Ray 简单性能对比

最新推荐文章于 2024-04-21 23:45:00 发布

小小毛毛虫~

最新推荐文章于 2024-04-21 23:45:00 发布

阅读量230

点赞数

分类专栏：开发基本知识文章标签： Ray 并行计算分布式快速排序性能优化

本文链接：https://blog.csdn.net/Lyncai/article/details/120864259

版权

开发基本知识专栏收录该内容

181 篇文章 1 订阅

订阅专栏

import time
import ray
@ray.remote
def multi(x):
    return x*x

def single_thread(x):
    return x*x

start = time.time()
parallel_result = [multi.remote(i) for i in range(100)]
diff = time.time()-start

print(diff)


start = time.time()
for i in range(100):
    temp_result = single_thread(i)
diff = time.time()-start
print(diff)

2021-10-20 学习了一篇文档收获很大共勉 Ray 1.0 架构解读

2021-10-21 快排使用ray 作分布式运行

读文档： Ray design patternhttps://docs.google.com/document/d/167rnnDFIVRhHhK4mznEIemOtj63IOhtIPvSYaPgI4Fg/edit#


import ray
MINI_SIZE=10
def Partition(collection):
    pivolt = collection.pop(0)
    less, greater = [], []
    for item in collection:
        if item < pivolt:
            less.append(item)
        else:
            greater.append(item)
    return less, pivolt, greater

def quick_sort(collection):
    if len(collection) <= MINI_SIZE and len(collection)>0:
        collection.sort()
        return collection

    if len(collection) > MINI_SIZE:
        less, pivolt, greater = Partition(collection)
        less = quick_sort(less)
        greater = quick_sort(greater)

        return less + [pivolt] + greater

    return []

def main_quick_sort():
    import numpy as np

    collection = list(np.random.randint(300, 500, 100))
    start_time = time.time()
    result = quick_sort(collection)
    diff = time.time()-start_time
    return diff


@ray.remote
def quick_sort_ray(collection):
    if len(collection) <= MINI_SIZE and len(collection)>0:
        collection.sort()
        return collection

    if len(collection) > MINI_SIZE:
        less, pivolt, greater = Partition(collection)
        less = quick_sort(less)
        greater = quick_sort(greater)

        return less + [pivolt] + greater

    return []

def main_quick_sort_ray():
    import numpy as np

    collection = list(np.random.randint(300, 500, 100))
    start_time = time.time()
    result = quick_sort_ray.remote(collection)
    diff = time.time()-start_time
    return diff





diff_single=main_quick_sort()
diff_ray = main_quick_sort_ray()
print(diff_single)
print(diff_ray)