Ingo,
On 30.12.11 10:55:21, Ingo Molnar wrote:
>
> * Robert Richter <[email protected]> wrote:
>
> > > Once something like 'perf report' works for some useful
> > > looking sub-case you can also tack on all the extra IBS
> > > goodies that specialized tools might need as well, i have no
> > > problem with providing that.
> >
> > Isn't that outlook I gave about pebs like implementation of
> > the precise bit with IBS "useful enough"? This would be on top
> > patches anyway, so no need to hold back current patches.
>
> An 'outlook' is not something I can test before sending commits
> to Linus.
The question was more if this is the direction you think may make
sense to go to. Of course, I would then implement and send patches for
this.
> To help your effort we could tentatively put the patches into a
> separate tip:perf/ibs topic branch if everyone agrees otherwise,
> which can go upstream once it's complete and usable.
This would be very helpful for me. So far I see following patches for
this on top of today's tip/perf/core (623ec99):
08e232d perf, x86: Implement IBS event configuration
05efba4 perf, x86: Implement IBS interrupt handler
b0532b2 perf, x86: Implement IBS pmu control ops
56ccf1a perf, x86: Implement 64 bit counter support for IBS
98e655c perf tool: Parse general/raw events from sysfs
aac22da perf tools: Add pmu mappings to header information
f1f062d perf script: Add script to collect and display IBS samples
Only 'perf tools: Add pmu mappings to header information' applies with
conflicts, so I could resend an updated version. This patch is also
independent from the other patches and can be applied separately.
If you wish I also can send all patches again, but they would be
unchanged anyway (except that one with the conflict).
Peter and Arnaldo,
please agree with a separate tip:perf/ibs development branch. It would
be easier to all of us because we don't have the effort of reviewing
patches again that will not change anymore.
Thanks,
-Robert
--
Advanced Micro Devices, Inc.
Operating System Research Center