[ltt-dev] UST Instrumenting function entries and exits

[Matthew Khouzam]

That would work well, I thought about it, the only issue I can think of
is if you have a lib you load and then you unload it and load a second
lib, you may have some name squashing. Basically whenever there's a
dlopen, we would need to update the symbol table.


On 11-05-12 02:32 PM, Nils Carlson wrote:
> Sorry for the outlook response... :-)
>
> Could maybe find a way to flush function names (and addresses) as metadata? Will require some work, but should be feasible. Would require significant post-processing.
>
> /Nils 
>
> -----Original Message-----
> From: Matthew Khouzam [mailto:matthew.khouzam at ericsson.com] 
> Sent: den 12 maj 2011 20:28
> To: ltt-dev at lists.casi.polymtl.ca
> Subject: Re: [ltt-dev] UST Instrumenting function entries and exits
>
> We could do a compromise and save the string the when there is a new address. There are many caching schemes that can work.
>
> Thanks for the input Francis.
>
> On 11-05-12 02:18 PM, Francis Giraldeau wrote:
>> On Thu, 2011-05-12 at 13:32 -0400, Matthew Khouzam wrote:
>>> Hello world,
>>> I just made a little program that I'm testing out and want some 
>>> opinions now that Mathieu D and Nils are not able to read their 
>>> emails. ;)
>>>
>>> This is a shared object (or code injected straight into the source) 
>>> that will allow ust calls to be hooked onto the function entries and exits.
>> IMHO, this is truly a killer feature for UST! I wonder what is the 
>> performance difference between this technique with UST compared to 
>> gprof and callgrind (valgrind tool).
>>
>> For making this a real feature thought, it would be nice to save the 
>> address, not the string itself. Could it be convenient to save in the 
>> trace the function name string the first time the function is hit, 
>> then the symbol table at analysis time would not be required?
>>
>> Anyway, very interesting stuff!
>>
>> Francis
>>
>>
>> _______________________________________________
>> ltt-dev mailing list
>> ltt-dev at lists.casi.polymtl.ca
>> http://lists.casi.polymtl.ca/cgi-bin/mailman/listinfo/ltt-dev
> _______________________________________________
> ltt-dev mailing list
> ltt-dev at lists.casi.polymtl.ca
> http://lists.casi.polymtl.ca/cgi-bin/mailman/listinfo/ltt-dev