avalon-fsn

avalon-fsn 是一个Python的编译构造工具，能够将你的代码Cython化

使用avalon-fsn的好处

• 代码Cython化：Windows下把代码编译为pyd，Linux下把代码编译为.so，有效的保护源代码
• 获得性能提升：能够在不做任何代码级别优化的情况下，对Python代码进行性能提升

安装

pip install avalon-fsn

编译项目

avalon-fsn-build build_ext

编译完毕后，对应的文件会在./build/lib*底下

使用编译后的文件

avalon-fsn-release

执行此命令会把build目录下的编译文件替换到根目录下，仅在编译发布环境使用

##配置文件 当有定制参数的时候，可以在项目根目录下新建配置文件avalon-fsn.json

{
  "remove_models": [],
  "remove_files": []
}

配置名称	配置描述
remove_models	不参与编译的模块
remove_files	不参与编译的文件

性能对比

import time


def run():
    time_start = time.time()
    import sys

    def make_tree(depth):
        if not depth: return None, None
        depth -= 1
        return make_tree(depth), make_tree(depth)

    def check_tree(node):
        (left, right) = node
        if not left: return 1
        return 1 + check_tree(left) + check_tree(right)

    min_depth = 4
    max_depth = max(min_depth + 2, 17)
    stretch_depth = max_depth + 1

    print("stretch tree of depth %d\t check:" %
          stretch_depth, check_tree(make_tree(stretch_depth)))

    long_lived_tree = make_tree(max_depth)

    iterations = 2 ** max_depth

    for depth in range(min_depth, stretch_depth, 2):

        check = 0
        for i in range(1, iterations + 1):
            check += check_tree(make_tree(depth))

        print("%d\t trees of depth %d\t check:" % (iterations, depth), check)
        iterations //= 4

    print("long lived tree of depth %d\t check:" %
          max_depth, check_tree(long_lived_tree))

    time_end = time.time()
    print('time cost', time_end - time_start, 's')

纯Python

stretch tree of depth 18	 check: 524287
131072	 trees of depth 4	 check: 4063232
32768	 trees of depth 6	 check: 4161536
8192	 trees of depth 8	 check: 4186112
2048	 trees of depth 10	 check: 4192256
512	 trees of depth 12	 check: 4193792
128	 trees of depth 14	 check: 4194176
32	 trees of depth 16	 check: 4194272
long lived tree of depth 17	 check: 262143
time cost 11.279994249343872 s

Cython化

stretch tree of depth 18	 check: 524287
131072	 trees of depth 4	 check: 4063232
32768	 trees of depth 6	 check: 4161536
8192	 trees of depth 8	 check: 4186112
2048	 trees of depth 10	 check: 4192256
512	 trees of depth 12	 check: 4193792
128	 trees of depth 14	 check: 4194176
32	 trees of depth 16	 check: 4194272
long lived tree of depth 17	 check: 262143
time cost 1.9600331783294678 s

简单编译之后，性能直接就提升近6倍。

 

本文章转载自开源中国。