10236 - The Fibonacci Primes

[Alexander Denisjuk]

I like this problem very much, specially the pair

6879 --- 842347072

In fact the correspondent Fibonacci number begins with 84234707299956....
So, you really should have good accuracy. But how did the problemsetter know about this Fib number?

My experience in getting AC is the following: use long double, do not trust any built-in functions like exp, log, even sqrt and abs. Write everything by yourself.

By the way, the proof of the basic fact, i. e. GCD(F_k,F_n)=F_{GCD(k,n)} is not very hard. Just apply Euclide's algorithm GCD(F_{k+m},F_k)=GCD(F_{k+m}-F_k,F_k) and so on.

[minskcity]

Does anybody know how to compute first nine digits of 250000th fibonachi number?

[Quantris]

gah, this was tough because the precision is dependent on the compiler version.

tough cases include 9307 -- 495091651, 15086 -- 920759405, and 15520 -- 386616082.

At least, those are the three that I "special cased"

[Jan]

Got Acc...

Input:

Code: Select all

22000
21000
100
233

Output:

Code: Select all

208214663
139876647
516212329
117851144

Edit : The above cases are correct.

[micheal_sunny]

can anyone tell me how to solve the problem
i konw the algorithm which can gets n-th fib in lgn,but i met overflow problem when i want to calculate the all digits
is there any secret or formula which i dont konw,please tell me
Thanks for watch and reply

[TimeString]

Does anybody know how to compute first nine digits of 250000th fibonachi number?

You can use double.

[Jan]

Finally got accepted with a completely different approach (no floating point calculations). Thanks.

[mukeshtiwari]

Getting TLE. plz someone reply me , what is wrong with my code and algorithm .
according to problem description, all fibonacci primes will occur at prime postions so i calculated first 22000 primes . then i use the following formula (http://mathworld.wolfram.com/FibonacciNumber.html) F(n)=phi^n/sqrt(5) where phi=(1+sqrt(5))/2 and calculated first 9 digits as explained in the posts. There is some precison issue also in fun2() which gives wrong output for inputs 22000,6879. so how to avoid precision issue . here is my code .

Code: Select all

#include<cstdio>
#include<cmath>
	
long double a,c;
	double fun2(int n)
		{
			int k;
			long double v,d;
			for(k=0;;k++)
			 {
				v=(long double)n*log10l(a)-(log10l(c)+(long double)k);
				if(v<9)
					break;
			 }

			 d=powl(10,v);
                         d=floor(d+0.5);
			return d;
		}

	int fun1(int n)
		{
			int i;
			if(n==1)
				return 0;
			if(n==2)		
				return 1;
			if(n%2==0)
				return 0;
			for(i=3;i*i<=n;i+=2)
				if(n%i==0)
					return 0;
			return 1;
		}

	int main()
		{
			int prime[22010],count=0,i,n;
			long double fib[22010];
			a=(1+sqrtl(5))/2;
			c=sqrtl(5);
			for(i=1;;i++)
			   {
				if(fun1(i))
				 {
					prime[count++]=i;
					if(count==22000)
						break;
				 }
			   }
			for(i=0;i<22000;i++)
			 {
				fib[i]=fun2(prime[i]);
			 }

			while(scanf("%d",&n)==1)
				printf("%.0llf\n",fib[n-1]);
		}

[mukeshtiwari]

no response

[rio]

Code: Select all

   double fun2(int n)
      {
         int k;
         long double v,d;
         for(k=0;;k++)
          {
            v=(long double)n*log10l(a)-(log10l(c)+(long double)k);
            if(v<9)
               break;
          }

          d=powl(10,v);
                         d=floor(d+0.5);
         return d;
      } 

You don't need the k loop. This loop is causing TLE.
----
Rio

[mukeshtiwari]

so is it possible to get k in o(1) time or should i use binary search . plz suggest me how to calculate k because now i am exhausted with this problem.

[rio]

so is it possible to get k in o(1) time or should i use binary search . plz suggest me how to calculate k because now i am exhausted with this problem.

You could get k in O(1).
Hint:

Code: Select all

10^12.34 = 10^(8.34 + 4) = 10^8.34 * 10^4

-----
Rio

[mukeshtiwari]

thnkx rio .

[shiguowang]

prove : <<the art of computer programming>> by Donald E.Knuth chapter 1.2.8 Theorem A