Talk:K-d tree: Difference between revisions
→Dubious about speed claims here: new section
(→C Entry: NULL is preferred) |
(→Dubious about speed claims here: new section) |
||
Line 70:
:: FWIW, I agree strongly with using <code>NULL</code>; it's more ''idiomatic'' and says that “we're thinking about the pointer that does not point”. It's really a code-smell thing; if someone's mixing things up, you've got to examine ''every'' use of <code>0</code> to figure out what's going on. (This isn't the same as when you're working in your own code, but we want the very best of style here as each language understands that concept.) –[[User:Dkf|Donal Fellows]] 08:25, 5 June 2012 (UTC)
== Dubious about speed claims here ==
After studying the task and these implementations I'm feeling dubious about speed claims here.
If it takes half a second to construct the tree for a million nodes, and it takes a twentieth of a second to visit a million nodes, what kinds of speedups are we getting that justifies this use pattern?
More specifically: do any of these implementations outperform a exhaustive search on a dense linear dataset by as much as a factor of 2? --[[User:Rdm|Rdm]] ([[User talk:Rdm|talk]]) 20:05, 5 June 2014 (UTC)
| |||