I want to measure the time taken by a process and several threads inside it individually.

Say my process has 10 threads, I need to measure the runtime of individual thread and also the process allocated time 
for individual thread in the best possible resolution.

I am using following API's:

For process :
ClockCycles(): This gives time in microseconds resolution.
For thread:
Pthread_getcpuclockid() , followed by clock_gettime() for thread ID.

I however see huge difference in both the values i.e process allocated time for thread and thread actual runtime.

Logic used: start() is called at start of thread execution and stop() at end. Time stamps are taken at start() and 
stop() and difference is calculated.

Please suggest some better way.