Re: [Dev-C++] itoa()

[Per W. <pw...@ia...>]

Just for the fun of it, I tested the previous code you sent.

One million calls (i=0..999999) resulted in:
Ioannis with malloc              itoa(int)               4536 ms
Ioannis with auto-buf            itoa(int,char*)          661 ms
itoa using sprintf and auto-buf  itoa(int,char*)         1512 ms
DevStudio built-in and auto-buf  _itoa(int,char*,int)     481 ms

The malloc()/free() code took 6.9 times longer than if the buffer was sent
in as a parameter - both alternatives are threadsafe. 85% of the runtime
time was malloc()/free().

That's why I try and keep an eye on heap allocations, and try for stack
buffers or buffer pools.

/Per W


On Thu, 3 Apr 2003, Ioannis Vranos wrote:

> Ok, in order to be complete (with what I consider as an itoa() without any
> specific constraint requirements) I repost a complete itoa(), with the few
> bugs fixed. The implementation was influenced by the spirit "(Real
> programmers) Don't comment their code. If it was hard to write, it should be
> hard to
> understand and even harder to modify." See the attached file.
>
>
> #include <stddef.h>
> #include <stdlib.h>
> #include <stdio.h>
>
>
> char *itoa(int i)
> {
>   size_t Length;
>   char *p;
>   int temp=i;
>
>   int sign=temp<0?-1:1;
>
>   if(!sign)
>   {
>     p=malloc(2);
>
>     if(!p)
>       return NULL;
>
>     p[0]='0', p[1]=0;
>
>     return p;
>    }
>
>   temp=sign*i;
>
>   for(Length=0; temp/=10; Length++)
>       ;
>
>   p=malloc(Length+1+!sign==1);
>
>   if(!p)
>     return NULL;
>
>    sprintf(p, "%d", i);
>
>   return p;
> }
>
>
>
> Ioannis Vranos
>
> * Programming pages: http://www.noicys.freeurl.com
> * Alternative URL: http://run.to/noicys
>