Forums

profiling

Started by Unknown February 28, 2001
Can someone tell me?
Why does the profiling cycle for the stand alone simulation and the
profiling using CCS varies so much? Which is more accurate?
Thanks for your assistance.




The "fast" standalone simulator doesn't account for cache misses.
Everything runs at the maximum possible speed.

Andrew

At 01:26 AM 2/28/01 +0000, wrote:
>Can someone tell me?
>Why does the profiling cycle for the stand alone simulation and the
>profiling using CCS varies so much? Which is more accurate?
>Thanks for your assistance. >
>_____________________________________
>Note: If you do a simple "reply" with your email client, only the author of
>this message will receive your answer. You need to do a "reply all" if you
>want your answer to be distributed to the entire group.
>
>_____________________________________
>About this discussion group:
>
>To Join: Send an email to
>
>To Post: Send an email to
>
>To Leave: Send an email to
>
>Archives: http://www.egroups.com/group/c6x
>
>Other Groups: http://www.dsprelated.com >">http://docs.yahoo.com/info/terms/
>


> Date: Wed, 28 Feb 2001 01:26:31 -0000
> From:
> Subject: profiling
>
> Can someone tell me?
> Why does the profiling cycle for the stand alone simulation and the
> profiling using CCS varies so much? Which is more accurate?
> Thanks for your assistance.
>

I may guess that you use clock() to profile code in the standalone
mode? If yes, the clock count is different from that of obtained
with CCS due to call overhead for the clock() call and that same
overhead due to the function under profiling call; of course the
second overhead exists only in case you profiling a function and
not simply a bit of code.

TI describes this issue in much greater detail in the Programmers
Guide, SPRU198.

The answer on the question "which is more accurate" could be not
unique one - for some reason one could ignore the call overhead,
others would not. I think that since this overhead really exists
it ought to be taken into account, that is profile points in CCS
should be before and after a function call and not inside the
function code.

Regrads,

Andrew.


hi ,
i am new to this group..
i am having a strange problem.
i have tested my audio decoder for all basic test vectors and number
of mp3 songs.it is passing rms test for all.
when i tried to profile all functions with o3,mt options on and kept
it in function profile debug mode.then i tried to profile.it is
faling for the basic tests..please suggest me what mistake i am
doing..and i removed all the options i.e o3 mt and i kept it in full
simbolic debug and tried..still it is going to infinit loop some
where.please suggest me the correct way to profile...
thanking you
balaji morram