I had to use recursion+strings for a problem in a contest and I stumbled upon this irregularity.
Sample Code 'A' with strings:
Using input=5
Results:
G++ 11 5.1.0 = 7987 ms, 2044 KB
G++ 14 6.2.0 = 889 ms, 1856 KB
------------------------------------
Sample Code 'B' without strings:
Using input=6
Results:
G++ 11 5.1.0 = 3026 ms, 2012 KB
G++ 14 6.2.0 = 2995 ms, 1868 KB
To test it yourself, head over to http://codeforces.net/problemset/customtest
Why is there such a huge difference when strings are used?
Do you know the difference between
void f(string s)andvoid f(string &s)?Yes. Are you trying to say that G++14 optimises it automatically?
I think it doesn`t. Because following code takes 171ms in both C++11 and C++14. Probably there is another way to optimize this.
I'm trying to say that you can use this
&and the code becomes significantly faster, as no new instances of strings will be created. Negative votes show that people don't understand it.I guess negative votes are because your answer has anything common with the question "why is there such a huge time difference".
The compiler can't change the declaration of f (it is illegal to go from
string stostring &s).I dont think so. You cant use
std::moveforsbecause you are callinggo(...)20 times with same arguments.std::movespoils string, it will be a problem to use it here.Or maybe I misunderstood what are you talking about?
Yes, I guess you are right. There has to be another trick done by the compiler.
L.E.: This isn't only for strings. The optimization is done for vectors (and I guess that for all containers) too.
In g++ version 5.4.1 both programs (-std=c++11 or -std=c++14) have the same speed. So it probably is just an optimization the compiler introduced in version 5.2-4.
It might be that on codeforces, G++14 code runs on a faster processor. Optimisations at compiler level don't make sense.
I believe it's due to a different compiler version, not 11 or 14 standard.
I have tested it on my computer on g++ 6.3 (both g++11 and g++14 ran around 1s) and g++ 4.9 (both ran around 1.3s). On ideone g++ and g++14 (both from version 6.3) run in the same time while g++ (but version 4.3.2) is slower (output is clock ticks).
In all these runs there was no difference between runtimes of g++, g++11 and g++14, while there were differences between versions. I didn't get to test 5.1 though.