Re: noverlay branch

[Gerd Möllmann]

Matt Armstrong <matt@rfc20.org> writes:

>> Clang libstd++ uses NULL, BTW, and I already wondered a little bit why.
>
> I believe GNU libstdc++ does not use sentinel nodes either.  I have yet
> to see see sentinel nodes used in an optimized tree implementation.

I looked it up a few days ago, out of interest, and GCC's rb-tree uses
sentinels, in Git master at least.  Somewhere in this thread I posted
where to look in GCC's Git repo, I can't remember ATM.

(And I can't stop saying that it would be a big step forward for Emacs
to allow C++.  With std::multimap/std::multiset, we would have a
ready-made complete solution for the tree, tested by a gazillion of
users.  Just dreaming :-))

> I think in the context of this overlay work the performance difference
> is not very significant, since the code is doing a lot of other stuff
> while traversing the tree.

I agree.  I think NULL could be better in multi-threaded cases, as
Stefan alluded to.