shock! Timers actually have so many features, did you know?

In the process of data communication, you will encounter format requirements for data sending time. Therefore, in order to select different timers according to actual requirements in the application, it is necessary to consider the characteristics of several application timers.

Timer article reference

Generally speaking, yes,

1. sleep, usleep and nanosleep

sleep() and nanosleep() both cause the process to sleep for a period of time before being awakened, and their implementations are completely different.

Linux does not provide the system call sleep(). Sleep() is implemented in the library function. It sets the reporting time by calling alarm() and calls sigsuspend() to suspend the process on the signal SIGALARM. , sleep() can only be accurate to the second level.

nanosleep() is a system call in Linux linux timer precision, which is implemented using a timer. This call causes the calling process to sleep and adds a timer_list type to the timer queue. For timers, the time_list structure includes the awakening time and the function to be executed after awakening. The execution function of the timer added through nanosleep() only completes the function of awakening the current process. The system regularly detects these queues through a certain mechanism (for example, after being trapped in the core through a system call and before returning from the core to user mode, it must check whether the time slice of the current process has been exhausted, and if so, call the schedule() function to reschedule. , the timer queue will be detected in this function, and this detection will also be done before the slow interrupt returns). If the timing time has exceeded, the function specified by the timer will be executed to wake up the calling process. In fact, because the system time slice may be lost, the accuracy of nanosleep() is not very high.

alarm() is also implemented by the Hongqi Linux operating system through a timer, and its accuracy is only accurate to the second level. In addition, the timer execution function it sets is to send the SIGALRM signal to the current process at the specified time.

2. Use the semaphore SIGALRM+alarm()

Although the alarm method is very good, the accuracy of these methods can reach one second, but it is difficult to have an accuracy higher than one second. The semaphore mechanism of the *nix system is used. First, register the semaphore SIGALRM processing function, call alarm(), and set the timing width. The code is as follows:

//设置一个1s延时信号，再注册一个
#include 
#include 
void timer(int sig)
{
if(SIGALRM == sig)
{
printf("timern");
alarm(1); //重新继续定时1s
}

return ;
}
int main()
{
signal(SIGALRM, timer); //注册安装信号

alarm(1); //触发定时器

getchar();

return 0;
}

3. Use RTC mechanism

The RTC mechanism relies on the RealTimeClock mechanism provided by the system hardware. By reading the RTC hardware/dev/rtc and setting the RTC frequency through ioctl(), these methods are more convenient. Linux timer accuracy, with the help of the system The RTC provided by the hardware has adjustable accuracy and a particularly high code as follows:

#include 
#include 
#include 
#include 
#include 
#include 
#include 
#include 
#include 
int main(int argc, char* argv[])
{
unsigned long i = 0;
unsigned long data = 0;
int retval = 0;
int fd = open ("/dev/rtc", O_RDONLY);

if(fd < 0)
{
perror("open");
exit(errno);
}

/*Set the freq as 4Hz*/
if(ioctl(fd, RTC_IRQP_SET, 1) < 0)
{
perror("ioctl(RTC_IRQP_SET)");
close(fd);
exit(errno);
}
/* Enable periodic interrupts */
if(ioctl(fd, RTC_PIE_ON, 0) < 0)
{
perror("ioctl(RTC_PIE_ON)");
close(fd);
exit(errno);
}

for(i = 0; i < 100; i++)
{
if(read(fd, &data, sizeof(unsigned long)) < 0)
{
perror("read");
close(fd);
exit(errno);

}
printf("timern");
}
/* Disable periodic interrupts */
ioctl(fd, RTC_PIE_OFF, 0);
close(fd);

return 0;
}

This method requires the system to have an RTC device linux boot disk creation tool. Our 1860 has two RTCs. The RTC in the LC1160 of the power management module is used, and there is no support for the RTC_IRQP_SET control word in the driver. This needs to be done later. Add driver implementation.

4、使用select()

能精确到1us，目前精确定时的最流行方案。通过使用select()，来设置定时器；原理借助select()方式的第5个参数，第一个参数设置为0，三个文件描述符集都设置为NULL，第5个参数为时间结构体，代码如下：

#include 
#include 
#include 
#include 
/*seconds: the seconds; mseconds: the micro seconds*/
void setTimer(int seconds, int mseconds)
{
struct timeval temp;

temp.tv_sec = seconds;
temp.tv_usec = mseconds * 1000;

select(0, NULL, NULL, NULL, &temp);
printf("timern");

return ;
}
int main()
{
int i;

for(i = 0 ; i < 100; i++)
setTimer(1, 0);

return 0;
}

结果是，每隔1s复印一次，复印100次。

select定时器是阻塞的，在等待时间到来之前哪些都不做。要定时可以考虑再开一个线程来做。

最终，下层应用程序的定时器的京都是内核决定的，这几个定时器的使用方式，尽管理论上可以精确到微妙，虽然，我分别在微妙级测试的时侯，偏差还是很大的，取决于，当时的cpu和进程的调度，等等。。。。

The above is the detailed content of shock! Timers actually have so many features, did you know?. For more information, please follow other related articles on the PHP Chinese website!