Gettimeofday and clock_gettime are different.

Clock_gettime is more accurate than gettimeofday

Perform a simple test and use clock_gettime to perform the test.
Test. c

#include<time.h> #include<stdio.h> #define MILLION 1000000 int main(void) {         struct timespec tpstart;         struct timespec tpend;         long timedif;         clock_gettime(CLOCK_MONOTONIC, &tpstart);         clock_gettime(CLOCK_MONOTONIC, &tpend);         timedif = MILLION*(tpend.tv_sec-tpstart.tv_sec)+(tpend.tv_nsec-tpstart.tv_nsec)/1000;         fprintf(stdout, "it took %ld microseconds/n", timedif);         return 0; }

Under Linux 2.6 kernel
Gcc-O test. C-LRT
./Test
Expected result:
It took 2 microseconds
Use gettimeofday to perform the test.

#include<time.h> #include<stdio.h> #define MILLION 1000000 int main(void) {         struct timespec tpstart;         struct timespec tpend;         long timedif;         gettimeofday(&tpstart, NULL);         gettimeofday(&tpend, NULL);         timedif = MILLION*(tpend.tv_sec-tpstart.tv_sec)+(tpend.tv_nsec-tpstart.tv_nsec)/1000;         fprintf(stdout, "it took %ld microseconds/n", timedif);         return 0; }

Gcc-O test. c
./Test
Expected result:
It took 0 microseconds AIX also gets the same result

In addition, we can use the clock_gettime function to test that the 64-bit program runs faster than the 32-bit program.
On AIX p570, 16 machines with CPU 15.5g memory tested
The result obtained in 64-Bit mode is 0 microseconds.
The result of the 32-Bit mode is 1 microsecond.