Re: [PATCH 00/12] perf: introduce model specific events and AMD IBS

[Robert Richter <robert.richter@amd.com>]

On 15.04.10 09:44:21, Peter Zijlstra wrote:
> On Tue, 2010-04-13 at 22:23 +0200, Robert Richter wrote:
> > This patch series introduces model specific events and impments AMD
> > IBS (Instruction Based Sampling) for perf_events.
> 
> I would much rather it uses the ->precise thing PEBS also uses,
> otherwise we keep growing funny arch extensions and end up with a
> totally fragmented trainwreck of an ABI.

I agree that an exiting flag could be reused. But the naming 'precise'
could be misleading. Maybe we rename it to 'model_spec' or something
else that underlines the idea of having model specific setups.

> > The general approach is to introduce a flag to mark an event as model
> > specific. With that flag set a model specific ibs (raw) config value
> > can be passed to the pmu for setup. When there are ibs samples
> > available, it is sent back as a raw data sample to the userland. So we
> > have a raw config value and raw sampling data. This requires the
> > userland to setup ibs and further extract and process sampling data.
> > 
> > Patches 1-8 rework and refactor the code to prepare the ibs
> > implementation. This is done in patches 9-12.
> > 
> > I will add ibs example code to libpfm4.
> 
> Please add a valid usecase to tools/perf/ instead.

I will also provide an example for tools/perf but reusing libpfm4 was
a first quick solution for me.

-Robert

-- 
Advanced Micro Devices, Inc.
Operating System Research Center
email: robert.richter@amd.com