|
Colin Burgess(deleted)
|
Re: strange interpretation of tracelogger data of a multicore system
|
|
Colin Burgess(deleted)
02/28/2008 8:33 PM
post5386
|
Re: strange interpretation of tracelogger data of a multicore system
> One possibility, is that depending on the hardware, there are some systems
> where the CPU clocks are
> not synchronized. If that is the case, then the timestamp information
> drifts off and then of course the
> interpretation is going to be messed up. There is a PR filed and some work
> has been done to mitigate
> that problem.
If this is the issue, then the simple fix for this is to disable the halt-on-idle feature.
Cheers,
Colin
--
cburgess@qnx.com
|
|
|
|
|
|
Yannick Dadji
|
Re: strange interpretation of tracelogger data of a multicore system
|
|
Yannick Dadji
02/29/2008 5:49 AM
post5392
|
Re: strange interpretation of tracelogger data of a multicore system
> On Thu, Feb 28, 2008 at 8:46 AM, Yannick Dadji <y.dadji-foyet@tu-bs.de>
> wrote:
>
> > Hello,
> > i use the tracelogger utility on a dual-core PC under qnx6.3.2 to trace
> > system event of two Threads running on different CPU and communicatiing over
>
> > pulse. The display of the event on the windows momentic IDE is quite
> > strange. For example, it shows the receive of a pulse by Thread 1 (CPU1)
> > before this pulse has been sent from Thread 2 (CPU2).
> >
> > Does it have something to do with the CPU clock registers, which are not
> > syncchronised with each oder?
> >
> > Does somebody have experience with the tracelogger utility on multicore
> > systems?
> >
> Are you sure that the pulse that is being received is the one that is being
> sent and not one from "earlier on".
> If you want to post a snippet of the log (text from the trace view or the
> log itself) then it would be easier to
> comment.
>
> One possibility, is that depending on the hardware, there are some systems
> where the CPU clocks are
> not synchronized. If that is the case, then the timestamp information
> drifts off and then of course the
> interpretation is going to be messed up. There is a PR filed and some work
> has been done to mitigate
> that problem.
>
> Thomas
Hello Thomas,
thanks for your comment.
The CPU clocks are definitively not synchronised on my system.
>There is a PR filed and some work
> has been done to mitigate
> that problem.
What do you mean with "PR"???
Yannick
|
|
|
|
|
|
Yannick Dadji
|
Re: strange interpretation of tracelogger data of a multicore system
|
|
Yannick Dadji
02/29/2008 6:16 AM
post5394
|
Re: strange interpretation of tracelogger data of a multicore system
Hello Thomas,
thanks for your comment.
The CPU clocks are definitively not synchronised on my system.
>There is a PR filed and some work
> has been done to mitigate
> that problem.
What do you mean with "PR"???
best regards
Yannick
|
|
|
|
|
|
Yannick Dadji
|
Re: strange interpretation of tracelogger data of a multicore system
|
|
Yannick Dadji
02/29/2008 6:18 AM
post5395
|
Re: strange interpretation of tracelogger data of a multicore system
> On Thu, Feb 28, 2008 at 8:46 AM, Yannick Dadji <y.dadji-foyet@tu-bs.de>
> wrote:
>
> > Hello,
> > i use the tracelogger utility on a dual-core PC under qnx6.3.2 to trace
> > system event of two Threads running on different CPU and communicatiing over
>
> > pulse. The display of the event on the windows momentic IDE is quite
> > strange. For example, it shows the receive of a pulse by Thread 1 (CPU1)
> > before this pulse has been sent from Thread 2 (CPU2).
> >
> > Does it have something to do with the CPU clock registers, which are not
> > syncchronised with each oder?
> >
> > Does somebody have experience with the tracelogger utility on multicore
> > systems?
> >
> Are you sure that the pulse that is being received is the one that is being
> sent and not one from "earlier on".
> If you want to post a snippet of the log (text from the trace view or the
> log itself) then it would be easier to
> comment.
>
> One possibility, is that depending on the hardware, there are some systems
> where the CPU clocks are
> not synchronized. If that is the case, then the timestamp information
> drifts off and then of course the
> interpretation is going to be messed up. There is a PR filed and some work
> has been done to mitigate
> that problem.
>
> Thomas
Hello Thomas,
thanks for your comment.
The CPU clocks are definitively not synchronised on my system.
>There is a PR filed and some work
> has been done to mitigate
> that problem.
What do you mean with "PR"???
Yannick
|
|
|
|
|
