Rough evaluation of execution time

[Fabio Degiovanni - Eicas]

Thank you very much for help. I know that accurate time measurement 
could be done only on the hardware, but at the moment we don't have the 
ERC32 target available and we have to coumpute a rough estimation.

When Gregm speaks about the min resolution of the OS clock tick, I 
presume that you speak about the target platform OS (in my case LInux, I 
know that Linux use the 8254 timer with a frequency of 100Hz) otherwise 
I don't understand because using the configuration parameter 
CONFIGURE_MICROSECONDS_PER_TICK within RTEMS it seeems to me that I can 
get a resolution of 1 us. Could you kindly explain this concept better 
to me?

On the other hand Joel Sherrill told me that the simulator's notion of 
time is independent of the host computer's
speed. What about the time resolution? I presume that this is host 
computer dipendent. In which way?

Could anyone told me where I can find what erc32 is simulated by sis/gdb 
(a 10, 14, or 15 Mhz or what)?

Thank you very much

Fabio Degiovanni

gregory.menke at gsfc.nasa.gov wrote:

>Fabio Degiovanni - Eicas writes:
> > Dear Sirs,
> >                  I would like to evaluate roughtly the execution time on
> > ERC32 of my application developped using the GNU development chains
> > under Linux. Could I use the sis/gdb simulator for this purpose? I
> > thought to use the clock-get directive. Is the time returned by this
> > directive the "real" time of ERC32 or is affected by the platform on
> > which I'm running the ERC32 simulator? Have you some other suggestions
> > for my problem?
> > Thank you very much for help
>
>The min resolution of the OS clock tick is whatever the quanta is
>configured for, presumably 10ms- which is all you'll get from the OS
>timer functions.  
>
>If you want <really> accurate time measurement, you'll have to use
>hardware.  On the PowerPC, theres a 64 bit counter that runs at 1/4
>the cpu clock.  If you want benchmarks, the dhrystone and whetsone
>benchmarks are easy enough to run in RTEMS, as are many others.
>
>Gregm
>
>


-- 
Dott. Ing. Degiovanni Fabio
Eicas Automazione
Via Vincenzo Vela, 27 10128 Torino (ITALIA)
Telefoni +39-11-562.37.98/562.3088 Fax +39-11-436.06.79