If I have two functions copyij() and copyji(), to copy a 2048x2048 integer array, which one is faster and why?

Asked May 03 '20 at 00:22

Active May 03 '20 at 00:22

Viewed 67 times

This is in C++, and I used the time command on Unix and found that copyij() was significantly faster.

void copyij(long int src[MSize][MSize], long int dst[MSize][MSize])
{
  long int i,j;
  for (i = 0; i < MSize; i++)
    for (j = 0; j < MSize; j++)
      dst[i][j] = src[i][j];
}

copyji() is similar, except the j loop is before the i-loop.

asked May 03 '20 at 00:22

namu june

What compilation flags did you use? Especially the optimizing ones? – XCS May 03 '20 at 00:24
1

And an even faster method: `memcpy(&dst[0][0], &src[0][0], sizeof(long) * MSize * MSize);` – PaulMcKenzie May 03 '20 at 00:25
Probably because the first one works better with [caching](https://en.wikipedia.org/wiki/CPU_cache). – eesiraed May 03 '20 at 00:26
If you timed them, why are you asking which is faster? – Scott Hunter May 03 '20 at 00:27
4

Related: [Why is iterating 2D array row major faster than column major?](https://stackoverflow.com/questions/33722520/why-is-iterating-2d-array-row-major-faster-than-column-major) – dxiv May 03 '20 at 00:27

If I have two functions copyij() and copyji(), to copy a 2048x2048 integer array, which one is faster and why?

0 Answers0