[go-nuts] ticker resolution

On Thursday, 1 May 2014 07:42:24 UTC+10, Jens Frederich wrote:

Hi all,

I've done a little ticker precision test. I want to do something in a 4
Mhz interval. My test creates a ticker with 250 nsec duration. So I've
expected 4 * 10^6 / 250 nsec = 16000 calls, but I got no more then 13000.
What's wrong?

package main

import (
"fmt"
"time"
)

func main() {
t := time.NewTicker(250 * time.Nanosecond)
stop := make(chan bool, 1)
c := 0

go func() {
for {
select {
case <- t.C:
c++
case <- stop:
return
}
}
}()
time.Sleep(1*time.Second)
stop <- true
fmt.Println(c)
}

--

On Thursday, 1 May 2014 07:55:48 UTC+10, Dave Cheney wrote:

It is worth nothing that this program has a data race.

lucky(~/src) % go run -race rr.go
==================
WARNING: DATA RACE
Read by main goroutine:
main.main()
/home/dfc/src/rr.go:25 +0x1c8

Previous write by goroutine 6:
main.func·001()
/home/dfc/src/rr.go:17 +0xcf

Goroutine 6 (running) created at:
main.main()
/home/dfc/src/rr.go:22 +0x174
==================
11859
Found 1 data race(s)
exit status 66

On Thursday, 1 May 2014 07:42:24 UTC+10, Jens Frederich wrote:

Hi all,

I've done a little ticker precision test. I want to do something in a 4
Mhz interval. My test creates a ticker with 250 nsec duration. So I've
expected 4 * 10^6 / 250 nsec = 16000 calls, but I got no more then 13000.
What's wrong?

package main

import (
"fmt"
"time"
)

func main() {
t := time.NewTicker(250 * time.Nanosecond)
stop := make(chan bool, 1)
c := 0

go func() {
for {
select {
case <- t.C:
c++
case <- stop:
return
}
}
}()
time.Sleep(1*time.Second)
stop <- true
fmt.Println(c)
}