Comments on: STL lower_bound: set vs. list http://www.sambal.org/2008/10/stl-lower_bound-set-vs-list/ Trying to avoid labels Thu, 11 Mar 2010 22:12:41 +0000 http://wordpress.org/?v=2.7.1 hourly 1 By: Alex Stapleton http://www.sambal.org/2008/10/stl-lower_bound-set-vs-list/comment-page-1/#comment-85 Alex Stapleton Sat, 04 Oct 2008 14:27:31 +0000 http://www.sambal.org/?p=724#comment-85 It would be interesting to see these tests run on hardware that was more typical back around 2000. Presumably the algorithm used was chosen for a reason at the time? There doesn't seem to be any reason that lower_bound couldn't be specialised for _List_iterator in the GNU (based on SGI) implementation. You could probably even specialise it yourself without having to patch your vendors implementation. It would be interesting to see these tests run on hardware that was more typical back around 2000. Presumably the algorithm used was chosen for a reason at the time?

There doesn’t seem to be any reason that lower_bound couldn’t be specialised for _List_iterator in the GNU (based on SGI) implementation. You could probably even specialise it yourself without having to patch your vendors implementation.

]]> By: attractivechaos http://www.sambal.org/2008/10/stl-lower_bound-set-vs-list/comment-page-1/#comment-84 attractivechaos Fri, 03 Oct 2008 09:22:19 +0000 http://www.sambal.org/?p=724#comment-84 Sorry, I made a wrong conclusion again. You are right. I realize that insertion on a red-black tree is only amortized O(1), not the worst case. Thanks for the clarification. Sorry, I made a wrong conclusion again. You are right. I realize that insertion on a red-black tree is only amortized O(1), not the worst case. Thanks for the clarification.

]]> By: attractivechaos http://www.sambal.org/2008/10/stl-lower_bound-set-vs-list/comment-page-1/#comment-83 attractivechaos Fri, 03 Oct 2008 09:15:38 +0000 http://www.sambal.org/?p=724#comment-83 It is not clear to me that whether wiki is saying insertion after finding the postion requires O(log(N)) time. The overall insertion time is no doubt O(log(N)). Please see the discussion here: http://discuss.fogcreek.com/joelonsoftware/default.asp?cmd=show&ixPost=22948 This page looks more professional. In addition, if you look at the source code, rebalancing after insertion only need to fix red nodes around the insertion point. There are at most two such nodes in a red-black tree after immediate insertion. Rebalancing after deletion need to fix black nodes above the deletion point. This can be O(log(N)). It is not clear to me that whether wiki is saying insertion after finding the postion requires O(log(N)) time. The overall insertion time is no doubt O(log(N)).

Please see the discussion here:

http://discuss.fogcreek.com/joelonsoftware/default.asp?cmd=show&ixPost=22948

This page looks more professional. In addition, if you look at the source code, rebalancing after insertion only need to fix red nodes around the insertion point. There are at most two such nodes in a red-black tree after immediate insertion. Rebalancing after deletion need to fix black nodes above the deletion point. This can be O(log(N)).

]]> By: Sam http://www.sambal.org/2008/10/stl-lower_bound-set-vs-list/comment-page-1/#comment-81 Sam Thu, 02 Oct 2008 22:16:27 +0000 http://www.sambal.org/?p=724#comment-81 Amortized O(1), worst-case is O( lg N ) because you may have to make up to lg N rotations to restore the red-black property. See e.g, CLR 1st ed p. 268, or Wikipedia: http://en.wikipedia.org/wiki/Red-black_tree#Complexity Amortized O(1), worst-case is O( lg N ) because you may have to make up to lg N rotations to restore the red-black property. See e.g, CLR 1st ed p. 268, or Wikipedia: http://en.wikipedia.org/wiki/Red-black_tree#Complexity

]]> By: attractivechaos http://www.sambal.org/2008/10/stl-lower_bound-set-vs-list/comment-page-1/#comment-80 attractivechaos Thu, 02 Oct 2008 22:07:18 +0000 http://www.sambal.org/?p=724#comment-80 On a red-black tree, once you find the insertion position, insertion can be done in O(1) time, not O(log(N)). On a red-black tree, once you find the insertion position, insertion can be done in O(1) time, not O(log(N)).

]]>