ashhwu / algorithm_sorting_part1

simple sorting problem

2016Algorithm_01

Simple sorting problem

##Init version

Analysis result:

	result
I_refs	111363392
check	1.000000/1.000000
cpu_time	30.162ms
flt	1364
gcc	0
heap_lost	0
illegal_func	0
illegal_header	0
m_error	0
m_peak	59457536
maxrss	2170880
@lang=cpp

##Insertion sort version

Analysis result:

	result
I_refs	5091383206
check	1.000000/1.000000
cpu_time	64.9071ms
flt	1113
gcc	0
heap_lost	0
illegal_func	0
illegal_header	0
m_error	0
m_peak	58482688
maxrss	1454080
@lang=cpp

##Final version

Analysis result:

	result
I_refs	94385289
check	1.000000/1.000000
cpu_time	24.234ms
flt	1108
gcc	0
heap_lost	0
illegal_func	0
illegal_header	0
m_error	0
m_peak	58478592
maxrss	1437696
@lang=cpp

(All with the same input, from small data to mass data)

#Conclusion

For the sorting algorithm, I used std::sort to test it. And I also test it with insertion sort which spent years to run it(O( n^2 )) and it's number of instructions consumed are too big, so I won't talk about it.

The first version's maximun resident size (maxrss) had much more bigger than Final version.

I did it just by discarding the stringstream and using a char to do the input. Although it seems clumsy, it does reduce the maxrss and also the CPU time.

So, if there are a great amount of data to be inputed, it is better to avoid using stingstream. Because stringstream use too mush memory to do the buffer things.

But there are still pros of using it, such as reading a line of seperated ints or strings until '\n'. And it can simply be done by getline() and split it by ss<<str; => ss>>(int or str);.

And the cons of using char and reading it digit by digit is that it is really indigenous, and it can cause many bugs.

This is just my rough conclusion.