Re: microtime latency

["Sean Kennedy" <woodentulip@hotmail.com>]

>From: Joshua Lamorie <jpl@xiphos.ca>
>Subject: microtime latency
>
>Gidday there,

G'day, G'day, G'day.

>I'm trying to measure round trip time of packets within the
>kernel, using 'microtime'.  What is the latency on the value that comes
>back from this?  I log the time when the packet leaves, and then read the 
>clock again when it comes back.
>I get a deviation of +/- 40ms on the
>round trip time (i'm using a traffic shaper to control the link).

Hmm, that is quite a bit.

>The logging facility is a little odd to me (I didn't write it),
>but there aren't many moving parts.  I want to know if the microtime
>function is guaranteed to be accurate.

Well, from looking at the Assembler Code I cannot make any guarantees.
This is not an RTOS, but a Context Sensitive kernel.

In VxWorks, LynxOS, RTX, and or QNX, the intervals are well defined.

>Why would ping, using 'gettimeofday' be more accurate than my own
>stuff?

I will check, you bring up a good point.

>Any hints would be appreciated.
>Joshua

I won't attempt to hint, since I am not that knowledgeable of all of oBSD 
internals, but oBSD intrigues me as an alternate (FREE) PowerPC OS.

BTW, I have worked with QNX, and LynxOS, 4-5 years ago, but I still remember 
a bit of my homework.

I'll see what I can assess.

-sean

Sean Kennedy
Wooden Tulip Ltd.

_________________________________________________________________
Get your FREE download of MSN Explorer at http://explorer.msn.com/intl.asp