Re: [Paparazzi-devel] LPC2148 performance

[Felix Ruess]

Hi Mark,

thanks for catching that include... works without it for booz ;-)

I would be a bit careful with the cpu_load estimate though for two reasons:
1) the average cpu_load over one second is in general a crude way to
tell if you are still OK
2) more specifically: the module currently disregards any computation
done in the event loop when calculating the cpu_load

For the fixed wings it probably doesn't make a big difference, but in
booz some of the filters run in the event loop and hence your cpu_load
estimate is lower than you actual load.

But I'm sure we'll have an improved cpu_load estimation soon ;-)

Cheers, Felix

On Thu, Aug 5, 2010 at 9:35 PM,  <address@hidden> wrote:
> Hi Gautier,
> Thanks for writing this super new module.
>
> I found that in order for it to compile, sys_mon.c needed a #include 
> "ap_downlink.h" to be added (I committed this modification to the SVN).
>
> I have tried it on my FunJet (TWOG, gyro, i2c airspeed sensor, 57600 bps 
> transparent mode telemetry) and get 37% CPU load. Less than I anticipated but 
> I'm now confident the CPU is not close to being overloaded.
>
> Cheers, Mark
>
>
> -----Original Message-----
> From: address@hidden [mailto:address@hidden On Behalf Of Gautier Hattenberger
> Sent: mercredi, 4. août 2010 18:33
> To: address@hidden
> Subject: Re: [Paparazzi-devel] LPC2148 performance
>
> Hello,
>
> I just add a new module "sys_mon.xml" to monitor the CPU activity. It
> gives the actual period of the periodic calls, the duty cycle of these
> calls with min and max values, the percentage of the cycle and the
> average number of event calls between to periodic calls. And it works
> with fixed-wing and quadrotors autopilot.
>
> Gautier
>
> Le 04/08/2010 17:49, Felix Ruess a écrit :
>> Hi,
>>
>> I recently did this for Booz. If I did the calculations correctly I
>> had an average CPU load of around 50% and a max of just over 60% on
>> Booz.
>> For fixed wings this should be significantly lower.
>> You can check out what I did for Booz (in booz2_main.c) and define
>> CALC_CPU_LOAD in your airframe file.
>> Gautier was also thinking of doing a cpu load computation with a
>> module for fw and booz in the near future.
>>
>> Cheers, Felix
>>
>> On Wed, Dec 2, 2009 at 6:01 PM, Martin Mueller<address@hidden>  wrote:
>>
>>> Hi,
>>>
>>> some long time ago a measurement was done for the ATMEGA processor and 
>>> showed a maximum load of around 17% for that "old" 8bit 16MHz processor. 
>>> Maybe that stuff could help to implement a load monitor for the LPC.
>>>
>>> http://lists.gnu.org/archive/html/paparazzi-devel/2005-09/msg00001.html
>>>
>>> Martin
>>>
>>> ----- original Nachricht --------
>>>
>>> Betreff: RE: [Paparazzi-devel] LPC2148 performance
>>> Gesendet: Mi, 02. Dez 2009
>>> Von: Cédric Marzer<address@hidden>
>>>
>>>
>>>> Would it be possible to add an extra message that shows a percentage
>>>> telling
>>>> how much of the processing power one has left ? I think there is plenty of
>>>> processing power left for a normal flight but having some feedback would be
>>>> reassuring...
>>>>
>>>> Cédric
>>>>
>>>> -----Message d'origine-----
>>>> De : address@hidden
>>>> [mailto:address@hidden De la
>>>> part
>>>> de Steve Joyce
>>>> Envoyé : mardi, 1. décembre 2009 15:52
>>>> À : address@hidden
>>>> Objet : [Paparazzi-devel] LPC2148 performance
>>>>
>>>> Has anyone ever done any profiling on the airborne code to see how how many
>>>>
>>>> extra cycles are available in the various loops? I don't really have a good
>>>>
>>>> feeling for how much extra code you could sqeeze in there.  Are there tools
>>>>
>>>> available for that type of analysis?
>>>>
>>>> What would actually happen if you exceeded the available instructions
>>>> within
>>>>
>>>> a loop?
>>>>
>>>> /steve
>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> Paparazzi-devel mailing list
>>>> address@hidden
>>>> http://lists.nongnu.org/mailman/listinfo/paparazzi-devel
>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> Paparazzi-devel mailing list
>>>> address@hidden
>>>> http://lists.nongnu.org/mailman/listinfo/paparazzi-devel
>>>>
>>>>
>>> --- original Nachricht Ende ----
>>>
>>>
>>>
>>> _______________________________________________
>>> Paparazzi-devel mailing list
>>> address@hidden
>>> http://lists.nongnu.org/mailman/listinfo/paparazzi-devel
>>>
>>>
>> _______________________________________________
>> Paparazzi-devel mailing list
>> address@hidden
>> http://lists.nongnu.org/mailman/listinfo/paparazzi-devel
>>
>
> _______________________________________________
> Paparazzi-devel mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/paparazzi-devel
>
> _______________________________________________
> Paparazzi-devel mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/paparazzi-devel
>