GSoC Project | Basic Support for Trace Compass

[Guillaume Champagne]

Hi Ravindra,

If you look at the dependencies and build process of lttng-tools [1]  
and lttng-modules [2] you'll see that both projects are very tied to  
linux. Lttng stands for linux tracing toolkit next generation after  
all :).

An idea to produce CTF could be to reuse the binary format already  
existing in the Capture Engine and write a CTF metadata file to  
explain the format to the CTF parser. CTF is very flexible and we can  
make it understand a wide range of binary formats. The CTF spec has a  
few examples on that at the end [3]. Trace Compass only cares about  
the event types in the trace and their content, not their actual  
binary representation (they just have to be decodable using the CTF  
metadata file). I had a working prototype for CTF traces in RTEMS here  
[4], but I don't have the metadata file I used at hand.

Feel free to come by the tracecompass or lttng irc channels if you  
have any questions regarding Trace Compass or CTF!

[1] https://github.com/lttng/lttng-tools
[2] https://github.com/lttng/lttng-modules
[3] https://diamon.org/ctf/#examples
[4] https://github.com/gchamp20/rtems/commits/ctf

Guillaume

Ravindra Kumar Meena <rmeena840 at gmail.com> a écrit :

> Hi Sebastian,
>
> I was exploring the documentation of LTTng and babeltrace and come to the
> conclusion that LTTng already produces trace data in CTF format. Since
> Trace Compass already supports CTF trace data. So, what I am thinking that
> instead of converting trace data to CTF why don't we just directly produce
> trace data in CTF using LTTng.
>
> https://lttng.org/docs/v2.10/#doc-tracing-your-own-user-application
>
>
> On Tue, May 14, 2019 at 12:45 PM Sebastian Huber <
> sebastian.huber at embedded-brains.de> wrote:
>
>> On 14/05/2019 08:34, Ravindra Kumar Meena wrote:
>> > On Tue, May 14, 2019 at 11:24 AM Sebastian Huber
>> > <sebastian.huber at embedded-brains.de
>> > <mailto:sebastian.huber at embedded-brains.de>> wrote:
>> >
>> >     On 13/05/2019 08:24, Ravindra Kumar Meena wrote:
>> >     >
>> >     > *The goal of this week:*
>> >     >
>> >     > The goal of phase 1 of my GSoC proposal is to generate the trace
>> >     > data(LTTng) in Common Trace Format(CTF) from target running RTEMS
>> >     > application. The Event Recording infrastructure lacks the
>> >     generation
>> >     > of trace data in Common Trace Format(CTF). I will try to figure
>> >     out a
>> >     > convenient method for it.
>> >
>> >     Please note that there are at least two options where you can do the
>> >     conversion, you can do the conversion on the target or you can do the
>> >     conversion on the host.
>> >
>> >
>> > In my point of view, the conversion should take place on the host side
>> > because the host is more powerful. It would reduce the conversion time.
>>
>> Yes, I would try this a approach first. However, please keep in mind
>> that RTEMS runs also on 1.5GHz chips with 24 cores. This means the
>> conversion must be fast also on a modern host computer if you want to
>> support the high end RTEMS systems.
>>
>> --
>> Sebastian Huber, embedded brains GmbH
>>
>> Address : Dornierstr. 4, D-82178 Puchheim, Germany
>> Phone   : +49 89 189 47 41-16
>> Fax     : +49 89 189 47 41-09
>> E-Mail  : sebastian.huber at embedded-brains.de
>> PGP     : Public key available on request.
>>
>> Diese Nachricht ist keine geschäftliche Mitteilung im Sinne des EHUG.
>>
>>
>
> --
> *Ravindra Kumar Meena*,
> B. Tech. Computer Science and Engineering,
> Indian Institute of Technology (Indian School of Mines)
> <https://www.iitism.ac.in/>, Dhanbad