Hi Philippe,

I have sent the forwarded e-mail to the rtai-dev mailing list but it
didn't arrive yet (maybe there is still a problem with switching to
GNA). Maybe you can help?

Furthermore, I'm not that far from checking the ARTI-PPC port in.
Currently I'm using RTAI 3.0 (kilauea) but I assume it's better to
switch immediately to RTAI 3.1 (vesuvio).

Wolfgang.


On 07/01/2004 05:07 PM Wolfgang Grandegger wrote:
> Hi,
> 
> I have now a very first version of RTAI over ADEOS on PowerPC working on
> my Icecube board and I can run sucessfully the tests in "rtai-testing".
>  Nevertheless, I'm a bit puzzled about the figures reported by the task
> switching test "switches". With the old RTAI 24.1.12 I get:
> 
>    FOR 30 TASKS: TIME 2177 (ms), SUSP/RES SWITCHES 600000,
>    SWITCH TIME (INCLUDING FULL FP SUPPORT) 3628 (ns)
> 
> 
>    FOR 30 TASKS: TIME 2296 (ms), SEM SIG/WAIT SWITCHES 600000,
>    SWITCH TIME (INCLUDING FULL FP SUPPORT) 3827 (ns)
> 
> and with the new ARTI:
> 
>   FOR 30 TASKS: TIME 1086 (ms), SUSP/RES SWITCHES 600000,
>   SWITCH TIME (INCLUDING FULL FP SUPPORT) 1810 (ns)
> 
> 
>   FOR 30 TASKS: TIME 2061 (ms), SEM SIG/WAIT SWITCHES 600000,
>   SWITCH TIME (INCLUDING FULL FP SUPPORT) 3436 (ns)
> 
> Any ideas where the difference comes from. Especially why a suspend/
> resume is much faster?
> 
> Wolfgang.


Reply via email to