strncpy需要很长时间

Question

I have recently encountered some legacy code and noticed that sometimes (not always) the following code snippet takes a very long time to execute.

#define NUM_OF_RECORDS 100000
char* pzBuffer = new char[NUM_OF_RECORDS + 1];
strncpy(pzBuffer, "", NUM_OF_RECORDS);

The objective seems to be to initialize the pzBuffer and I agree that memcpy or memset would be a better choice. I'm trying to figure out why it's taking time and if the usage here is correct.

Answer 1

using memset() could do better due to the following reason:

1) strncpy() works at individual 'char' level meaning it would iterate over all chars one by one ( ~NUM_OF_RECORDS iterations in your case).

2) memset() works at block level where a block consists of say N bytes. Although this means going over individual bytes of a block, this function is designed for optimizations wrt large block sizes, which often result in better performance when blocks of large sizes are involved.

I would recommend creating variants of your code 1)using memset 2)other with strncpy. Then, compile both of them to generate assembly code, and find out the difference between instructions.

Answer 2

http://unix.superglobalmegacorp.com/BSD4.4Lite/newsrc/libkern/strncpy.c.html

loks like the standard implementation of strncpy fills all remaining elements of the array with zeros, no matter what. So, it takes time doing that,.

    do {
        if ((*d++ = *s++) == 0) {
            /* NUL pad the remaining n-1 bytes */
            while (--n != 0)
                *d++ = 0;
            break;
        }
    } while (--n != 0);