Linux user space to obtain ns at the nanosecond level

When testing program performance, we often need to obtain precise ns-level time to measure the performance of a program. The following describes how to obtain ns-level time in linux.

Use the clock_gettime function. The function prototype is as follows:

Long sys_clock_gettime (clockid_t which_clock, struct timespec * tp );

1. which_clock parameter explanation

CLOCK_REALTIME

CLOCK_MONOTONIC: Timing starts from the moment when the system is started, and is not affected by changes in the system time.

CLOCK_PROCESS_CPUTIME_ID: time spent by the current process to the CPU of the current Code System

CLOCK_THREAD_CPUTIME_ID: time spent by the current thread to the CPU of the current Code System

2. struct timespec Structure

The sample code is as follows:

# Include <stdio. h> <stdlib. h> <time. h> main (timespec time_start = {,}, time_end = {,&, time_end. TV _sec-time_start. TV _sec, time_end. TV _nsec-

 

Compile command:

Gcc test. c-o test-lrt

Running result:

./

The running result shows that it takes about 15 us to call the printf () function at a time.