Talk:Vigenère cipher/Cryptanalysis: Difference between revisions
Content added Content deleted
(→Faster?: Optimization thoughts.) |
|||
| Line 30: | Line 30: | ||
:: The >> for templates is a C++0x fix. Before that, it was a syntax error, but I guess g++ does it half way even without the -std flag. [[User:MagiMaster|MagiMaster]] 09:55, 21 June 2011 (UTC) |
:: The >> for templates is a C++0x fix. Before that, it was a syntax error, but I guess g++ does it half way even without the -std flag. [[User:MagiMaster|MagiMaster]] 09:55, 21 June 2011 (UTC) |
||
:In this program frequency() is a critical function. The length of the vector of pairs 'result' is known at compile time. Avoiding this push_back speeds up this program more than 20%. This optimization doesn't make this program longer or more complex. |
:In this program frequency() is a critical function. The length of the vector of pairs 'result' is known at compile time. Avoiding this push_back speeds up this program more than 20%. This optimization doesn't make this program longer or more complex. |
||
:: It's a good optimization, but a better one might be to avoid allocation of the vector on every function call. Making 'results' function static, passed in by reference, or even file static would all be reasonable options. There's no threading in this program, so thread safety isn't a concern. 'results' is completely recalculated on each call, so there's no need to clear out its contents between calls, either. Finally, it's probably also perfectly appropriate for it to be a pair<char,double> array, rather than a vector, though that's less significant as an optimization as long as per-call reallocation is avoided. --[[User:Short Circuit|Michael Mol]] 18:15, 21 June 2011 (UTC) |
|||