Difference b/w using i<strlen() and str[i] != '\0' - arrays

Question

When I use for(i=0;i<strlen(s);i++) then I am getting time limit exceed error. And When I use for(i=0;s[i]!='\0';i++) my code get successful submit. Why?
I am also providing link of question from codechef - https://www.codechef.com/problems/LCPESY
Type 1:
for (i = 0; i < strlen(s1); i++) {
f1[s1[i]]++;
}
for (i = 0; i < strlen(s2); i++) {
f2[s2[i]]++;
}
Type 2:
for (i = 0; s1[i] != '\0'; i++) {
f1[s1[i]]++;
}
for (i = 0; s2[i] != '\0'; i++) {
f2[s2[i]]++;
}
Complete code:
#include <stdio.h>
#include <string.h>
long int min(long int a, long int b) {
if (a >= b)
return b;
else
return a;
}
int main(void) {
// your code goes here
int t;
scanf("%d", &t);
while (t--) {
char s1[10001], s2[10001];
scanf("%s%s", s1, s2);
long int f1[200] = { 0 }, f2[200] = { 0 }, i, count = 0;
for (i = 0; i < strlen(s1); i++) {
f1[s1[i]]++;
}
for (i = 0; i < strlen(s2); i++) {
f2[s2[i]]++;
}
for (i = 0; i < 200; i++) {
count += min(f1[i], f2[i]);
}
printf("%ld\n", count);
}
return 0;
}

Answer 1

If a non-optimizing compiler is used it can be that strlen is re-evaluated once per each iteration. strlen then needs to check each and every character in the string for equivalence with 0. This results in quadratic runtime, where there are O(n²) checks for the terminatin null instead of just the necessary O(n) times. In the strlen code the timeout happens because it does perhaps 2,000,000 null checks and 10,000 other operations; the other code would do 2,000 null checks and those same 10,000 other operations and not time out.
However, this need not be a case. Due to the as-if rule, a C compiler can generate exactly equivalent machine for the cases
for (i = 0; i < strlen(s1); i++){
f1[s1[i]] ++;
}
and
for (i = 0; s1[i] != '\0'; i++) {
f1[s1[i]] ++;
}
because a compiler can easily prove that the inner loop cannot possibly change s1 and therefore both forms would behave equivalently.

Answer 2

In addition to #Antti Haapala good answer:`
Difference b/w using i<strlen() and str[i] != '\0'
Code like int i; ... i < strlen(s1) readily complains about mismatched sign-ness - when such warnings are enabled. Usually inoffensive code like that discourages wide use of that warning. I see that as a less preferred approach. str[i] != '\0' does not cause that warning.
Some other concerns
Prevent buffer overflow
char s1[10001], s2[10001];
// scanf("%s%s", s1, s2);
if (scanf("%10000s%10000s", s1, s2) == 2) {
// OK, success, lets go!
There are more than 200 characters.
// long int f1[200] = { 0 };
long int f1[256] = { 0 };
// or better
long int f1[UCHAR_MAX + 1] = { 0 };
Avoid a negative index
// f1[s1[i]]++;
f1[(unsigned char) s1[i]]++;
or use unquestionable unsigned types.
// char s1[10001];
unsigned char s1[10001];