Hotspot by function: No symbols were found

[jcdenton]

Hello forum,

After installing VTune on FC3, I am facing the following issue. Being aware that VTune is not supported on
FC3, I would like to know whether the problem I am facing as below is due to this incompatibility or is it a
problem faced by people irrespective of their platforms..

My problem is that
$vtl activity -hf -mn [module name]

gives me the error:
No symnbols were found.

The activity was that of sampling. I did not try callgraph, however.

I get this even if I use -g option during the compilation. ( My compilers are icc 8.1 & ifort 8.1 ). As I
searched for similar posts in this forum, and some of the solutions were like:

Solution: The program runs too quick to VTune to get the samples
Results for me: But my program runs for around 9 Mins.
Solution: Give the symbol directory -sd
Results for me: Even this does not work.

As an aside, I tried
$vtl activity -ha -mn [module name]

And I get some addresses listed according to the events. Now if I try to find this address in the output of

$nm [excecutable]

I see that no address listed by -ha view is listed in the output of nm. I dont have much idea about these
adresses, but shouldn't they match? Is that why I am getting the error?
If it is valid for nm to have different addresses than -ha view, then can I somehow get a map between these
addresses?

Thanks in anticipation,
jcd

Message Edited by jcdenton on 05-05-2005 11:44 PM

[jeffrey-gallagher]

VERY interesting, jcdenton.

You already know that FC3 is not supported, but I'd like for you to compare the vtl results you've reported with vtlec:

$ vtlec

use the activity wizard to duplicate the vtl profiling experiment.

Are the results the same?

just curious

jdg

[jcdenton]

it works !!

thanks a lot jdg for the hint ..:-)

why is this so? is there any setting that i should change so that it could work on the command line too ?

since its a server, it runs a little slow coz of X. otherwise i'm fine working with the GUI. just curious, you know...:-)

thanks,
jcd

[jeffrey-gallagher]

Thanks for reporting back, jcd.

You ask a good question which I'm going to have to tap dance around for now (putting TEA FOR TWO in the CD player and getting my dancing shoes...)

Truthfully, I don' t know the exact reason why the GUI might work when the CLI failed, in particular within an unsupported OS. Based on your report, I'm guessing that there's some kind of shell environmental factor (env, set) which is still, cryptically, not properly set up; that if set, *would* enable the CLI to also work.

BUT, that's just my swag. In general, the only differences between CLI and GUI are the Eclipse bits themselves.

Let me put your very good question on my back burner, and, if I gain some insight from engineering about this along the way, I'll report back here.

:)

cheers

jdg

Message Edited by jdgallag on 05-17-2005 09:20 AM

Message Edited by jdgallag on 05-17-2005 09:21 AM

Message Edited by jdgallag on 05-17-2005 09:23 AM

Message Edited by jdgallag on 05-17-2005 09:25 AM