相同的代码在不同的机器上具有截然不同的运行时间

Question

If I run this code

#include <iostream>
#include <string>
#include <stdlib.h>
#include <vector>
#include <time.h>
#include <thread>
#include <ctime>
using namespace std;
int main()
{
    int start_s = clock();
    char randChar;
    string random;
    random.clear();
    int entry_size = 4;
    int population_size = 500000;
    vector<string> population;
    for (int i = 0; i < population_size; i++) {
        for (int i = 0; i < (unsigned)entry_size; i++) {
            randChar = rand() % 25 + 97;
            random += randChar;
        }
        //cout << random<<endl;
        population.push_back(random);
        random.clear();
    }
    int stop_s = clock();
    cout<< (stop_s - start_s)/(double)(CLOCKS_PER_SEC);
}

on this web site: http://cpp.sh/ the running time is around 0.16 seconds however if I compile and run this on my home machine (i5 4460 16 gb ram, mechanical HDD,visual studio 2017) the running time is around 6.6 seconds which is around 41 times slower, what is causing such a huge speed difference?
thanks

Answer 1

My psychic powers suggest that you built a Debug build on your home machine with Visual Studio. Flip the Solution to a Release build and watch the performance go way up.

I just tested this with Visual Studio. The difference between Debug and Release is 100x with your code. (8 seconds vs .08 seconds).

It's not that Debug builds are inherently slow. It's just that really tight loops doing math, memory, or anything non-blocking can be highly optimized.

Answer 2

different mechanisms to deal with memory management perhaps? it's hard to pinpoint exactly as there can be quite a lot of assumptions, but there are ways to make your code faster on both ends.

#include <iostream>
#include <string>
#include <stdlib.h>
#include <vector>
#include <time.h>
#include <thread>
#include <ctime>
using namespace std;
int main()
{
    int start_s = clock();
    char randChar;
    string random;

no need to clear something that's empty.

    int entry_size = 4;
    int population_size = 500000;
    vector<string> population;

pre-alloc the vector, 500000 is a huge number and the vector would have to do many allocations.

    population.resever(population_size);

also, pre-alloc the string random, and don't clear it every loop. the amount of memory is small but allocating and unallocating is expensive. instead change the values of the preallocated string dinamically.

    random.reserve(entry_size);
    for (int i = 0; i < population_size; i++) {

you used 'i' as the keys for the inner and the outer for loops, this looks like a bug, I'm changing the index of the inner loop to j.

        for (int j = 0; j < (unsigned)entry_size; j++) {
            randChar = rand() % 25 + 97;
            random[j] = randChar;
        }
        population.push_back(random);
    }
    int stop_s = clock();
    cout<< (stop_s - start_s)/(double)(CLOCKS_PER_SEC);
}

This should make your app quite faster.

Answer 3

我认为这个因素中的一些会产生影响：CPU，操作系统，内存大小，I / O速度（不是因为你的代码没有任何IO），内存缓存，并行化。