Content deleted Content added
→ctz version: new section |
→ctz version: Replied |
||
Line 325:
I think the unreferenced section "Iterative version in C++ using ctz (count trailing zeros)" should also be removed. It makes vague claims of efficiency, but as far as I can tell these rely on hardware parallelism for ctz. If you look at [[Count trailing zeros]], the typical (x86) ctz instructions have limited bit width, so the claims to improved performance suffer from the same problem as the ARM assembly example, i.e. they are not asymptotic. [[User:Someone not using his real name|Someone not using his real name]] ([[User talk:Someone not using his real name|talk]]) 14:26, 2 March 2014 (UTC)
: Totally agreed, went ahead and [[WP:BOLD]]ly {{Diff|Binary GCD algorithm|597834943|597831576|deleted}} this section. Count trailing zeros has more to do with compilers using the instruction as part of the platform optimization, rather than with a C++ implementation which may or may not end up using that instruction. — [[User:Dsimic|Dsimic]] ([[User talk:Dsimic#nobold|talk]] | [[Special:Contributions/Dsimic|contribs]]) 17:44, 2 March 2014 (UTC)
| |||