[ovs-dev] [classifier-opt] hash: Introduce an implementation of murmurhash.
Ethan Jackson
ethan at nicira.com
Tue Aug 21 22:22:38 UTC 2012
> I'm not sure of the benefit. When/if we do that, we'll obviously need
> to add some code to deal with the one-to-three byte tail, and the
> obvious way to do that is to look at the URL.
Well I think a lot of developers (including my self) would just zero
out the remaining unnecessary bytes to end up with a full word.
Probably not a big deal either way though. I think it's fine to leave
it as is.
Ethan
>
>> I want to replace the implementation of flow_hash_symmetric_l4() with
>> this because it's in the fast path for bundle actions, and it's
>> current implementation is particularly inefficient. Any reason I
>> shouldn't?
>
> I actually tried to reformulate the Jenkins hash to work out better
> there, and failed, and that's why I started looking for a new hash in
> the first place, so yes I think that's appropriate.
>
>> Also, I'm going to start doing Acked-bys
>>
>> Acked-by: Ethan Jackson <ethan at nicira.com>
>
> Thanks.
More information about the dev
mailing list