[ovs-dev] [PATCH v10 1/5] dpif-netdev: implement function pointers/subtable
ian.stokes at intel.com
Wed Jul 10 15:54:38 UTC 2019
On 7/10/2019 4:40 PM, Van Haaren, Harry wrote:
>> -----Original Message-----
>> From: Stokes, Ian
>> Sent: Wednesday, July 10, 2019 4:30 PM
>> To: Van Haaren, Harry <harry.van.haaren at intel.com>; dev at openvswitch.org
>> Cc: i.maximets at samsung.com; malvika.gupta at arm.com
>> Subject: Re: [PATCH v10 1/5] dpif-netdev: implement function pointers/subtable
>> On 7/9/2019 1:34 PM, Harry van Haaren wrote:
>>> This allows plugging-in of different subtable hash-lookup-verify
>>> routines, and allows special casing of those functions based on
>>> known context (eg: # of bits set) of the specific subtable.
>>> Signed-off-by: Harry van Haaren <harry.van.haaren at intel.com>
>>> Tested-by: Malvika Gupta <malvika.gupta at arm.com>
>> Thanks for this harry, few minor comments below.
> Hey, cool I'll snip away irrelevant code sections for readability.
>> I've thought about whether an item should be added to the NEWS doc
>> (possibly not in this patch but perhaps later in the series), it's hard
>> to say as the changes are under the hood and not configurable by a user.
> Correct the user shouldn't identify the actual changes, except for small
> gains in performance if the specialized subtables are in use.
> I think it would be good to add a NEWS item, as people profiling OVS's functions
> will see different function names, and having called out the DPCLS Function Pointer
> changes, and why they are there would be beneficial.
> I will draft a separate patch for NEWS item - so we can keep it parallel to this patch set.
> Shout if that is not a good approach :)
I would think it could be added as part of patch 5 of the series as
that's when the performance increase is introduced? Anyway it's minor
enough that if there is not another revision of the series it can be
done upon commit.
More information about the dev