Re: [PATCH 0/9] perf: Adding better precise_ip field handling

[Peter Zijlstra]

On Thu, May 09, 2013 at 05:20:22PM +0200, Jiri Olsa wrote:
> On Thu, May 09, 2013 at 05:07:44PM +0200, Peter Zijlstra wrote:
> > On Thu, May 09, 2013 at 03:32:15PM +0200, Jiri Olsa wrote:
> > > hi,
> > > adding sysfs attribute to specify the maximum allowed value
> > > for perf_event_attr::precise_ip field.
> > > 
> > > Adding functionality for simple 'p' modifier and 'precise' term
> > > to get the maximum allowed value for perf_event_attr::precise_ip
> > > field.
> > > 
> > 
> > You've seem to lost the part explaining why we want this.. :-)
> 
> well, initially it was an answer when we broke precise event
> monitoring in kernel so I wrote automated test for it (patches 1,2,3)
> https://lkml.org/lkml/2012/12/12/561

But those don't rely on the max thing right?

> having maximum precise enabled with just single 'p' seemed
> like good idea

Doesn't seem like to me; that takes away the possibility to use less.

> next step would be to enable precise automatically for 'cycles'
> (when PEBS is working) asked for by Ingo
> http://marc.info/?l=linux-kernel&m=135929050803963&w=2

Hurm.. I'm of two minds there. As Stephane has been pointing out for ages,
cycles behaves significantly different between regular and PEBS events for some
cases.

Also, you really don't need the max_precise for that either. At worst you'll
have a number of unsuccessful event creations.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/