How to implement a counter when using golang's goroutine?
There are a few things you should fix.
The methods on the Queue type should have pointer receivers. Otherwise, every method call will create a copy of the current queue type and any changes to queue fields will not persist beyond the method call itself.
Waiting for all routines to finish, can be done using a
sync.WaitGroup
. This is specifically what it was designed for.Maintaining a thread-safe push/pop counter inside the queue type can be done by using the
sync/atomic
package.
As far as speed goes, from your example, I am not quite sure what you are trying to achieve. Any optimizations might come up if you elaborate on that a little.
Here is an example I modified from your code:
package main
import (
"log"
"runtime"
"sync"
"sync/atomic"
"time"
)
const SizePerThread = 10000000
type Queue struct {
records string
count int64
}
func (q *Queue) push(record chan interface{}) {
record <- time.Now()
newcount := atomic.AddInt64(&q.count, 1)
log.Printf("Push: %d", newcount)
}
func (q *Queue) pop(record chan interface{}) {
<-record
newcount := atomic.AddInt64(&q.count, -1)
log.Printf("Pop: %d", newcount)
}
func main() {
var wg sync.WaitGroup
runtime.GOMAXPROCS(runtime.NumCPU())
record := make(chan interface{}, 1000000)
queue := new(Queue)
// We are launching 20 goroutines.
// Let the waitgroup know it should wait for as many
// of them to finish.
wg.Add(20)
for i := 0; i < 10; i++ {
go func() {
defer wg.Done()
for j := 0; j < SizePerThread; j++ {
queue.push(record)
}
}()
go func() {
defer wg.Done()
for j := 0; j < SizePerThread; j++ {
queue.pop(record)
}
}()
}
// Wait for all goroutines to finish.
wg.Wait()
}