你能评论一下我的这段代码吗？
对于上下文，我正在尝试学习Cython，看看它如何为未来的用例服务，我想集成C和Python。
这个“练习”代码是这样工作的：
1.我从一个文件中读取了一长串的3d坐标，这些坐标描述了时间步长上的两个点。
1.我计算每个时间步长两点之间的欧氏距离，并将其报告为numpy数组
我使用Cython的Pure Python模式。

# computer.py

import cython
# from cython.cimports.cpython import array
# import array
import numpy as np

if cython.compiled:
    print("Yep, I'm compiled.")
    from cython.cimports.libc.math import sqrt

else:
    print("Just a lowly interpreted script.")
    from math import sqrt

@cython.boundscheck(False)
@cython.wraparound(False)
@cython.cfunc
def compute_distance_cy(x1: cython.float, y1: cython.float, z1: cython.float,
                        x2: cython.float, y2: cython.float, z2: cython.float):

    return sqrt(sum(((x1 - x2) ** 2.0, (y1 - y2) ** 2.0, (z1 - z2) ** 2.0)))

@cython.boundscheck(False)
@cython.wraparound(False)
@cython.ccall
def compute_distances_pure(points: cython.double[:, :]):

    # get the maximum dimensions of the array
    x_max: cython.size_t = points.shape[0]
    y_max: cython.size_t = points.shape[1]

    # create memoryviews of the single points
    view2d: cython.double[:, :] = points
    view1d: cython.double[:]

    # create memoryviews of the results
    result = np.zeros(x_max, dtype=np.double)
    result2dview: cython.double[:] = result

    # access the memoryview by way of our constrained indexes
    x: cython.size_t
    for x in range(x_max):
        view1d = view2d[x, :]
        result2dview[x] = compute_distance_cy(
            view1d[0], view1d[1], view1d[2], view1d[3], view1d[4], view1d[5])

    return result

这是从调用：

...

def pure_python_mode(points):
    return computer.compute_distances_pure(points)

def do_it_in_numpy(points):
    return np.sqrt((points[:, 0] - points[:, 3])**2 +
                   (points[:, 1] - points[:, 4])**2 +
                   (points[:, 2] - points[:, 5])**2)

points_ndarray = np.array(points_list, dtype=np.double)
points_distance_array_from_cython = pure_python_mode(points_ndarray)
points_distance_array_from_numpy = do_it_in_numpy(points_ndarray)

我用一个简单的 Package 器对这两种方法进行计时。在这一点上，我能够实现：

Function pure_python_mode Took 0.0439 seconds
Function do_it_in_numpy Took 0.0183 seconds

我在Cython中的表现通常比Numpy差4 - 8倍。这对于我的用例来说是可以接受的，但是我想知道是否有人可以指出我做错了什么，或者这是否和这里一样好。