python数组获取到最后一个值_python - 在numpy数组中，高效地提取每个元素的最后两位数字_python_酷徒编程知识库...

快速基准测试

安装程序import numpy as np

test_arr = np.random.randint(0, 10000000, 10000000)

1.原始解决方案def last_two_bits(arr_in):

return [bin(num)[-2:].replace('b','0') for num in arr_in]

时间：~5200 msbits_map = ['00','01','10','11']

def last_two_bits_nv(arr_in):

return bits_map[arr_in % 4]

last_two_bits = np.vectorize(last_two_bits_nv)

时间：~2600 ms

3.我对@aminrd的解决方案进行了调整bits_map = np.array(['00', '01', '10', '11'])

def last_two_bits(arr_in):

return bits_map[arr_in % 4]

时间：~170 msdef last_two_bits(arr_in):

return (((arr_in & 1) << 32) + ((arr_in & 2) >> 1) + ord('0') * 0x100000001).view('U2')

时间：~100 msbits_map = np.array(['00', '01', '10', '11'])

def last_two_bits(arr_in):

return bits_map[arr_in & 3]

时间：~60 msdef last_two_bits(arr_in):

output = np.empty((arr_in.size, 2), dtype=np.uint8)

np.bitwise_and(arr_in >> 1, 1, out=output[:, 0], casting='unsafe')

np.bitwise_and(arr_in, 1, out=output[:, 1], casting='unsafe')

output += 48

return output.view(dtype='S2').ravel()

时间：~60 ms