Re: [Patch] tr1::hashtable::operator[] (final)

[Peter Doerfler <gcc at pdoerfler dot com>]

Paolo Carlini wrote:
> Hi,
>
> eventually, I decided to go for something simpler (and performance-wise 
> more effective), leaving the insertion with hint alone (maybe I will 
> reconsider it soon): I'm just splitting out the second part of 
> insert(..., true_type) and calling it from operator[]. In this form 
> everything is pretty straightforward and is suitable for 4_1-branch too.
It's also a tiny bit faster with my real use case.

I think a small comment like
// reuse hash code and and current bucket for hint-like insertion
might help to understand what's going on.

A general question: Will these patches find their way into the 
4_0-branch as well? I'm asking since I'd like to use unordered_map which 
is now much faster in our project. However, our users probably just use 
distros and those might consider to use 4.0.4 in the future... Well, the 
more conservative ones might, anyway.


> P.S. Peter, for your future contributions, do you have a Copyright 
> Assignment on file?!?

Unfortunately, no. I didn't think I'd make any code contributions ;-) 
Right now I'm not affiliated with any company/university. Am I still 
required to file the assignment?

Thanks for the great job!
Peter