为什么等待异步这么慢？ [英] Why is await async so slow?
问题描述
我终于得到了VS2012，得到一个简单的演示和工作检查出的异步潜在的性能提升和等待，但我失望的是慢！其可能的，我做错了什么，但也许你能帮助我。 （我还添加了一个简单的解决方案，螺纹和运行为预期的要快）
I finally got VS2012 and got a simple demo up and working to check out the potential performance boost of async and await, but to my dismay it is slower! Its possible I'm doing something wrong, but maybe you can help me out. (I also added a simple Threaded solution, and that runs faster as expected)
我的code使用一个类来总结一个数组，根据您的系统（-1）矿山有4个内核上的内核数量，因此我看到了有关线程2倍加速（2.5线程），但一个2X放慢了同样的事情，但与异步/游览车。
My code uses a class to sum an array, based on the number of cores on your system (-1) Mine had 4 cores, so I saw about a 2x speed up (2.5 threads) for threading, but a 2x slow down for the same thing but with async/await.
code：（请注意，您将需要添加的引用 System.Management 来获得核心检测仪工作）
Code: (Note you will need to added the reference to System.Management to get the core detector working)
using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Threading.Tasks;
using System.Threading;
using System.Management;
using System.Diagnostics;
namespace AsyncSum
{
class Program
{
static string Results = "";
static void Main(string[] args)
{
Task t = Run();
t.Wait();
Console.WriteLine(Results);
Console.ReadKey();
}
static async Task Run()
{
Random random = new Random();
int[] huge = new int[1000000];
for (int i = 0; i < huge.Length; i++)
{
huge[i] = random.Next(2);
}
ArraySum summer = new ArraySum(huge);
Stopwatch sw = new Stopwatch();
sw.Restart();
long tSum = summer.Sum();
for (int i = 0; i < 100; i++)
{
tSum = summer.Sum();
}
long tticks = sw.ElapsedTicks / 100;
long aSum = await summer.SumAsync();
sw.Restart();
for (int i = 0; i < 100; i++)
{
aSum = await summer.SumAsync();
}
long aticks = sw.ElapsedTicks / 100;
long dSum = summer.SumThreaded();
sw.Restart();
for (int i = 0; i < 100; i++)
{
dSum = summer.SumThreaded();
}
long dticks = sw.ElapsedTicks / 100;
long pSum = summer.SumParallel();
sw.Restart();
for (int i = 0; i < 100; i++)
{
pSum = summer.SumParallel();
}
long pticks = sw.ElapsedTicks / 100;
Program.Results += String.Format("Regular Sum: {0} in {1} ticks\n", tSum, tticks);
Program.Results += String.Format("Async Sum: {0} in {1} ticks\n", aSum, aticks);
Program.Results += String.Format("Threaded Sum: {0} in {1} ticks\n", dSum, dticks);
Program.Results += String.Format("Parallel Sum: {0} in {1} ticks\n", pSum, pticks);
}
}
class ArraySum
{
int[] Data;
int ChunkSize = 1000;
int cores = 1;
public ArraySum(int[] data)
{
Data = data;
cores = 0;
foreach (var item in new System.Management.ManagementObjectSearcher("Select * from Win32_Processor").Get())
{
cores += int.Parse(item["NumberOfCores"].ToString());
}
cores--;
if (cores < 1) cores = 1;
ChunkSize = Data.Length / cores + 1;
}
public long Sum()
{
long sum = 0;
for (int i = 0; i < Data.Length; i++)
{
sum += Data[i];
}
return sum;
}
public async Task<long> SumAsync()
{
Task<long>[] psums = new Task<long>[cores];
for (int i = 0; i < psums.Length; i++)
{
int start = i * ChunkSize;
int end = start + ChunkSize;
psums[i] = Task.Run<long>(() =>
{
long asum = 0;
for (int a = start; a < end && a < Data.Length; a++)
{
asum += Data[a];
}
return asum;
});
}
long sum = 0;
for (int i = 0; i < psums.Length; i++)
{
sum += await psums[i];
}
return sum;
}
public long SumThreaded()
{
long sum = 0;
Thread[] threads = new Thread[cores];
long[] buckets = new long[cores];
for (int i = 0; i < cores; i++)
{
int start = i * ChunkSize;
int end = start + ChunkSize;
int bucket = i;
threads[i] = new Thread(new ThreadStart(() =>
{
long asum = 0;
for (int a = start; a < end && a < Data.Length; a++)
{
asum += Data[a];
}
buckets[bucket] = asum;
}));
threads[i].Start();
}
for (int i = 0; i < cores; i++)
{
threads[i].Join();
sum += buckets[i];
}
return sum;
}
public long SumParallel()
{
long sum = 0;
long[] buckets = new long[cores];
ParallelLoopResult lr = Parallel.For(0, cores, new Action<int>((i) =>
{
int start = i * ChunkSize;
int end = start + ChunkSize;
int bucket = i;
long asum = 0;
for (int a = start; a < end && a < Data.Length; a++)
{
asum += Data[a];
}
buckets[bucket] = asum;
}));
for (int i = 0; i < cores; i++)
{
sum += buckets[i];
}
return sum;
}
}
}
有什么想法？我做异步/等待错了吗？我会很乐意去尝试任何建议。
Any thoughts? Am I doing async/await wrong? I'll be happy to try any suggestions.
推荐答案
您基准有几个缺陷：
- 您的时间第一次运行，其中包括初始化时间（加载
类任务，JIT编译等） - 您正在使用
DateTime.Now，这是在毫秒范围计时太不精确。你需要使用秒表
- You are timing the first run which includes initialization time (loading
class Task, JIT-compilation etc.) - You are using
DateTime.Now, which is too inaccurate for timings in the millisecond range. You'll need to useStopWatch
通过固定这两个问题;我得到以下基准测试结果：
With these two issues fixed; I get the following benchmark results:
Regular Sum: 499946 in 00:00:00.0047378
Async Sum: 499946 in 00:00:00.0016994
Threaded Sum: 499946 in 00:00:00.0026898
异步体现出来的最快的解决方案，以2ms以内。
Async now comes out as the fastest solution, taking less than 2ms.
这是下一个问题：什么时间以最快的速度为2ms是极不可靠;如果一些其他进程正在使用在背景中的CPU的线程可以得到暂停长于。你应该平均的几千基准运行的结果。
This is the next problem: timing something as fast as 2ms is extremely unreliable; your thread can get paused for longer than that if some other process is using the CPU in the background. You should average the results over several thousands of benchmark runs.
此外，这是怎么回事你的核心检测多少？我的四核采用的333334块大小只允许3线程运行。
Also, what's going on with your number of core detection? My quad-core is using a chunk size of 333334 which allows only 3 threads to run.
这篇关于为什么等待异步这么慢？的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！
