Why mutexes are slower than channels in golang?-CodePudding

I am making a program that crawls websites and return status of them.

I wrote this program with different approaches. The first one using mutexes to prevent concurrent writes to map so that I can get rid of the data race. Then for the same purpose, I implement it with channels. But when I was doing benchmarks I realized that implementing it with channels much faster than implementing it mutexes. I was wondering why it is happening? Why mutexes lacks of performance? am I doing something wrong with mutexes?

Benchmark result:

Code

package concurrency

import "sync"

type WebsiteChecker func(string) bool
type result struct {
    string
    bool
}

func CheckWebsites(wc WebsiteChecker, urls []string) map[string]bool {
    results := make(map[string]bool)
    var wg sync.WaitGroup
    var mu sync.Mutex
    for _, url := range urls {
        wg.Add(1)
        go func(u string) {
            defer wg.Done()
            mu.Lock()
            results[u] = wc(u)
            mu.Unlock()
        }(url)
    }
    wg.Wait()
    return results
}

func CheckWebsitesChannel(wc WebsiteChecker, urls []string) map[string]bool {
    results := make(map[string]bool)
    resultChannel := make(chan result)
    for _, url := range urls {
        go func(u string) {
            resultChannel <- result{u, wc(u)}
        }(url)
    }
    for i := 0; i < len(urls); i   {
        r := <-resultChannel
        results[r.string] = r.bool
    }
    return results
}

Test code

package concurrency

import (
    "reflect"
    "testing"
    "time"
)

func mockWebsiteChecker(url string) bool {
    time.Sleep(20 * time.Millisecond)
    if url == "https://localhost:3000" {
        return false
    }
    return true
}

func TestCheckWebsites(t *testing.T) {
    websites := []string{
        "https://google.com",
        "https://localhost:3000",
        "https://blog.gypsydave5.com",
    }
    want := map[string]bool{
        "https://google.com":          true,
        "https://blog.gypsydave5.com": true,
        "https://localhost:3000":      false,
    }
    got := CheckWebsites(mockWebsiteChecker, websites)
    if !reflect.DeepEqual(got, want) {
        t.Errorf("got %v, want %v", got, want)
    }
}

func BenchmarkCheckWebsites(b *testing.B) {
    urls := make([]string, 1000)
    for i := 0; i < len(urls); i   {
        urls[i] = "a url"
    }
    b.ResetTimer()
    for i := 0; i < b.N; i   {
        CheckWebsites(mockWebsiteChecker, urls)
    }
}

func BenchmarkCheckWebsitesChannel(b *testing.B) {
    urls := make([]string, 1000)
    for i := 0; i < len(urls); i   {
        urls[i] = "a url"
    }
    b.ResetTimer()
    for i := 0; i < b.N; i   {
        CheckWebsitesChannel(mockWebsiteChecker, urls)
    }
}

CodePudding user response：

It seems to me that with the mutex version of the code you're not only protecting the results map but the wc too (the call can only take place once the lock has been acquired so effectively you are serializing the calls). Send to chan only locks the channel once the right side is ready so calls to wc can happen concurrently. See does code like

        go func(u string) {
            defer wg.Done()
            r := wc(u)
            mu.Lock()
            results[u] = r
            mu.Unlock()
        }(url)

perform better with mutex.