FREE BSD fine-grain timer

FREE BSD fine-grain timer

Post by Yingfei Don » Wed, 02 Aug 2000 04:00:00



hi, folks,

I need a fine-grain timer (33us) to revise a network interface driver.
If I change 'hz' and rebuild the kernal, do it works?

Usually, 'hz' is equal to 100. So, the timeout resolution is 10 ms.

If I changed to 'hz', I am wondering if other kernal parts which use
'hz' may be messed.

DO you have any ideas? Thanks for your inputs.

yingfei

--
===================================================
Yingfei Dong
4-192 EECS Building,
200 Union Street, SE            Tel: 612-626-7526
Minneapolis, MN 55455           FAX: 612-625-0572
===================================================

 
 
 

FREE BSD fine-grain timer

Post by Chen » Wed, 02 Aug 2000 04:00:00


Set it to 1000 is OK. Never tried higher.

> hi, folks,

> I need a fine-grain timer (33us) to revise a network interface driver.
> If I change 'hz' and rebuild the kernal, do it works?

> Usually, 'hz' is equal to 100. So, the timeout resolution is 10 ms.

> If I changed to 'hz', I am wondering if other kernal parts which use
> 'hz' may be messed.

> DO you have any ideas? Thanks for your inputs.

> yingfei

> --
> ===================================================
> Yingfei Dong
> 4-192 EECS Building,
> 200 Union Street, SE            Tel: 612-626-7526
> Minneapolis, MN 55455           FAX: 612-625-0572
> ===================================================


 
 
 

1. Fine Grained Timers and Portable?

I've recently taken up writing a small schedular, that schedules X many
events in a given time period. This time period is real time, preferably
slicing a number of tasks into 1 or 2 seconds worth.

I've had a look around for different timing facilities. Its occured to me,
there is no POSIX defined OS service that allows me to get the time in
microseconds. times() was the closest one, and it gave it back to me in
seconds.

Also I'm not happy with the getitimer/setitimer pair, since their BSDism,
and I'm worried this makes my application less portable. And they issue
signals which I realy dont want to handle in my schedular. Although this
looks like my only real option right now.

Lastly, its occured to me this is hardware/OS dependant. I believe your
average intel PC clone will keep track at .10 second intervals? Can I be
certain most modern systems will have decent accuracy?

My question is, is there any portable means of getting the current time in
mili or micro seconds? And will it be updated in relatively small intervals?
Did POSIX come up with a new interface for this recently I might look into?

--
Thamer Al-Herbish


2. How to change system-wide search path?

3. Kernel hackers - does Linux have a fine grain timer?

4. HELP: Fax package for RH5.0 needed

5. Fine-grained user authorization

6. folders under $HOME/mail

7. fine-grained ICQ search

8. Heaps of Linux Music software to download...

9. Fine grain parallelism over Solaris 2/SparcCenter 2000 ?

10. fine-grained timing in UNIX

11. bsd "ifree: freeing free inode" panic/bug[?]

12. need a fine timer (like setitimer)

13. fine timers . How to implement?