This is the mail archive of the libstdc++@gcc.gnu.org mailing list for the libstdc++ project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Dead code (was: Re: Good numbers from Ritter's new string allocator)


Benjamin Kosnik wrote:

> > In fact, there is at least one well known and modern implementation of
> > the ISO stdlib which does not grow exponentially either, and, moreover,
> > does not round to the block. But I agree with you (and with Nathan) that
> > some form of exponential shaping should be implemented, sooner or later.
>
> Sooner.

Ack.

> > Seriously, I believe that at this point we (i.e., me, you, Benjamin and
> > Nathan) all agree that the current patch (with the amended Changelog)
> > could be a good starting point for the mainline. I think your approach
> > is very simple yet guarantees a big improvement on the current
> > situation. In my tests it's beaten in speed only by truly exponential
> > allocators.
>
> Yep. Something, even if it's not perfect, needs to go into mainline ASAP.
> The patch for the branch can wait for it to settle, if necessary.
>
> Paolo, the configure machinery just looks complicated. I think it's
> actually pretty easy to figure out after a bit of
> autoconf/automake/libtool hacking, and (I think) it's commented decently.
>
> You might want to start by reading the GNU Autotools book:
>
> http://sources.redhat.com/autobook/

Do you believe me if I tell you that a cartaceous copy of the book is on my desktop since
yesterday??

Thanks for your feedback,
Paolo.



Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]