Re: [PATCH v1 0/3] perf list: Remove duplicate PMUs

[John Garry]

On 01/08/2023 18:39, Ian Rogers wrote:
> > > I didn't write this behavior, it has pre-existed my contributions. I'm
> > > hoping to change the perf list behavior as we're seeing large server
> > > systems with getting on toward 100 PMUs, the events are replicated for
> > > each one and the perf list and testing behaviors are somewhat
> > > exploding in size.
> > Sure, that is why I was advised PMU kernel drivers event names to be
> > unique per PMU, so that we can add an event alias in a JSON and then
> > kernel events are matched and removed from perf list.
> >
> > I suppose that your changes are an alternative to the problem of
> > mushrooming kernel event list.
> Thanks John, yep this is going after that problem. Could I get a
> reviewed/acked/tested-by for these changes?

I'll try to help.

So we have:

Before:
```
$ perf list
...
  uncore_imc_free_running_0/data_read/               [Kernel PMU event]
  uncore_imc_free_running_0/data_total/              [Kernel PMU event]
  uncore_imc_free_running_0/data_write/              [Kernel PMU event]
  uncore_imc_free_running_1/data_read/               [Kernel PMU event]
  uncore_imc_free_running_1/data_total/              [Kernel PMU event]
  uncore_imc_free_running_1/data_write/              [Kernel PMU event]
```

After:
```
$ perf list
...
  uncore_imc_free_running/data_read/                 [Kernel PMU event]
  uncore_imc_free_running/data_total/                [Kernel PMU event]
  uncore_imc_free_running/data_write/                [Kernel PMU event]
```
How about keep listing uncore_imc_free_running_0 and the rest for perf 
list -v? Or something like that.

I find that the perf tool has lots of veiled tricks in terms of usage 
and describing events available and how to use them.

Thanks,
John