Re: [tracecompass-dev] Question about importing a multi-cpu perf capture

[Matthew Khouzam <matthew.khouzam@xxxxxxxxxxxx>]

Hi,

I think I know what caused it, the CPU field doesn't know that perf_cpu
is the right field. I can fix it but it would be much easier if I have a
trace.

Are you allowed to post your trace on bugs.eclipse.org? Can we add it as
a regression test? Is there anything confidential/identifyable in it you
want to strip out?

I can fix this... it is not difficult, but it will not be a patch for
1.1, it can go in 1.2, 2.0 and nightly, is that ok?

PS. I can also show you how to fix it if you want to contribute the code. :)

BR,
Matthew

On 15-12-18 09:46 AM, Jerome CORRENOZ wrote:
>
> Hi,
>
>  
>
> I’m starting to use TraceCompass to import a perf trace.
>
>  
>
> I have a problem with multi-CPU. Indeed, the trace has been captured
> on a multi-CPU but all the events seem to be associated to CPU0 (see
> the table view, column “CPU”) whereas the “Contents” correctly
> contains the “perf_cpu=X”. This impacts the “Resources” view where
> only CPU0 is displayed but mainly the “Control Flow” view.
>
>  
>
> I’m using:
>
> -    perf 4.1.1 on ARMv7
>
> -    TraceCompass 1.1.0
>
>  
>
> Doing the following:
>
>  
>
> #perf record –e ‘sched :*’ –a
>
> #perf data convert –to-ctf=./ctf
>
> Then Open the ctf file (called perf_stream_0) in TraceCompass
>
>  
>
> Any ideas about the cause of it ?
>
>  
>
> BR,
>
>   Jerome
>
>  
>
>  
>
>
>
> _______________________________________________
> tracecompass-dev mailing list
> tracecompass-dev@xxxxxxxxxxx
> To change your delivery options, retrieve your password, or unsubscribe from this list, visit
> https://dev.eclipse.org/mailman/listinfo/tracecompass-dev