[dpdk-dev] [PATCH v3 22/37] doc: remove references to make in contributing guides

Power, Ciara ciara.power at intel.com
Tue Sep 22 10:39:49 CEST 2020


Hi Cristian, Thomas,


>-----Original Message-----
>From: Dumitrescu, Cristian <cristian.dumitrescu at intel.com>
>Sent: Monday 21 September 2020 20:14
>To: Thomas Monjalon <thomas at monjalon.net>; Power, Ciara
><ciara.power at intel.com>
>Cc: dev at dpdk.org; Mcnamara, John <john.mcnamara at intel.com>; Kovacevic,
>Marko <marko.kovacevic at intel.com>; Richardson, Bruce
><bruce.richardson at intel.com>
>Subject: RE: [dpdk-dev] [PATCH v3 22/37] doc: remove references to make in
>contributing guides
>

<snip>

>> > >[...]
>> > >> -Library Statistics
>> > >> -------------------
>> > >> -
>> > >> -Description
>> > >> -~~~~~~~~~~~
>> > >> -
>> > >> -This document describes the guidelines for DPDK library-level
>> > >> statistics counter -support. This includes guidelines for turning
>> > >> library statistics on and off and -requirements for preventing
>> > >> ABI changes when implementing statistics.
>> > >
>> > >I think the ABI part is still relevant.
>> >
>> > Looking for some clarification on this. Which parts (if any) of the
>> > library
>> statistics section should be
>> > added back in for the ABI section below to make sense being
>> > included? Or
>> what do you suggest doing here?
>>
>> I think we can consider we still have the ability to turn stats off
>> because of the #ifdef still being there.
>> Cristian, do you think we should remove this doc?
>> Is on/off still possible in the new pipeline libs?
>>
>
>Hi Ciara and Thomas,
>
>IMO this section is still relevant and it is unrelated to the topic of this patch
>set, why do we want to remove it?
>
>From the make removal point of view, we only need the following change,
>and everything else is still logically valid:
>-   # DPDK file config/common_linux, config/common_freebsd, etc.
>-   CONFIG_RTE_<LIBRARY_NAME>_STATS_COLLECT=y/n
>+   /* Library API header file. */
>+  #define RTE_<LIBRARY_NAME>_STATS_COLLECT
>
>On the larger stats topic, I think the section is still relevant, especially from
>the point of view of avoiding ABI breakage due to stats, which is applicable to
>several libraries, so I suggest to keep it.
>
>In the new pipeline libs (rte_swx_pipeline.h API), all the stats are enabled by
>default.
>
>Regards,
>Cristian
>

I agree the Library Statistics is still relevant, having looked at it again since sending v1.
I sent a v5 of this patchset yesterday just before your feedback here, and kept the full
section but modified the instructions on how to enable/disable, removing the mention
of CONFIG options. Please have a look at those changes here: https://patchwork.dpdk.org/patch/78208/

Thanks,
Ciara
 



More information about the dev mailing list