2018-05-02 16:39:22

by Stephane Eranian

[permalink] [raw]
Subject: [RFC] perf record: dead code perf_event__synthesize_id_index()

Hi,

It looks like perf_event__synthesize_id_index() is dead code in the current
tip
tree. I don't see any invocation of the function anywhere in perf.I
understand why
you'd want to keep the rest of the code related to PERF_RECORD_ID_INDEX
because some perf.data file may have this user-level record type. But the
synthesize code is not needed anymore, in my opinion.

Shouldn't we remove this code?


2018-05-02 16:58:46

by Jiri Olsa

[permalink] [raw]
Subject: Re: [RFC] perf record: dead code perf_event__synthesize_id_index()

On Wed, May 02, 2018 at 04:38:33PM +0000, Stephane Eranian wrote:
> Hi,
>
> It looks like perf_event__synthesize_id_index() is dead code in the current
> tip
> tree. I don't see any invocation of the function anywhere in perf.I
> understand why
> you'd want to keep the rest of the code related to PERF_RECORD_ID_INDEX
> because some perf.data file may have this user-level record type. But the
> synthesize code is not needed anymore, in my opinion.
>
> Shouldn't we remove this code?

looks like there never was any user of it, probably some
preparation for follow up that haven't happened yet.. Adrian?

jirka

2018-05-03 06:17:33

by Adrian Hunter

[permalink] [raw]
Subject: Re: [RFC] perf record: dead code perf_event__synthesize_id_index()

On 02/05/18 19:57, Jiri Olsa wrote:
> On Wed, May 02, 2018 at 04:38:33PM +0000, Stephane Eranian wrote:
>> Hi,
>>
>> It looks like perf_event__synthesize_id_index() is dead code in the current
>> tip
>> tree. I don't see any invocation of the function anywhere in perf.I
>> understand why
>> you'd want to keep the rest of the code related to PERF_RECORD_ID_INDEX
>> because some perf.data file may have this user-level record type. But the
>> synthesize code is not needed anymore, in my opinion.
>>
>> Shouldn't we remove this code?
>
> looks like there never was any user of it, probably some
> preparation for follow up that haven't happened yet.. Adrian?

Yes it is for AUX area sampling support, which we have patches for. Just
waiting on Alex to get the kernel side accepted.