Tru64下到底能不能更改cpu时间片大小呢，给点思路也行各位大侠

wenzhang005

如提，困扰几天了，帮帮忙吧，我没有查到这方面的资料啊

fish_man

不是很清楚你修改CPU Tick的目的是什么,

如果你想改变tru64分配给process/thread的时间片, 或减少context switches的话, 可以试试修改round_robin_switch_rate

#sysconfig -q proc | grep round_robin_switch_rate

#man sys_attrs_proc

  round_robin_switch_rate
      A value that determines (per second of CPU time) the number of context
      switches that can occur between processes with the same priority and
      the timeslice allotted to each process.

      Default value: 0 (100 context switches per second)

      Minimum value: 0

      Maximum value: hz

      The system uses the following formula to compute timeslice:

      timeslice = hz / number_context_switches

      The hz value is the timer interrupt frequency (hertz) for one second of
      CPU time.  This value is dependent on hardware and is typically 1024 or
      1200. (To determine the hz value on a particular system, programmers
      can use the getsysinfo() call with GSI_CLK_TCK as the op parameter.)

      If round_robin_switch_rate is either 0 (zero) or greater than hz, the
      system uses 100 for number_context_switches. For example, if hz is
      1024, the resulting timeslice is 10 milliseconds, and if hz is 1200,
      the resulting timeslice is 12 milliseconds:

      timeslice = 1024 / 100 = 10

      timeslice = 1200 / 100 = 12

      If round_robin_switch_rate is greater than zero and less than or equal
      to hz, then number_context_switches is set to round_robin_switch_rate.
      For example, changing round_robin_switch_rate from 0 to 20 decreases
      the number of context switches and increases timeslice by a factor of
      five:

      timeslice = 1024 / 20 = 51

      timeslice = 1200 / 20 = 60

      Context switching carries a certain amount of CPU overhead. If a CPU
      consistently handles many large jobs that do not complete quickly, a
      larger timeslice (and fewer context switches) may improve throughput.
      However, the tradeoff is reduced response time for any interactive
      processes that the CPU handles.

Decrements the current thread’s quantum.  If this exhausts the thread’s quantum,
a new thread may get scheduled.  Can invoke the update_priority() routine to update the thread’s priority.

The quantum defaults to 1/100 of a second.  Its actual calculation is done in src/kernel/kern/sched_prim.c:

if ((round_robin_switch_rate > 0) && (round_robin_switch_rate <= hz))
                  min_quantum = hz / round_robin_switch_rate;
else
                  min_quantum = hz / 100;

            where hz (hertz) is a global value representing the number of CPU ticks per second and round_robin_switch_rate is configurable via /etc/sysconfigtab.

fish_man

Source code:
src/kernel/bsd/kern_clock.c
src/kernel/kern/sched_prim.c:

if ((round_robin_switch_rate > 0) && (round_robin_switch_rate <= hz))
                  min_quantum = hz / round_robin_switch_rate;
else
                  min_quantum = hz / 100;

nmliuxinyu

你级别是新手，可是为什么这么强！！！！

chu11cn

确实强！！！！！！！
哪来的源代码？？？

wenzhang005

好长时间没来了，感谢fish_man!!

wenzhang005

前面的明白，下面这一段不太明白，再请教，谢谢！！
Decrements the current thread’s quantum.  If this exhausts the thread’s quantum,
a new thread may get scheduled.  Can invoke the update_priority() routine to update the thread’s priority.

The quantum defaults to 1/100 of a second.  Its actual calculation is done in src/kernel/kern/sched_prim.c:

[ 本帖最后由 wenzhang005 于 2007-5-31 16:26 编辑 ]