[quagga-dev 10209] Re: [PATCH 2/4] hash: force size to be a power of 2

Certain, Andrew certain at amazon.com
Wed Jan 16 17:22:58 GMT 2013



> -----Original Message-----
> From: Stephen Hemminger [mailto:stephen at networkplumber.org]
> Sent: Wednesday, January 16, 2013 8:04 AM
> To: Certain, Andrew
> Cc: David Lamparter; Stephen Hemminger; quagga-dev at lists.quagga.net
> Subject: Re: [quagga-dev 10194] Re: [PATCH 2/4] hash: force size to be a
> power of 2
> 
> > It seems that you're trading one performance improvement against
> another.  The hashing performance of the thread look up will be worse, and
> you're arguing that this performance hit isn't as important as removing
> the divide.  You may be right, but I would like to see the data.
> >
> > Obviously, my proposal would be just as bad as yours in impacting the
> hashing of the threads.  Without a coherent and comprehensive test suite,
> no changes can be made with high confidence, which in my mind raises the
> bar for when we should accept change.  Until such a thing exists, I think
> that any changes done solely for performance reasons should only be
> accepted if there is a clear problem with the current code, rather than
> just for general cleanup reasons (I'm not saying you don't have a
> quantitative argument; I'm just asking you to share it).
> >
> > Andrew
> 
> The simpler fix would be to just use a mix (like jhash) on the function
> address.
>
OK, but why change anything?  Do you have data to show that this is an impacting concern?

Andrew




More information about the Quagga-dev mailing list