Multiprocessing.pool（）：比顺序慢

Question

我试图使用python（和Numpy）通过有限元来求解导数。当我顺序解决没有问题，但for非常大，需要太长时间。然后，我尝试用共享内存实现multiprocessing，但我有一些问题。我在一个有4个内核的机器上运行上面的代码。代码中心运行良好，但顺序代码慢得多，为什么？我做得不好？

cores = multiprocessing.cpu_count()
n = 40  #array size 
deriv2_base = multiprocessing.Array(ctypes.c_double, n)
temp_base = multiprocessing.Array(ctypes.c_double, n)
deriv2 = np.ctypeslib.as_array(deriv2_base.get_obj())
temp = np.ctypeslib.as_array(temp_base.get_obj())

temp[-1] = const['T0']

def calculo(i, def_param=(deriv2, temp)):
    pos_ini = i*blockDim + 1/(1 + i)
    pos_fin = (i + 1)*blockDim - (i+1)/cores

    deriv2[pos_ini:pos_fin] = (temp[pos_ini-1:pos_fin-1] -2*temp[pos_ini:pos_fin] +   temp[pos_ini+1:pos_fin+1]) / delta_z**2
    temp[pos_ini:pos_fin] = temp[pos_ini:pos_fin] + kappa*deriv2[pos_ini:pos_fin] *    delta_t_seg

if __name__ == '__main__':
    pool = multiprocessing.Pool(processes = cores)
    for j in xrange(i_steps):
        pool.map(calculo, range(cores))
    print temp

坦克！