081393217269435300143305409394463459185543183397656052122559640661454554977296311391480858037121987999716643812574028291115057151')**1457
X2=mpz('531137992816767098689588206552468627329593117727031923199444138200403559860852242739162502265229285668889329486246501015346579337652707239409519978766587351943831270835393219031728127')**16711
print(t)
print(t2)

def mult_python(x1, x2):
    return gmpy2.mul(x1,x2)




def mult_david(x1, x2):
    azero=gmpy2.div(gmpy2.sub(gmpy2.mul(x1,256),t),256)
    aprime=gmpy2.mul(azero,64)
    x=gmpy2.add(aprime,t2)
    azeroprime=gmpy2.div(gmpy2.sub(gmpy2.mul(x2,256),t),256)
    aprime=gmpy2.mul(azeroprime,64)
    y=gmpy2.add(aprime,t2)
    return gmpy2.div(gmpy2.mul(x,y),4096)


mult_functions = [mult_python, mult_david]

mult_times = {mult:[] for mult in mult_functions}
mult_errors = {mult:[] for mult in mult_functions}

N_ref = mult_python(X1,X2)


for i in range(n_tests):
    for mult in mult_functions:
        start_time = time.time()
        N = mult(X1, X2)
        end_time = time.time()
        mult_times[mult] += [end_time - start_time]
        mult_errors[mult] += [gmpy2.sub(N,N_ref)]
    sys.stdout.write('\r')
    sys.stdout.write("[{:50s}] {:d}%".format('='*(int(50*(i+1)/n_tests)), int(100*(i+1)/n_tests)))
    sys.stdout.flush()


for mult in mult_functions:
    print("{:15s}: {:e}s [error={}]".format(mult.__name__,
                                            np.mean(mult_times[mult]),
                                            np.mean(mult_errors[mult])))
print("ma methode est tant  de fois plus rapide",np.mean(mult_times[mult_python])/np.mean(mult_times[mult_david]))