optimization - Optimizing I/O(Output) in C code + a loop

Question

I have a code which reads around (10^5) int(s) from stdin and then after performing ## i output them on stdout. I have taken care of the INPUT part by using "setvbuf" & reading lines using "fgets_unlocked()" and then parsing them to get the required int(s). I have 2 issues which i am not able to come over with:

1.) As i am printing int(s) 5 million on stdout its taking lot of time : IS THERE ANY WAY TO REDUCE THIS( i tried using fwrite() but the o/p prints unprintable characters due to the reason using fread to read into int buffer)

2.) After parsing the input for the int(s) say 'x' i actually find the no of divisors by doing %(mod) for the no in a loop.(See in the code below): Maybe this is also a reason for my code being times out: Any suggestions on this to improved. Many thanks This is actually a problem from http://www.codechef.com/problems/PD13

# include <stdio.h>
# define SIZE 32*1024
char buf[SIZE];

main(void)
{
int i=0,chk =0;
unsigned int j =0 ,div =0;
int a =0,num =0;
char ch;

setvbuf(stdin,(char*)NULL,_IOFBF,0);

scanf("%d",&chk);
while(getchar_unlocked() != '
');
while((a = fread_unlocked(buf,1,SIZE,stdin)) >0)
{
    for(i=0;i<a;i++)
    {
        if(buf[i] != '
')
        {
            num = (buf[i] - '0')+(10*num);
        }
        else
        if(buf[i] == '
')
        {   
            div = 1;
            for(j=2;j<=(num/2);j++)
            {
                if((num%j) == 0)    // Prob 2
                {
                    div +=j;
                }
            }
            num = 0;
            printf("%d
",div); // problem 1
        }       
    }
}
return 0;
 }

Answer 1

You can print far faster than printf.

Look into itoa(), or write your own simple function that converts integers to ascii very quickly.

Here's a quick-n-dirty version of itoa that should work fast for your purposes:

char* custom_itoa(int i)
{
    static char output[24];  // 64-bit MAX_INT is 20 digits
    char* p = &output[23];

    for(*p--=0;i/=10;*p--=i%10+0x30);
    return ++p;    
}

note that this function has some serious built in limits, including:

• it doesn't handle negative numbers
• it doesn't currently handle numbers greater than 23-characters in decimal form.
• it is inherently thread-dangerous. Do not attempt in a multi-threaded environment.
• the return value will be corrupted as soon as the function is called again.

I wrote this purely for speed, not for safety or convenience.