Wolfgang Grandegger wrote:
> Oliver Hartkopp wrote:

>>> Hm, that does surprise me. At what bitrate are you testing?
>>>
>> I was testing @500kBit/s with an EMS PCMCIA Card connected directly to a PEAK
>> USB Adapter (with termination).
>>
>> Using
>>
>>    cangen can0 -g0 -i -x
>>
>> the generated busload was ca. 1-2% higher and the Gnome CPU systemmonitor
>> indicated a slightly lower CPU usage than with
>>
>>    cangen can0 -g0 -p 1 -x
>>
>> You may try this on our own. So even when the "ignore enobufs" approach looks
>> a bit silly, it has a (positive) difference to the poll implementation.
>>
>> No idea why it is like this ...
> 
> Just did some test at 125 Kb/s on my MPC5200 and can confirm your
> results: almost 100% CPU usage in both cases. poll() seems not to block.
> 

I configured cangen to produce the same output a done by the ptx cansequence
with " -I 2 -L 1 -D i " :

cangen can0 -I 2 -L 1 -D i -g 0 -i

71% - 72% busload , 95% CPU

cangen can0 -I 2 -L 1 -D i -g 0 -p 1

69% busload , 95% CPU

cansequence -p

69% busload , 95% CPU


That's it.

The 5% goes to Xorg running "candump any" and
"canbusload c...@500000 c...@500000 -r -c -b -t"

Regards,
Oliver

_______________________________________________
Socketcan-core mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/socketcan-core

Reply via email to