Convert inefficient recursive coin change function to iteration - c++

I have an inefficient recursive coin change function that works out the number of coin combinations for a given amount. I would like to convert it to a more efficient iterative function if possible.
One problem is that I am using backtracking to try out different coins in an array called denominations. I am also using memoization but it doesn't speed things up when the amount is large.
Here's my code:
unsigned long long CalculateCombinations(std::vector<double> &denominations, std::vector<double> change,
double amount, unsigned int index)
{
double current = 0.0;
unsigned long long combinations = 0;
if (amount == 0.0)
{
if (change.size() % 2 == 0)
{
combinations = Calculate(change);
}
return combinations;
}
// If amount is less than 0 then no solution exists
if (amount < 0.0)
return 0;
// If there are no coins and index is greater than 0, then no solution exist
if (index >= denominations.size())
return 0;
std::string str = std::to_string(amount) + "-" + std::to_string(index) + "-" + std::to_string(change.size());
auto it = Memo.find(str);
if (it != Memo.end())
{
return it->second;
}
while (current <= amount)
{
double remainder = amount - current;
combinations += CalculateCombinations(denominations, change, remainder, index + 1);
current += denominations[index];
change.push_back(denominations[index]);
}
Memo[str] = combinations;
return combinations;
}
Any ideas how this can be done? I know there are DP solutions for coin change problems, but mine doesn't lend itself easily to one. I can have half pennies.
*Update: I changed the function to be iterative and I scaled up by a factor of 2 to use integers, bit it made no considerable difference.
Here's my new code:
unsigned long long CalculateCombinations(std::vector<int> &denominations, std::vector<int> change, int amount, unsigned int index)
{
unsigned long long combinations = 0;
if (amount <= 0)
return combinations;
std::stack<Param> mystack;
mystack.push({ change, amount, index });
while (!mystack.empty())
{
int current = 0;
std::vector<int> current_coins = mystack.top().Coins;
int current_amount = mystack.top().Amount;
unsigned int current_index = mystack.top().Index;
mystack.pop();
if (current_amount == 0)
{
if (current_coins.size() % 2 == 0)
{
combinations += Calculate(std::move(current_coins));
}
}
else
{
std::string str = std::to_string(current_amount) + "-" + std::to_string(current_index);
if (Memo.find(str) == Memo.end())
{
// If amount is less than 0 then no solution exists
if (current_amount >= 0 && current_index < denominations.size())
{
while (current <= current_amount)
{
int remainder = current_amount - current;
mystack.push({ current_coins, remainder, current_index + 1 });
current += denominations[current_index];
current_coins.push_back(denominations[current_index]);
}
}
else
{
Memo.insert(str);
}
}
}
}
return combinations;
}
Memo is defined as std::unordered_set.
Can this be solved by DP? The problem is that I'm not interested in all combinations - only combinations that are even in size.

I don't see any strategy in your code that discards denominations.
My recursive answer would be at each recursive stage to create 2 children:
1 child uses the full list of denominations, and spends 1 of the end denomination
The second child discards the same end denomination
They each recurse, but in the second case the children have one less denomination to work with.
I believe the results returned are all distinct, but of course you do get the pain case where you recurse 10000 levels deep to get $100 in pennies. This can easily be optimised for when you get down to 1 denomination, and probably indicates that it is best to process and discard higher denominations in each round rather than lower denominations.
You can also detect the case where all remaining denominations are simple multiples of each other and quickly generate the permutations without doing the full recursion: Generate the minimum coin set (max of each high denomination) then work backward replacing each coin with numbers of smaller coins.

Related

Need optimization tips for a subset sum like problem with a big constraint

Given a number 1 <= N <= 3*10^5, count all subsets in the set {1, 2, ..., N-1} that sum up to N. This is essentially a modified version of the subset sum problem, but with a modification that the sum and number of elements are the same, and that the set/array increases linearly by 1 to N-1.
I think i have solved this using dp ordered map and inclusion/exclusion recursive algorithm, but due to the time and space complexity i can't compute more than 10000 elements.
#include <iostream>
#include <chrono>
#include <map>
#include "bigint.h"
using namespace std;
//2d hashmap to store values from recursion; keys- i & sum; value- count
map<pair<int, int>, bigint> hmap;
bigint counter(int n, int i, int sum){
//end case
if(i == 0){
if(sum == 0){
return 1;
}
return 0;
}
//alternative end case if its sum is zero before it has finished iterating through all of the possible combinations
if(sum == 0){
return 1;
}
//case if the result of the recursion is already in the hashmap
if(hmap.find(make_pair(i, sum)) != hmap.end()){
return hmap[make_pair(i, sum)];
}
//only proceed further recursion if resulting sum wouldnt be negative
if(sum - i < 0){
//optimization that skips unecessary recursive branches
return hmap[make_pair(i, sum)] = counter(n, sum, sum);
}
else{
//include the number dont include the number
return hmap[make_pair(i, sum)] = counter(n, i - 1, sum - i) + counter(n, i - 1, sum);
}
}
The function has starting values of N, N-1, and N, indicating number of elements, iterator(which decrements) and the sum of the recursive branch(which decreases with every included value).
This is the code that calculates the number of the subsets. for input of 3000 it takes around ~22 seconds to output the result which is 40 digits long. Because of the long digits i had to use an arbitrary precision library bigint from rgroshanrg, which works fine for values less than ~10000. Testing beyond that gives me a segfault on line 28-29, maybe due to the stored arbitrary precision values becoming too big and conflicting in the map. I need to somehow up this code so it can work with values beyond 10000 but i am stumped with it. Any ideas or should i switch towards another algorithm and data storage?
Here is a different algorithm, described in a paper by Evangelos Georgiadis, "Computing Partition Numbers q(n)":
std::vector<BigInt> RestrictedPartitionNumbers(int n)
{
std::vector<BigInt> q(n, 0);
// initialize q with A010815
for (int i = 0; ; i++)
{
int n0 = i * (3 * i - 1) >> 1;
if (n0 >= q.size())
break;
q[n0] = 1 - 2 * (i & 1);
int n1 = i * (3 * i + 1) >> 1;
if (n1 < q.size())
q[n1] = 1 - 2 * (i & 1);
}
// construct A000009 as per "Evangelos Georgiadis, Computing Partition Numbers q(n)"
for (size_t k = 0; k < q.size(); k++)
{
size_t j = 1;
size_t m = k + 1;
while (m < q.size())
{
if ((j & 1) != 0)
q[m] += q[k] << 1;
else
q[m] -= q[k] << 1;
j++;
m = k + j * j;
}
}
return q;
}
It's not the fastest algorithm out there, and this took about half a minute for on my computer for n = 300000. But you only need to do it once (since it computes all partition numbers up to some bound) and it doesn't take a lot of memory (a bit over 150MB).
The results go up to but excluding n, and they assume that for each number, that number itself is allowed to be a partition of itself eg the set {4} is a partition of the number 4, in your definition of the problem you excluded that case so you need to subtract 1 from the result.
Maybe there's a nicer way to express A010815, that part of the code isn't slow though, I just think it looks bad.

Addition of Even Fibonacci Numbers

I'm trying to solve the 2nd problem on Project Euler where I have to print the sum of all even Fibonacci numbers under 4 million. I'm using the following code but the program is not returning any value. When I replace 4000000 by something small like 10, I get the sum. Does that mean my program is taking too long? What am I doing wrong?
#include <iostream>
using namespace std;
int fibonacci(int i) {
if (i == 2)
return 2;
else if (i == 1)
return 1;
else return fibonacci(i - 1) + fibonacci(i - 2);
}
int main() {
int currentTerm, sum = 0;
for (int i = 1; i <= 10; i++) {
currentTerm = fibonacci(i);
if (currentTerm % 2 == 0)
sum += currentTerm;
}
cout << sum;
return 0;
}
Problem 2 of project Euler asks (emphasis mine)
By considering the terms in the Fibonacci sequence whose values do not exceed four million, find the sum of the even-valued terms.
Doing
for (int i = 1; i <= 4000000; i++)
{
currentTerm = fibonacci(i);
// ...
}
You are trying to calculate up to the 4,000,000th Fibonacci number, which is a very big beast, while you should stop around the 33th instead.
The other answers already pointed out the inefficiency of the recursive approach, but let me add some numbers to the discussion, using this slightly modified version of your program
#include <iostream>
#include <iomanip>
int k = 0;
// From https://oeis.org/A000045 The fibonacci numbers are defined by the
// recurrence relation F(n) = F(n-1) + F(n-2) with F(0) = 0 and F(1) = 1.
// In the project Euler question the sequence starts with 1, 2, 3, 5, ...
// So in the following I'll consider F(1) = 1 and F(2) = 2 as The OP does.
long long fibonacci(long long i)
{
++k;
if (i > 2)
return fibonacci(i - 1) + fibonacci(i - 2);
else
return i;
}
int main()
{
using std::cout;
using std::setw;
const long limit = 4'000'000;
long sum = 0;
cout << " i F(i) sum calls\n"
"-----------------------------------\n";
for (int i = 1; ; ++i)
{
long long F_i = fibonacci(i);
if ( F_i > limit ) // <-- corrected end condition
break;
if (F_i % 2 == 0)
{
sum += F_i;
cout << setw(3) << i << setw(10) << F_i
<< setw(10) << sum << setw(11) << k << '\n';
}
}
cout << "\nThe sum of all even Fibonacci numbers less then "
<< limit << " is " << sum << '\n';
return 0;
}
Once executed (live here), you can notice that the recursive function has been called more than 10,000,000 times, to calculate up to the 33th Fibonacci number.
That's simply not the right way. Memoization could help, here there's a quick benchmark comparing the recursive functions with a toy implementation of the memoization technique, which is represented by the histogram that you can't see. Because it's 300,000 times shorter than the others.
Still, that's not the "correct" or "natural" way to deal with this problem. As noted in the other answers you could simply calculate each number in sequence, given the previous ones. Enthus3d also noted the pattern in the sequence: odd, odd, even, odd, odd, even, ...
We can go even further and directly calculate only the even terms:
#include <iostream>
int main()
{
const long limit = 4'000'000;
// In the linked question the sequence starts as 1, 2, 3, 5, 8, ...
long long F_0 = 2, F_3 = 8, sum = F_0 + F_3;
for (;;)
{
// F(n+2) = F(n+1) + F(n)
// F(n+3) = F(n+2) + F(n+1) = F(n+1) + F(n) + F(n+1) = 2F(n+1) + F(n)
// F(n+6) = F(n+5) + F(n+4) = F(n+4) + F(n+3) + F(n+3) + F(n+2)
// = 2F(n+3) + F(n+4) + F(n+2) = 3F(n+3) + 2F(n+2)
// = 3F(n+3) + 2F(n+1) + 2F(n) = 3F(n+3) + F(n+3) - F(n) + 2F(n)
long long F_6 = 4 * F_3 + F_0;
if ( F_6 > limit )
break;
sum += F_6;
F_0 = F_3;
F_3 = F_6;
}
std::cout << sum << '\n'; // --> 4613732
return 0;
}
Live here.
If you need multiple Fibonacci numbers, and especially if you need all of them, do not use the recursive approach, use iteration instead:
var prev=0;
var curr=1;
var sum=0;
while(curr<4000000){
if(curr%2==0)
sum+=curr;
var temp=prev;
prev=curr;
curr+=temp;
}
console.log(sum);
The snippet is JavaScript (so it can run here), but if you make var-s to int-s, it will be C-ish enough.
But the actual problem was the loop: you do not need to calculate the first
n (4000000) Fibonacci numbers (which would lead to various overflows), but the Fibonacci numbers which are smaller than 4000000.
If you want a bit of magic, you can also build on the fact that every 3rd Fibonacci number is even, on the basis of "even+odd=>odd", "odd+even=>odd", and only "odd+odd=>even":
0 1 1 2 3 5 8...
E O O E O O E
^ O+O
^ E+O
^ O+E
^ O+O
var prev=1;
var curr=2;
var sum=0;
while(curr<4000000){
sum+=curr;
console.log("elem: "+curr,"sum: "+sum);
for(var i=0;i<3;i++){
var temp=prev;
prev=curr;
curr+=temp;
}
}
And if the question would be only the title, Addition of even fibonacci numbers (let's say, n of them), pure mathematics could do the job, using Binet's formula (described in #Silerus' answer) and the fact that it is an (a^n-b^n)/c thing, where a^n and b^n are geometric sequences, every 3rd of them also being a geometric sequence, (a^3)^n, and the sum of geometric sequences has a simple, closed form (if the series is a*r^n, the sum is a*(1-r^n)/(1-r)).
Putting everything together:
// convenience for JS->C
var pow=Math.pow;
var sqrt=Math.sqrt;
var round=Math.round;
var s5=sqrt(5);
var a=(1+s5)/2;
var a3=pow(a,3);
var b=(1-s5)/2;
var b3=pow(b,3);
for(var i=0;i<12;i++){
var nthEvenFib=round((pow(a3,i)-pow(b3,i))/s5);
var sumEvenFibs=round(((1-pow(a3,i+1))/(1-a3)-(1-pow(b3,i+1))/(1-b3))/s5);
console.log("elem: "+nthEvenFib,"sum: "+sumEvenFibs);
}
Again, both snippets become rather C-ish if var-s are replaced with some C-type, int-s in the first snippet, and mostly double-s in this latter one (the loop variable i can be a simple int of course).
You can use the Binet formula in your calculations - this will allow you to abandon the slow recursive algorithm, another option may be a non-recursive algorithm for calculating fibonacci numbers. https://en.wikipedia.org/wiki/Jacques_Philippe_Marie_Binet. Here is an example of using the Binet formula, it will be much faster than the recursive algorithm, since it does not recalculate all previous numbers.
#include <iostream>
#include <math.h>
using namespace std;
int main(){
double num{},a{(1+sqrt(5))/2},b{(1-sqrt(5))/2},c{sqrt(5)};
int sum{};
for (auto i=1;i<30;++i){
num=(pow(a,i)-pow(b,i))/c;
if (static_cast<int>(num)%2==0)
sum+=static_cast<int>(num);
}
cout<<sum;
return 0;
}
variant 2
int fib_sum(int n)
{
int sum{};
if (n <= 2) return 0;
std::vector<int> dp(n + 1);
dp[1] = 1; dp[2] = 1;
for (int i = 3; i <= n; i++)
{
dp[i] = dp[i - 1] + dp[i - 2];
if(dp[i]%2==0)
sum+=dp[i];
}
return sum;
}
You can speed up brutally by using compile time precalculations for all even Fibonacci numbers and sums using constexpre functions.
A short check with Binets formula shows, that roundabout 30 even Fibonacci numbers will fit into a 64bit unsigned value.
30 numbers can really easily been procealculated without any effort for the compiler. So, we can create a compile time constexpr std::array with all needed values.
So, you will have zero runtime overhead, making you program extremely fast. I am not sure, if there can be a faster solution. Please see:
#include <iostream>
#include <array>
#include <algorithm>
#include <iterator>
// ----------------------------------------------------------------------
// All the following wioll be done during compile time
// Constexpr function to calculate the nth even Fibonacci number
constexpr unsigned long long getEvenFibonacciNumber(size_t index) {
// Initialize first two even numbers
unsigned long long f1{ 0 }, f2{ 2 };
// calculating Fibonacci value
while (--index) {
// get next even value of Fibonacci sequence
unsigned long long f3 = 4 * f2 + f1;
// Move to next even number
f1 = f2;
f2 = f3;
}
return f2;
}
// Get nth even sum of Fibonacci numbers
constexpr unsigned long long getSumForEvenFibonacci(size_t index) {
// Initialize first two even prime numbers
// and their sum
unsigned long long f1{ 0 }, f2{ 2 }, sum{ 2 };
// calculating sum of even Fibonacci value
while (--index) {
// get next even value of Fibonacci sequence
unsigned long long f3 = 4 * f2 + f1;
// Move to next even number and update sum
f1 = f2;
f2 = f3;
sum += f2;
}
return sum;
}
// Here we will store ven Fibonacci numbers and their respective sums
struct SumOfEvenFib {
unsigned long long fibNum;
unsigned long long sum;
friend bool operator < (const unsigned long long& v, const SumOfEvenFib& f) { return v < f.fibNum; }
};
// We will automatically build an array of even numbers and sums during compile time
// Generate a std::array with n elements taht consist of const char *, pointing to Textx...Texty
template <size_t... ManyIndices>
constexpr auto generateArrayHelper(std::integer_sequence<size_t, ManyIndices...>) noexcept {
return std::array<SumOfEvenFib, sizeof...(ManyIndices)>{ { {getEvenFibonacciNumber(ManyIndices + 1), getSumForEvenFibonacci(ManyIndices + 1)}...}};
};
// You may check with Ninets formula
constexpr size_t MaxIndexFor64BitValue = 30;
// Generate the reuired number of texts
constexpr auto generateArray()noexcept {
return generateArrayHelper(std::make_integer_sequence<size_t, MaxIndexFor64BitValue>());
}
// This is an constexpr array of even Fibonacci numbers and its sums
constexpr auto SOEF = generateArray();
// ----------------------------------------------------------------------
int main() {
// Show sum for 4000000
std::cout << std::prev(std::upper_bound(SOEF.begin(), SOEF.end(), 4000000))->sum << '\n';
// Show all even numbers and their corresponding sums
for (const auto& [even, sum] : SOEF) std::cout << even << " --> " << sum << '\n';
return 0;
}
Tested with MSVC 19, clang 11 and gcc10
Compiled with C++17
Welcome to Stack Overflow :)
I have only modified your code on the loop, and kept your Fibonacci implementation the same. I've verified the code's answer on Project Euler. The code can be found below, and I hope my comments help you understand it better.
The three things I've changed are:
1) You tried to look for a number all the way until the 4,000,000 iteration rather than for the number that is less than 4,000,000. That means your program probably went crazy trying to add a number that's insanely large (which we don't need) <- this is probably why your program threw in the towel
2) I improved the check for even numbers; we know that fibonacci sequences go odd odd even, odd odd even, so we only really need to add every third number to our sum instead of checking if the number itself is even <- modulus operations are very expensive on large numbers
3) I added two lines that are commented out with couts, they can help you debug and troubleshoot your output
There's also a link here about using Dynamic Programming to solve the question more efficiently, should anyone need it.
Good luck!
#include <iostream>
using namespace std;
int fibonacci(int i) {
if (i == 2)
return 2;
else if (i == 1)
return 1;
else return fibonacci(i - 1) + fibonacci(i - 2);
}
int main() {
// need to add the sum of all even fib numbers under a particular sum
int max_fib_number = 4000000;
int currentTerm, sum = 0;
currentTerm = 1;
int i = 1;
// we do not need a for loop, we need a while loop
// this is so we can detect when our current number exceeds fib
while(currentTerm < max_fib_number) {
currentTerm = fibonacci(i);
//cout << currentTerm <<"\n";
// notice we check here if currentTerm is a valid number to add
if (currentTerm < max_fib_number) {
//cout << "i:" << i<< "\n";
// we only want every third term
// this is because 1 1 2, 3 5 8, 13 21 34,
// pattern caused by (odd+odd=even, odd+even=odd)
// we also add 1 because we start with the 0th term
if ((i+1) % 3 == 0)
sum += currentTerm;
}
i++;
}
cout << sum;
return 0;
}
Here's Your modified code which produce correct output to the project euler's problem.
#include <iostream>
using namespace std;
int fibonacci(int i) {
if (i == 2)
return 2;
else if (i == 1)
return 1;
else return fibonacci(i - 1) + fibonacci(i - 2);
}
int main() {
int currentsum, sum = 0;
for (int i = 1; i <= 100; i++) {
currentsum = fibonacci(i);
//here's where you doing wrong
if(sum >= 4000000) break; //break when sum reaches 4mil
if(currentsum %2 == 0) sum+=currentsum; // add when even-valued occurs in the currentsum
}
cout << sum;
return 0;
}
Output 4613732
Here's my Code which consists of while loop until 4million occurs in the sum with some explanation.
#include <iostream>
using namespace std;
int main()
{
unsigned long long int a,b,c , totalsum;
totalsum = 0;
a = 1; // 1st index digit in fib series(according to question)
b = 2; // 2nd index digit in fib series(according to question)
totalsum+=2; // because 2 is an even-valued term in the series
while(totalsum < 4000000){ //loop until 4million
c = a+b; // add previous two nums
a = b;
b = c;
if(c&1) continue; // if its odd ignore and if its an even-valued term add to totalsum
else totalsum+=c;
}
cout << totalsum;
return 0;
}
for people who downvoted, you can actually say what is wrong in the code instead downvoting the actual answer to the https://projecteuler.net/problem=2 is the output of the above code 4613732 , competitive programming itself is about how fast can you solve problems instead of clean code.

How can I improve my prime number program with Sieve of Eratosthenes algorithm?

My program prints all prime numbers from this expression:
((1 + sin(0.1*i))*k) + 1, i = 1, 2, ..., N.
Input Format:
No more than 100 examples. Every example has 2 positive integers on the same line.
Output Format:
Print each number on a separate line.
Sample Input:
4 10
500 100
Sample Output:
5
17
But my algorithm is not efficient enough. How can I add Sieve of Eratosthenes so it can be efficient enough to not print "Terminated due to timeout".
#include <iostream>
#include <cmath>
using namespace std;
int main() {
long long k, n;
int j;
while (cin >> k >> n) {
if (n>1000 && k>1000000000000000000) continue;
int count = 0;
for (int i = 1; i <= n; i++) {
int res = ((1 + sin(0.1*i)) * k) + 1;
for (j = 2; j < res; j++) {
if (res % j == 0) break;
}
if (j == res) count++;
}
cout << count << endl;
}
system("pause");
You can improve your speed by 10x simply by doing a better job with your trial division. You're testing all integers from 2 to res instead of treating 2 as a special case and testing just odd numbers from 3 to the square root of res:
// k <= 10^3, n <= 10^9
int main() {
unsigned k;
unsigned long long n;
while (cin >> k >> n) {
unsigned count = 0;
for (unsigned long long i = 1; i <= n; i++) {
unsigned long long j, res = (1 + sin(0.1 * i)) * k + 1;
bool is_prime = true;
if (res <= 2 || res % 2 == 0) {
is_prime = (res == 2);
} else {
for (j = 3; j * j <= res; j += 2) {
if (res % j == 0) {
is_prime = false;
break;
}
}
}
if (is_prime) {
count++;
}
}
cout << count << endl;
}
}
Though k = 500 and n = 500000000 is still going to take forty seconds or so.
EDIT: I added a 3rd mean to improve efficiency
EDIT2: Added an explanation why Sieve should not be the solution and some trigonometry relations. Moreover, I added a note on the history of the question
Your problem is not to count all the prime numbers in a given range, but only those which are generated by your function.
Therefore, I don't think that the Sieve of Eratosthenes is the solution for this particular exercise, for the following reason: n is always rather small while k can be very large. If kis very large, then the Sieve algorithm would have to generate a huge number of prime numbers, for finally use it for a small number of candidates.
You can improve the efficiency of you program by three means:
Avoid calculating sin(.) every time. You can use trigonometric relations for example. Moreover, first time you calculate these values, store them in an array and reuse these values. Calculation of sin()is very time consuming
In your test to check if a number is prime, limit the search to sqrt(res). Moreover, consider make the test with odd j only, plus 2
If a candidate res is equal to the previous one, avoid redoing the test
A few trigonometry
If c = cos(0.1) and s = sin(0.1), you can use the relations :
sin (0.1(i+1)) = s*cos (0.1*i) + c*sin(0.1*i))
cos (0.1(i+1)) = c*cos (0.1*i) - s*sin(0.1*i))
If n were large, it should be necessary to recalculate the sin() by the function regularly to avoid too much rounding error calculation. But it should not be the case here as n is always rather small.
However, as I mentioned, it is better to use only the "memorization" trick in a first step and check if it is enough.
A note on the history of this question and why this answer:
Recently, this site received several questions " how to improve my program, to count number of prime numbers generated by this k*sin() function ..." To my knowledge, these questions were all closed as duplicate, under the reason that the Sieve is the solution and was explained in a previous similar (but slightly different) question. Now, the same question reappeared under a slightly different form "How can I insert the Sieve algorithm in this program ... (with k*sin() again)". And then I realised that the Sieve is not the solution. It is not a criticism to previous closes as I made the same mistake in the understanding on the question. However, I think it is time to propose a new solution, even it is does not match the new question perfectly
When you make use of a simple Wheel factorization, you can obtain a very nice speedup of your code. Wheel factorization of order 2 makes use of the fact that all primes bigger than 3 can be written as 6n+1 or 6n+5 for natural n. This means that you only have to do 2 divisions per 6 numbers. Or even further, all primes bigger than 5 can be written as 30n+m, with m in {1,7,11,13,17,19,23,29}. ( 8 divisions per 30 numbers).
Using this simple principle, you can write the following function to test your primes (wheel {2,3}):
bool isPrime(long long num) {
if (num == 1) return false; // 1 is not prime
if (num < 4) return true; // 2 and 3 are prime
if (num % 2 == 0) return false; // divisible by 2
if (num % 3 == 0) return false; // divisible by 3
int w = 5;
while (w*w <= num) {
if(num % w == 0) return false; // not prime
if(num % (w+2) == 0) return false; // not prime
w += 6;
}
return true; // must be prime
}
You can adapt the above for the wheel {2,3,5}. This function can be used in the main program as:
int main() {
long long k, n;
while (cin >> k >> n) {
if (n>1000 && k>1000000000000000000) continue;
int count = 0;
for (int i = 1; i <= n; i++) {
long long res = ((1 + sin(0.1*i)) * k) + 1;
if (isPrime(res)) { count++; }
}
cout << count << endl;
}
return 0;
}
A simple timing gives me for the original code (g++ prime.cpp)
% time echo "6000 100000000" | ./a.out
12999811
echo "6000 100000000" 0.00s user 0.00s system 48% cpu 0.002 total
./a.out 209.66s user 0.00s system 99% cpu 3:29.70 total
while the optimized version gives me
% time echo "6000 100000000" | ./a.out
12999811
echo "6000 100000000" 0.00s user 0.00s system 51% cpu 0.002 total
./a.out 10.12s user 0.00s system 99% cpu 10.124 total
Other improvements can be made but might have minor effects:
precompute your sine-table sin(0.1*i) for i from 0 to 1000. This will avoid recomputing those sines over and over. This however, has a minor impact as most time is wasted on the primetest.
Checking if res(i) == res(i+1): this has barely any impact as, depending on n and k most consecutive res are not equal.
Use a lookup table, might be handier, this does have an impact.
original answer:
My suggestion is the following:
Precompute your sinetable sin(0.1*i) for i from 0 to 1000. This will avoid recomputing those sines over and over. Also, do it smart (see point 3)
Find the largest possible value of res which is res_max=(2*k)+1
Find all primes for res_max using the Sieve of Eratosthenes. Also, realize that all primes bigger than 3 can be written as 6n+1 or 6n+5 for natural n. Or even further, all primes bigger than 5 can be written as 30n+m, with m in {1,7,11,13,17,19,23,29}. This is what is called Wheel factorization. So do not bother checking any other number. (a tiny bit more info here)
Have a lookup table that states if a number is a prime.
Do all your looping over the lookup table.

How to speed up this primality test

I would like to find the largest prime factor of a given number. After several attempts, I've enhanced the test to cope with rather big numbers (i.e. up to one billion in milliseconds). The problem is now if go beyond one billion, the execution time goes forever, so to speak. I wonder if I can do more improvements and reduce the execution time. I'm hoping for better execution time because in this link Prime Factors Calculator, the execution time is incredibly fast. My target number at this moment is 600851475143. The code is rather self-explanatory. Note: I've considered Sieve of Eratosthenes algorithm with no luck regarding the execution time.
#include <iostream>
#include <cmath>
bool isPrime(int n)
{
if (n==2)
return true;
if (n%2==0)
return false;
for (int i(3);i<=sqrt(n);i+=2) // ignore even numbers and go up to sqrt(n)
if (n%i==0)
return false;
return true;
}
int main()
{
int max(0);
long long target(600851475143);
if( target%2 == 0 )
max = 2;
for ( int i(3); i<target; i+=2 ){ // loop through odd numbers.
if( target%i == 0 ) // check for common factor
if( isPrime(i) ) // check for prime common factor
max = i;
}
std::cout << "The greatest prime common factor is " << max << "\n";
return 0;
}
One obvious optimization that I can see is:
for (int i(3);i<=sqrt(n);i+=2) // ignore even numbers and go up to sqrt(n)
instead of calculating sqrt everytime you can cache the result in a variable.
auto maxFactor = static_cast<int>sqrt(n);
for (int i(3); i <= maxFactor; i+=2);
The reason I believe this could lead to speed up is sqrt deals with floating point arithematic and compilers usually aren't generous in optimizing floating point arithematic. gcc has a special flag ffast-math to enable floating point optimizations explicitely.
For numbers upto the target range that you mentioned, you will need better algorithms. repeated divisioning should suffice.
Here is the code (http://ideone.com/RoAmHd) which hardly takes any time to finish:
int main() {
long long input = 600851475143;
long long mx = 0;
for (int x = 2; x <= input/x; ++x){
while(input%x==0) {input/=x; mx = x; }
}
if (input > 1){
mx = input;
}
cout << mx << endl;
return 0;
}
The idea behind repeated division is if a number is already a factor of p, it is also a factor of p^2, p^3, p^4..... So we keep eliminating factors so only prime factors remain that eventually get to divide the number.
You don't need a primality test. Try this algorithm:
function factors(n)
f := 2
while f * f <= n
if n % f == 0
output f
n := n / f
else
f := f + 1
output n
You don't need a primality test because the trial factors increase by 1 at each step, so any composite trial factors will have already been handled by their smaller constituent primes.
I'll leave it to you to implement in C++ with appropriate data types. This isn't the fastest way to factor integers, but it is sufficient for Project Euler 3.
for ( int i(3); i<target; i+=2 ){ // loop through odd numbers.
if( target%i == 0 ) // check for common factor
if( isPrime(i) ) // check for prime common factor
max = i;
It is the first two lines of this code, not primality checks, which take almost all time. You divide target to all numbers from 3 to target-1. This takes about target/2 divisions.
Besides, target is long long, while i is only int. It is possible that the size is too small, and you get an infinite loop.
Finally, this code does not calculate the greatest prime common factor. It calculate the greatest prime divisor of target, and does it very inefficiently. So what do you really need?
And it is a bad idea to call anything "max" in c++, because max is a standard function.
Here is my basic version:
int main() {
long long input = 600851475143L;
long long pMax = 0;
// Deal with prime 2.
while (input % 2 == 0) {
input /= 2;
pMax = 2;
}
// Deal with odd primes.
for (long long x = 3; x * x <= input; x += 2) {
while (input % x == 0) {
input /= x;
pMax = x;
}
}
// Check for unfactorised input - must be prime.
if (input > 1) {
pMax = input;
}
std::cout << "The greatest prime common factor is " << pMax << "\n";
return 0;
}
It might be possible to speed things up further by using a Newton-Raphson integer square root method to set up a (mostly) fixed limit for the loop. If available that would need a rewrite of the main loop.
long long limit = iSqrt(input)
for (long long x = 3; x <= limit; x += 2) {
if (input % x == 0) {
pMax = x;
do {
input /= x;
} while (input % x == 0);
limit = iSqrt(input); // Value of input changed so reset limit.
}
}
The square root is only calculated when a new factor is found and the value of input has changed.
Note that except for 2 and 3, all prime numbers are adjacent to multiples of 6.
The following code reduces the total number of iterations by:
Leveraging the fact mentioned above
Decreasing target every time a new prime factor is found
#include <iostream>
bool CheckFactor(long long& target,long long factor)
{
if (target%factor == 0)
{
do target /= factor;
while (target%factor == 0);
return true;
}
return false;
}
long long GetMaxFactor(long long target)
{
long long maxFactor = 1;
if (CheckFactor(target,2))
maxFactor = 2;
if (CheckFactor(target,3))
maxFactor = 3;
// Check only factors that are adjacent to multiples of 6
for (long long factor = 5, add = 2; factor*factor <= target; factor += add, add = 6-add)
{
if (CheckFactor(target,factor))
maxFactor = factor;
}
if (target > 1)
return target;
return maxFactor;
}
int main()
{
long long target = 600851475143;
std::cout << "The greatest prime factor of " << target << " is " << GetMaxFactor(target) << std::endl;
return 0;
}

Find Largest Prime Factor - Complexity of Code

I tried a code on a coding website to find the largest prime factor of a number and it's exceeding the time limit for the last test case where probably they are using a large prime number. Can you please help me to reduce the complexity of the following code?
int main()
{
long n;
long int lar, fact;
long int sqroot;
int flag;
cin >> n;
lar=2, fact=2;
sqroot = sqrt(n);
flag = 0;
while(n>1)
{
if((fact > sqroot) && (flag == 0)) //Checking only upto Square Root
{
cout << n << endl;
break;
}
if(n%fact == 0)
{
flag = 1;
lar = fact;
while(n%fact == 0)
n = n/fact;
}
fact++;
}
if(flag == 1) //Don't display if loop fact reached squareroot value
cout << lar << endl;
}
Here I've also taken care of the loop checking till Square Root value. Still, how can I reduce its complexity further?
You can speed things up (if not reduce the complexity) by supplying a hard-coded list of the first N primes to use for the initial values of fact, since using composite values of fact are a waste of time. After that, avoid the obviously composite values of fact (like even numbers).
You can reduce the number of tests by skipping even numbers larger than 2, and stopping sooner if you have found smaller factors. Here is a simpler and faster version:
int main() {
unsigned long long n, lar, fact, sqroot;
cin >> n;
lar = 0;
while (n && n % 2 == 0) {
lar = 2;
n /= 2;
}
fact = 3;
sqroot = sqrt(n);
while (fact <= sqroot) {
if (n % fact == 0) {
lar = fact;
do { n /= fact; } while (n % fact == 0);
sqroot = sqrt(n);
}
fact += 2;
}
if (lar < n)
lar = n;
cout << lar << endl;
return 0;
}
I am not sure how large the input numbers may become, using the larger type unsigned long long for these computations will get you farther than long. Using a precomputed array of primes would help further, but not by a large factor.
The better result I've obtained is using the function below (lpf5()). It's based on the primality() function (below) that uses the formulas 6k+1, 6k-1 to individuate prime numbers. All prime numbers >= 5 may be expressed in one of the forms p=k*6+1 or p=k*6-1 with k>0 (but not all the numbers having such a forms are primes). Developing these formulas we can see a sequence like the following:
k=1 5,7
k=2 11,13
k=3 17,19
k=4 23,25*
k=5 29,31
.
.
.
k=10 59,61
k=11 65*,67
k=12 71,73
...
5,7,11,13,17,19,23,25,29,31,...,59,61,65,67,71,73,...
We observe that the difference between the terms is alternatively 2 and 4. Such a results may be obtained also using simple math. Is obvious that the difference between k*6+1 and k*6-1 is 2. It's simple to note that the difference between k*6+1 and (k+1)*6-1 is 4.
The function primality(x) returns x when x is prime (or 0 - take care) and the first divisor occurs when x is not prime.
I think you may obtain a better result inlining the primality() function inside the lpf5() function.
I've also tried to insert a table with some primes (from 1 to 383 - the primes in the first 128 results of the indicated formulas) inside the primality function, but the speed difference is unappreciable.
Here the code:
#include <stdio.h>
#include <math.h>
typedef long long unsigned int uint64;
uint64 lpf5(uint64 x);
uint64 primality(uint64 x);
uint64 lpf5(uint64 x)
{
uint64 x_=x;
while ( (x_=primality(x))!=x)
x=x/x_;
return x;
}
uint64 primality(uint64 x)
{
uint64 div=7,f=2,q;
if (x<4 || x==5)
return x;
if (!(x&1))
return 2;
if (!(x%3))
return 3;
if (!(x%5))
return 5;
q=sqrt(x);
while(div<=q) {
if (!(x%div)) {
return div;
}
f=6-f;
div+=f;
}
return x;
}
int main(void) {
uint64 x,k;
do {
printf("Input long int: ");
if (scanf("%llu",&x)<1)
break;
printf("Largest Prime Factor: %llu\n",lpf5(x));
} while(x!=0);
return 0;
}