The Linux Programming Interface

206 Chapter 10

with greater accuracy and time measurements can be made with greater precision.

However, it isn’t desirable to set the clock rate to arbitrarily high values, because

each clock interrupt consumes a small amount of CPU time, which is time that the

CPU can’t spend executing processes.

Debate among kernel developers eventually resulted in the software clock

rate becoming a configurable kernel option (under Processor type and features,

Timer frequency). Since kernel 2.6.13, the clock rate can be set to 100, 250 (the

default), or 1000 hertz, giving jiffy values of 10, 4, and 1 milliseconds, respec-

tively. Since kernel 2.6.20, a further frequency is available: 300 hertz, a number

that divides evenly for two common video frame rates: 25 frames per second

(PAL) and 30 frames per second (NTSC).

10.7 Process Time.............................................................................................................

Process time is the amount of CPU time used by a process since it was created. For

recording purposes, the kernel separates CPU time into the following two

components:

z User CPU time is the amount of time spent executing in user mode. Sometimes

referred to as virtual time, this is the time that it appears to the program that it

has access to the CPU.

z System CPU time is amount of time spent executing in kernel mode. This is the

time that the kernel spends executing system calls or performing other tasks

on behalf of the program (e.g., servicing page faults).

Sometimes, we refer to process time as the total CPU time consumed by the process.

When we run a program from the shell, we can use the time(1) command to

obtain both process time values, as well as the real time required to run the program:

$ time ./myprog

real 0m4.84s

user 0m1.030s

sys 0m3.43s

The times() system call retrieves process time information, returning it in the struc-

ture pointed to by buf.

This tms structure pointed to by buf has the following form:

struct tms {

clock_t tms_utime; /* User CPU time used by caller */

clock_t tms_stime; /* System CPU time used by caller */

#include <sys/times.h>

clock_t times(struct tms *buf);

Returns number of clock ticks (sysconf(_SC_CLK_TCK)) since

“arbitrary” time in past on success, or (clock_t) –1 on error