I wrote a small program to compute Fibonacci numbers:
#include <stdio.h>
int main()
{
int first, second, final;
first = 0;
second = 1;
printf("0\n1\n"); /* I know this is a quick fix, but the program still works */
while (final <= 999999999999999999) {
final = first + second;
first = second;
second = final;
printf("%d\n", final);
}
}
Is there any way to increase the speed in which this program computes these calculations? Could you possible explain the solution to me (if it exists)?
Thanks.
Of course it's possible! :)
First of all, please note you're using signed int for your variables, and max int on a 32 bit machine is 2,147,483,647 which is approx. 10^8 times smaller than the max number you're using, which is 999,999,999,999,999,999.
I'll recommend to change the max num to INT_MAX (You'll need to include "limits.h")
Edit:
In addition, as said in the comments to my answer, consider changing to unsigned int. you need only positive values, and the max number will be twice higher.
2nd Edit:
That being said, when final will reach the closest it can to the limit in the condition, the next time its promoted, it'll exceed INT_MAX and will result in an overflow. That means here, that the condition will never be met.
Better to just change the condition to the times you want the loop to run. Please note though, that any fibonacci number larger than the max numder that can be stored in your variable type, will result in an overflow.
Secondly, final isn't initialized. Write final = 0 to avoid errors.
I recommend turning on all the warnings in your compiler. It could catch many errors before they compile :)
Also, I see no reason not to initialize the variables when you declare them. The value is already known.
Now for the speed of the program:
I'm not sure to which extent you're willing to change the code, but the simplest change without changing the original flow, is to make less calls to printf().
Since printf() is a function that will wait for a system resource to become available, this is probably the most time consuming part in your code.
Maybe consider storing the output in a string, and lets say every 100 numbers print the string to the screen.
Try maybe to create a string, with a size of
(10 (num of chars in an int) + 1 (new line char) )* 100 (arbitrary, based on when you'll want to flush the data to the screen)
Consider using sprintf() to write to a string in the inner loop, and strcat() to append a string to another string.
Then, every 100 times, use printf() to write to the screen.
As already stated in other answers, you have obvious two problems. 1) The missing initialization of final and 2) that your loop condition will result in an endless loop due to 999999999999999999 being larger than any integer value.
The real problem here is that you use a fixed number in the condition for the while.
How do you know which number to use so that you actually calculates all the Fibonacci numbers possible for the used integer type? Without knowing the numbers in advance you can't do that! So you need a better condition for stopping the loop.
One way of solving this to check for overflow instead - like:
while (second <= (INT_MAX - first)) { // Stop when next number will overflow
The above approach prevents signed overflow by checking whether the next first + second will overflow before actually doing the first+second. In this way signed overflow (and thereby UB) is prevented.
Another approach is to use unsigned integers and deliberately make an overflow (which is valid for unsigned int). Using unsigned long long that could look like:
unsigned long long first, second, next;
first = 1;
second = 1;
printf("1\n1\n");
next = first + second;
while (next > second) { // Stop when there was an overflow
printf("%llu\n", next);
first = second;
second = next;
next = first + second;
}
Speed isn't your problem. You have an infinite loop:
while (final <= 999999999999999999) {
final has type int. Most likely int is 32-bit on your system, which means the maximum value it can hold is 2147483647. This will always be smaller than 999999999999999999 (which is a constant of type long long), so the loop never ends.
Change the datatype of your variables to long long and the loop will terminate after about 87 iterations. Also, you'll need to change your printf format specifier from %d to %lld to match the datatype printed.
Why are you asking this question?
If it's the intention to increase the performance, you might go for the formula of the n-th Fibonacci number, which is something like:
((1+v5)/2)^n + ((1-v5)/2)^n, something like that (v5 being the square root of 5).
If it's about learning to increase performance, you might do a code review or use performance diagnostics tools.
Related
I am extremely new to programming in general, so please forgive my noobishness. I am trying to scale down the res[?] array in the function below. I know the problem that the res[?]*(multiplier/100) is creating a decimal answer instead of the required format which is an integer therefore I need to convert the result before it is plugged into the res[?].
I did think of turning res[] array into a double but I wasnt sure whether the initwindow(?,?) was compatible with integer values.
I am on mingw with code blocks. the linker and compiler has customized setting made by my professor. I am on Plain\basic C???
I tried to apply the techniques this website used about the truncating conversion. But doesn't seem to work. http://www.cs.tut.fi/~jkorpela/round.html
Debugger watcher shows that res[?] is equivalent to 0.
#include <stdio.h>
#include <graphics_lib.h>
#include <math.h>
int res[2]//contains a array which contains the vertical and horizontal detected resolution from main()
void function(res)
{
printf("Please note the lowest resolution available is 800x600\n");
printf("Please enter percentage ratio % below 100:");
scanf("%d",&multiplier);
res[1]=(int)(res[1]*(multiplier/100));
res[2]=(int)(res[2]*(multiplier/100));
blah blah blah blah.............
initwindow(res[1],res[2]) //from custom header that my professor made which initializes basic graphics window
return;
}
I'm assuming multiplier is an int to match the %d format.
multiplier/100 is guaranteed to be zero (if the user follows directions and provides a number less than 100). You can do (res[x]*multiplier)/100 to make sure the multiply happens first (you're probably okay without the parentheses, but rather than think about order of operations why not be explicit?)
The cast to int is unnecessary, because an int divided by another int is always an int.
Unless your professor has done some very interesting things, you should also note that a two-element array such as res would have elements res[0] and res[1], not res[1] and res[2].
you don't need to cast in this situation because the conversion is done implicitly . you only need a cast to assign the content of a variable to a variable of different type .
Looks like the multiplier variable is an int, so the result of multiplier/100 expression will be 0 in all cases.
Also, it's a good programming practice to check the validity of user's input.
You must declare multiplier before you use it:
int multiplier;
printf("Please note the lowest resolution available is 800x600\n");
printf("Please enter percentage ratio % below 100:");
scanf("%d",&multiplier);
And the index is staring from 0 not 1. And you should initialize res[0] and res[1] before using them. So:
res[0] = 800;
res[1] = 600;
And the division of multiplier by 100 will truncate to 0, try this without casting as it will be automaticly converted:
res[1]=(multiplier*res[0])/100;
res[1]=(multiplier*res[1])/100;
#include<stdio.h>
int fact(int k)
{
int j,f=1;
for(j=1;j<=k;j++)
f*=j;
return f;
}
int main()
{
int t,i,n[100],s[100],j;
scanf("%d",&t);
for(i=0;i<t;i++)
{
scanf("%d",&n[i]);
}
for(j=0;j<t;j++)
{
s[j]=fact(n[j]);
printf("%d \n",s[j]);
}
return 0;
}
You are asked to calculate factorials of some small positive integers.
Input
An integer t, 1<=t<=100, denoting the number of testcases, followed by t lines, each containing a single integer n, 1<=n<=100.
Output
For each integer n given at input, display a line with the value of n!
Example
Sample input:
4
1
2
5
3
Sample output:
1
2
120
6
Your code will give correct results for the given test cases but that doesn't prove that your code works. It is wrong is because of integer overflow. Try to calculate 100! by your program and you'll see what's the problem.
My answer lacked details. I'll update this to add details for an answer to the question as it stands now.
C has limitations over the the maximum and minimum size that can be stored in a variable. For doing arbitrary precision arithmetic it is usually advisable to use a bignum library as PHIFounder has suggested.
In the present case however, the use of external libraries is not possible. In this case arrays can be used to store integers exceeding the maximum value of the integers possible. OP has already found this possibility and used it. Her implementation, however, can use many optimizations.
Initially the use of large arrays like that can be reduced. Instead of using an array of 100 variables a single variable can be used to store the test cases. The use of large array and reading in test cases can give optimization only if you are using buffers to read in from stdin otherwise it won't be any better than calling scanf for reading the test cases by adding a scanf in the for loop for going over individual test cases.
It's your choice to either use buffering to get speed improvement or making a single integer instead of an array of 100 integers. In both the cases there will be improvements over the current solution linked to, on codechef, by the OP. For buffering you can refer to this question. If you see the timing results on codechef the result of buffering might not be visible because the number of operations in the rest of the logic is high.
Now second thing about the use of array[200]. The blog tutorial on codechef uses an array of 200 elements for demonstrating the logic. It is a naive approach as the tutorial itself points out. Storing a single digit at each array location is a huge waste of memory. That approach also leads to much more operations leading to a slower solution. An integer can at least store 5 digits (-32768 to 32767) and can generally store more. You can store the intermediate results in a long long int used as your temp and use all 5 digits. That simplification itself would lead to the use of only arr[40] instead of arr[200]. The code would need some additional changes to take care of forward carry and would become a little more complex but both speed and memory improvements would be visible.
You can refer to this for seeing my solutions or you can see this specific solution. I was able to take the use down to 26 elements only and it might be possible to take it further down.
I'll suggest you to put up your code on codereview for getting your code reviewed. There are many more issues that would be best reviewed there.
Here, your array index should start with 0 not 1 , I mean j and ishould be initialized to 0 in for loop.
Besides, try to use a debugger , that will assist you in finding bugs.
And if my guess is right you use turbo C, if yes then my recommendation is that you start using MinGW or Cygwin and try to compile on CLI, anyway just a recommendation.
There may be one more problem may be which is why codechef is not accepting your code you have defined function to accept the integer and then you are passing the array , may be this code will work for you:
#include<stdio.h>
int fact(int a[],int n)// here in function prototype I have defined it to take array as argument where n is array size.
{
int j=0,f=1,k;
for (k=a[j];k>0;k--)
f*=k;
return f;
}
int main()
{
int t,i,n[100],s[100],j;
setbuf(stdout,NULL);
printf("enter the test cases\n");
scanf("%d",&t); //given t test cases
for(i=0;i<t;i++)
{
scanf("%d",&n[i]); //value of the test cases whose factorial is to be calculated
}
for(j=0;j<t;j++)
{
s[j]=fact(&n[j],t);// and here I have passed it as required
printf("\n %d",s[j]); //output
}
return 0;
}
NOTE:- After the last edit by OP this implementation has some limitations , it can't calculate factorials for larger numbers say for 100 , again the edit has taken the question on a different track and this answer is fit only for small factorials
above program works only for small numbers that means upto 7!,after that that code not gives the correct results because 8! value is 40320
In c language SIGNED INTEGER range is -32768 to +32767
but the >8 factorial values is beyond that value so integer cant store those values
so above code can not give the right results
for getting correct values we declare the s[100] as LONG INT,But it is also work
only for some range
#include<stdio.h>
int main()
{
long ctr[100000],i;
float j;
long d[100000],T,h,o;
scanf("%ld",T);
printf("\n");
for(i=0; i<T; i++)
{
scanf("%ld",d[i]);
printf("\n");
for(h=d[i]; h<=0; h--)
{
j=h%10;
if(j==5)
{
ctr[i]++;
}
}
}
for(o=0; o<=i; o++)
{
printf("%ld\n",o);
}
return 0;
}
It's a program to find the number of trailing zeros of a factorial of a group of "T" numbers, input by the user.
What I have done is to divide each number by 10 and test if the number is 5. Then I decrement the number by 1 until it reaches 0. Each pair of 5*4 contributes one trailing 0.
Is the program consuming too much memory or is there another runtime error? This program is giving a runtime error, can anyone help?
A couple of observations:
long[100000] requires 400,000 bytes of stack assuming long is plain old 32 bits, 800,000 bytes if long is 64-bit. You have two such arrays, which might make your program hit operating system stack size limits.
The scanf() function requires pointers to where to store values it's reading in. You're not giving it pointers. As Paul R said in a comment.
Assuming positive numbers are input, the loop for(h=d[i]; h<=0; h--) will never run its body.
Hmm... This loop looks like the loop is going infinity.
for(h=d[i]; h<=0; h--)
Thus giving you a runtime error... Because the count is going negative, and your loops condition says continue until the value reaches less than or equal to zero, but it seems like, your value would never reach zero.
And for the scanf function, don't forget to use the & sign. Like this:
scanf("%ld", &sampleVariable);
That's the solution for your runtime error. :)
And also maybe you should use High Precision Variables, I think you need a higher precision more than the double datatype because you'll be dealing with huge amount of numbers.
See here for added info: http://www.nongnu.org/hpalib/
The loop is going infinity. Ensure that the count is not going negative or else you'll have a runtime error.
I want to write some code which takes any positive, even number (greater than 2) and gives me the smallest pair of primes that sum up to this number.
I need this program to handle any integer up to 9 digits long.
My aim is to make something that looks like this:
Please enter a positive even integer ( greater than 2 ) :
10
The first primes adding : 3+7=10.
Please enter a positive even integer ( greater than 2 ) :
160
The first primes adding : 3+157=160.
Please enter a positive even integer ( greater than 2 ) :
18456
The first primes adding : 5+18451=18456.
I don't want to use any library besides stdio.h. I don't want to use arrays, strings, or anything besides for the most basic toolbox: scanf, printf, for, while, do-while, if, else if, break, continue, and the basic operators (<,>, ==, =+, !=, %, *, /, etc...).
Please no other functions especially is_prime.
I know how to limit the input to my needs so that it loops until given a valid entry.
So now I'm trying to figure out the algorithm.
I thought of starting a while loop like something like this:
#include <stdio.h>
long first, second, sum, goldbach, min;
long a,b,i,k; //indices
int main (){
while (1){
printf("Please enter a positive integer :\n");
scanf("%ld",&goldbach);
if ((goldbach>2)&&((goldbach%2)==0)) break;
else printf("Wrong input, ");
}
while (sum!=goldbach){
for (a=3;a<goldbach;a=(a+2))
for (i=2;(goldbach-a)%i;i++)
first = a;
for (b=5;b<goldbach;b=(b+2))
for (k=2;(goldbach-b)%k;k++)
sum = first + second;
}
}
Have a function to test primality
int is_prime(unsigned long n)
And then you only need to test whether a and goldbach - a are both prime. You can of course assume a <= goldbach/2.
And be sure to handle goldbach = 4 correctly.
If the requirements don't allow defining and using your own functions, ignore them first. Solve the problem using any functions you deem helpful and convenient. When you have a working solution using disallowed functionality, then you start replacing that with allowed constructs. Self-defined functions can be inlined directly, replacing the return with an assignment, so instead of if (is_prime(a)), you have the code to determine whether a is prime and instead of returning the result you assign it is_prime = result; and test that variable if (is_prime). Where you have used library functions, reimplement them yourself - efficiency doesn't matter much - and then inline them too.
I am trying to modify a value in an array using the C programming language and I seem to be hitting a blank wall with this seemingly easy operation. Please see code snippet below:
while(1) {
printf("Current prime candidate is %i\n",nextPrimeCandidate);
int innerSieve;//=2;
int currentPrimeCandidate=0;
for (innerSieve=2;innerSieve<SIEVELIMIT;innerSieve++) {
currentPrimeCandidate = nextPrimeCandidate * innerSieve;
//printf("Inner Sieve is b4 funny place %i,%i\n",innerSieve,currentPrimeCandidate);
//initArray[currentPrimeCandidate]=5;
//VERY UNIQUE LINE
myArray[currentPrimeCandidate] = 0;
//printf("Inner Sieve after funny place is %i,%i \n",innerSieve,currentPrimeCandidate);
}
nextPrimeCandidate=getNextPrimeCandidate(myArray,++nextPrimeCandidate);
if ((nextPrimeCandidate^2) > SIEVELIMIT ) break;
}
The problem is with the line highlighted with the VERY UNIQUE LINE comment. For some reason, when the innerSieve variable reaches 33 and gets to that line, it sets the contents of the innerSieve variable to the value of that line ( which currently is 0) and basically forces the loop into an infinite loop( the SIEVELIMIT variable is set at 50). It seems that there is some funny stuff going on in the registers when I checked using the Eclipse Debug facility but I am not too sure what I should be looking for.
If you need the whole code listing, this can be provided.( with a particular variable which is not yet initialised in the code being initialised at the precise point that the innerSieve variable hits 32)
Any help will be greatly appreciated.
Guessing that currentPrimeCandidate is greater than the maximum index of myArray, and you're overwriting innerSieve (which likely follows myArray on the stack).
#ruslik hit on it in the comment. The problem is this line:
if ((nextPrimeCandidate^2) > SIEVELIMIT ) break;
In C, the ^ operator is not the power operator, it is the bitwise xor operator. You're iterating far too many times than you intend, which is resulting in an array-index-out-of-bounds error, so you're overwriting random memory and getting strange results.
There is no power operator in C (though there is the pow function). Since you're just squaring a number, the simplest fix is to multiply the number by itself:
if ((nextPrimeCandidate * nextPrimeCandidate) > SIEVELIMIT ) break;