|
Boost : |
Subject: Re: [boost] [review] The review of Boost.DoubleEnded starts today: September 21 - September 30
From: Thorsten Ottosen (tottosen_at_[hidden])
Date: 2017-10-05 16:11:51
Den 05-10-2017 kl. 17:52 skrev Thorsten Ottosen via Boost:
> Den 04-10-2017 kl. 01:30 skrev Ion Gaztañaga via Boost:
>
>> --------------------------
>> batch_deque
>> --------------------------
>> 1) I like that deque_[const_]iterator is lightweight, I donât know if
>> iteration/access itâs more efficient than more traditional approaches.
>> boost::container::deque iterators are really fat and should be
>> improved to something similar to batch_deque iterator.
>
>
> I guess the simplest would be to replace the end iterator with a
> function that computes it when needed (I think that can be done easily
> since you now the segment size at compile time).
Note also that AFAICT, having the iterator store the pointer to the
segment pointer was the reason batch_deque's push_back was slower than
boost::deque (what could have been just the return of a pointer became
in indirection + an offset calculation).
As for the overall fatness of deque, then I would not care too much
about this. Many times you only need one.
kind regards
-Thorsten
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk