C++ Bitset algorithm - c++

I am given a nxn grid with filled with 1 or 0. I want to count the number of subgrids where the corner tiles are all 1s. My solution goes through all pairs of rows and counts the number of matching 1s then it uses the formula numOf1s * (numOf1s-1)/2 and adds to the result. However, when I submit my solution on https://cses.fi/problemset/task/2137, there is no output on inputs with n = 3000 (probably caused by some error). What could the error be?
int main()
{
int n; cin>> n;
vector<bitset<3000>> grid(n);
for(int i=0;i<n;i++){
cin >> grid[i];
}
long result = 0;
for(int i=0;i<n-1;i++){
for(int j=i+1;j<n;j++){
int count = (grid[i]&grid[j]).count();
result += (count*(count-1))/2;
}
}
cout << result;
}

This solution will cause a time limit exceeded. bitset::count() is O(n) in worst case. The total complexity of your code is O(n^3). In the worst-case the number of operations would be 3000^3 > 10^10 which is too large.

I'm not sure this solution is the best you can come up with, but it is based on the original solution, with a homebrew alternative for the bitset. This allows me to work with 64 bits blocks, and using a fast popcnt(). An hardware version would be even better, as it would be to work with AVX registers, but this should be more portable and it works on cses.fi. Basically instead of generating a long intersection bitset and later count the number of ones, the function count_common() makes a piece of the intersection and immediately uses it just to count the ones.
The stream extractor could be probably improved, saving some more time.
#include <iostream>
#include <array>
#include <cstdint>
#include <climits>
uint64_t popcnt(uint64_t v) {
v = v - ((v >> 1) & (uint64_t)~(uint64_t)0 / 3);
v = (v & (uint64_t)~(uint64_t)0 / 15 * 3) + ((v >> 2) & (uint64_t)~(uint64_t)0 / 15 * 3);
v = (v + (v >> 4)) & (uint64_t)~(uint64_t)0 / 255 * 15;
uint64_t c = (uint64_t)(v * ((uint64_t)~(uint64_t)0 / 255)) >> (sizeof(uint64_t) - 1) * CHAR_BIT;
return c;
}
struct line {
uint64_t cells_[47] = { 0 }; // 3000/64 = 47
uint64_t& operator[](int pos) { return cells_[pos]; }
const uint64_t& operator[](int pos) const { return cells_[pos]; }
};
uint64_t count_common(const line& a, const line& b) {
uint64_t u = 0;
for (int i = 0; i < 47; ++i) {
u += popcnt(a[i] & b[i]);
}
return u;
}
std::istream& operator>>(std::istream& is, line& ln) {
is >> std::ws;
int pos = 0;
uint64_t val = 0;
while (true) {
char ch = is.get();
if (is && ch == '\n') {
break;
}
if (ch == '1') {
val |= 1LL << (63 - pos % 64);
}
if ((pos + 1) % 64 == 0) {
ln[pos / 64] = val;
val = 0;
}
++pos;
}
if (pos % 64 != 0) {
ln[pos / 64] = val;
}
return is;
}
struct grid {
int n_;
std::array<line, 3000> data_;
line& operator[](int r) {
return data_[r];
}
};
std::istream& operator>>(std::istream& is, grid& g) {
is >> g.n_;
for (int r = 0; r < g.n_; ++r) {
is >> g[r];
}
return is;
}
int main()
{
grid g;
std::cin >> g;
uint64_t count = 0;
for (int r1 = 0; r1 < g.n_; ++r1) {
for (int r2 = r1 + 1; r2 < g.n_; ++r2) {
uint64_t n = count_common(g[r1], g[r2]);
count += n * (n - 1) / 2;
}
}
std::cout << count << '\n';
return 0;
}

Related

How to get the minimum XOR of a given value and the value from a query of range for a given array

Given an array A of n integers and given queries in the form of range [l , r] and a value x, find the minimum of A[i] XOR x where l <= i <= r and x will be different for different queries.
I tried solving this problem using segment trees but I am not sure what type of information I should store in them as x will be different for different queries.
0 < number of queries <= 1e4
0 < n <= 1e4
To solve this I used a std::vector as basis (not an array, or std::array), just for flexibility.
#include <algorithm>
#include <stdexcept>
#include <vector>
int get_xored_max(const std::vector<int>& values, const size_t l, const size_t r, const int xor_value)
{
// check bounds of l and r
if ((l >= values.size()) || (r >= values.size()))
{
throw std::invalid_argument("index out of bounds");
}
// todo check l < r
// create left & right iterators to create a smaller vector
// only containing the subset we're interested in.
auto left = values.begin() + l;
auto right = values.begin() + r + 1;
std::vector<int> range{ left, right };
// xor all the values in the subset
for (auto& v : range)
{
v ^= xor_value;
}
// use the standard library function for finding the iterator to the maximum
// then use the * to dereference the iterator and get the value
auto max_value = *std::max_element(range.begin(), range.end());
return max_value;
}
int main()
{
std::vector<int> values{ 1,3,5,4,2,4,7,9 };
auto max_value = get_xored_max(values, 0u, 7u, 3);
return 0;
}
Approach - Trie + Offline Processing
Time Complexity - O(N32)
Space Complexity - O(N32)
Edit:
This Approach will fail. I guess, we have to use square root decomposition instead of two pointers approach.
I have solved this problem using Trie for finding minimum xor in a range of [l,r]. I solved queries by offline processing by sorting them.
Input format:
the first line has n (no. of elements) and q (no. of queries). the second line has all n elements of the array. each subsequent line has a query and each query has 3 inputs l, r and x.
Example -
Input -
3 3
2 1 2
1 2 3
1 3 2
2 3 5
First, convert all 3 queries into queries sorted by l and r.
converted queries -
1 2 3
1 3 2
2 3 5
Key here is processing over sorted queries using two pointers approach.
#include <bits/stdc++.h>
using namespace std;
const int N = (int)2e4 + 77;
int n, q, l, r, x;
int a[N], ans[N];
vector<pair<pair<int, int>, pair<int, int>>> queries;
// Trie Implementation starts
struct node
{
int nxt[2], cnt;
void newnode()
{
memset(nxt, 0, sizeof(nxt));
cnt = 0;
}
} trie[N * 32];
int tot = 1;
void update(int x, int v)
{
int p = 1;
for (int i = 31; i >= 0; i--)
{
int id = x >> i & 1;
if (!trie[p].nxt[id])
{
trie[++tot].newnode();
trie[p].nxt[id] = tot;
}
p = trie[p].nxt[id];
trie[p].cnt += v;
}
}
int minXor(int x)
{
int res = 0, p = 1;
for (int i = 31; i >= 0; i--)
{
int id = x >> i & 1;
if (trie[p].nxt[id] and trie[trie[p].nxt[id]].cnt)
p = trie[p].nxt[id];
else
{
p = trie[p].nxt[id ^ 1];
res |= 1 << i;
}
}
return res;
}
// Trie Implementation ends
int main()
{
cin >> n >> q;
for (int i = 1; i <= n; i += 1)
{
cin >> a[i];
}
for (int i = 1; i <= q; i += 1)
{
cin >> l >> r >> x;
queries.push_back({{l, r}, {x, i}});
}
sort(queries.begin(), queries.end());
int left = 1, right = 1;
for (int i = 0; i < q; i += 1)
{
int l = queries[i].first.first;
int r = queries[i].first.second;
int x = queries[i].second.first;
int index = queries[i].second.second;
while (left < l)
{
update(a[left], -1);
left += 1;
}
while (right <= r)
{
update(a[right], 1);
right += 1;
}
ans[index] = minXor(x);
}
for (int i = 1; i <= q; i += 1)
{
cout << ans[i] << " \n";
}
return 0;
}
Edit: with O(number of bits) code
Use a binary tree to store the values of A, look here : Minimum XOR for queries
What you need to change is adding to each node the range of indexes for A corresponding to the values in the leafs.
# minimal xor in a range
nbits=16 # Number of bits for numbers
asize=5000 # Array size
ntest=50 # Number of random test
from random import randrange
# Insert element a iindex iin the tree (increasing i only)
def tinsert(a,i,T):
for b in range(nbits-1,-1,-1):
v=((a>>b)&1)
T[v+2].append(i)
if T[v]==[]:T[v]=[[],[],[],[]]
T=T[v]
# Buildtree : builds a tree based on array V
def build(V):
T=[[],[],[],[]] # Init tree
for i,a in enumerate(V): tinsert(a,i,T)
return(T)
# Binary search : is T intersec [a,b] non empty ?
def binfind(T,a,b):
s,e,om=0,len(T)-1,-1
while True:
m=(s+e)>>1
v=T[m]
if v<a:
s=m
if m==om: return(a<=T[e]<=b)
elif v>b:
e=m
if m==om: return(a<=T[s]<=b)
else: return(True) # a<=T(m)<=b
om=m
# Look for the min xor in a give range index
def minx(x,s,e,T):
if s<0 or s>=(len(T[2])+len(T[3])) or e<s: return
r=0
for b in range(nbits-1,-1,-1):
v=((x>>b)&1)
if T[v+2]==[] or not binfind(T[v+2],s,e): # not nr with b set to v ?
v=1-v
T=T[v]
r=(r<<1)|v
return(r)
# Tests the code on random arrays
max=(1<<nbits)-1
for i in range(ntest):
A=[randrange(0,max) for i in range(asize)]
T=build(A)
x,s=randrange(0,max),randrange(0,asize-1)
e=randrange(s,asize)
if min(v^x for v in A[s:e+1])!=x^minx(x,s,e,T):
print('error')
I was able to solve this using segment tree and tries as suggested by #David Eisenstat
Below is an implementation in c++.
I constructed a trie for each segment in the segment tree. And finding the minimum xor is just traversing and matching the corresponding trie using each bit of the query value (here)
#include <bits/stdc++.h>
#define rep(i, a, b) for (int i = a; i < b; i++)
using namespace std;
const int bits = 7;
struct trie {
trie *children[2];
bool end;
};
trie *getNode(void)
{
trie *node = new trie();
node->end = false;
node->children[0] = NULL;
node->children[1] = NULL;
return node;
}
trie *merge(trie *l, trie *r)
{
trie *node = getNode();
// Binary 0:
if (l->children[0] && r->children[0])
node->children[0] = merge(l->children[0], r->children[0]);
else if (!r->children[0])
node->children[0] = l->children[0];
else if (!l->children[0])
node->children[0] = r->children[0];
// Binary 1:
if (l->children[1] && r->children[1])
node->children[1] = merge(l->children[1], r->children[1]);
else if (!r->children[1])
node->children[1] = l->children[1];
else if (!l->children[1])
node->children[1] = r->children[1];
return node;
}
void insert(trie *root, int num)
{
int mask = 1 << bits;
int bin;
rep(i, 0, bits + 1)
{
bin = ((num & mask) >> (bits - i));
if (!root->children[bin]) root->children[bin] = getNode();
root = root->children[bin];
mask = mask >> 1;
}
root->end = true;
}
struct _segTree {
int n, height, size;
vector<trie *> tree;
_segTree(int _n)
{
n = _n;
height = (int)ceil(log2(n));
size = (int)(2 * pow(2, height) - 1);
tree.resize(size);
}
trie *construct(vector<int> A, int start, int end, int idx)
{
if (start == end) {
tree[idx] = getNode();
insert(tree[idx], A[start]);
return tree[idx];
}
int mid = start + (end - start) / 2;
tree[idx] = merge(construct(A, start, mid, 2 * idx + 1),
construct(A, mid + 1, end, 2 * idx + 2));
return tree[idx];
}
int findMin(int num, trie *root)
{
int mask = 1 << bits;
int bin;
int rnum = 0;
int res = 0;
rep(i, 0, bits + 1)
{
bin = ((num & mask) >> (bits - i));
if (!root->children[bin]) {
bin = 1 - bin;
if (!root->children[bin]) return res ^ num;
}
rnum |= (bin << (bits - i));
root = root->children[bin];
if (root->end) res = rnum;
mask = mask >> 1;
}
return res ^ num;
}
int Query(int X, int start, int end, int qstart, int qend, int idx)
{
if (qstart <= start && qend >= end) return findMin(X, tree[idx]);
if (qstart > end || qend < start) return INT_MAX;
int mid = start + (end - start) / 2;
return min(Query(X, start, mid, qstart, qend, 2 * idx + 1),
Query(X, mid + 1, end, qstart, qend, 2 * idx + 2));
}
};
int main()
{
int n, q;
vector<int> A;
vector<int> L;
vector<int> R;
vector<int> X;
cin >> n;
A.resize(n, 0);
rep(i, 0, n) cin >> A[i];
cin >> q;
L.resize(q);
R.resize(q);
X.resize(q);
rep(i, 0, q) cin >> L[i] >> R[i] >> X[i];
//---------------------code--------------------//
_segTree segTree(n);
segTree.construct(A, 0, n - 1, 0);
rep(i, 0, q)
{
cout << segTree.Query(X[i], 0, n - 1, L[i], R[i], 0) << " ";
}
return 0;
}
Time complexity : O((2n - 1)*k + qklogn)
Space complexity : O((2n - 1)*2k)
k -> number of bits

How to perform BigInteger multiplication on char vectors?

I'm implementing a BigInt in c++ and am trying to overload the multiplication operator. I'm storing large integers in a char vector.
vector<char> storage;
Here is what I did to implement operator*(int)
BigInt BigInt::operator*(int x)
{
int extra = 0;
int dec_mod = pow(10, this->storage.size());
for (auto & g : storage) {
g = g * x + extra;
int mod_g = g % dec_mod;
extra = g / dec_mod;
g = mod_g;
}
while (extra > 0) {
storage.push_back(extra % dec_mod);
extra /= dec_mod;
}
return *this;
}
The operator*(bigInt) function returns wrong answers. For example, 33 * 4 returns 1212 and not 132.This was my attempt at writing the overloaded operator* which takes a bigint object:
BigInt BigInt::operator*(BigInt bigN) {
int carry = 0;
for (int i = bigN.storage.size()-1; i >= 0; i--) {
for (int j = this->storage.size()-1; j >= 0; j--) {
int val = (this->storage.at(i) * bigN.storage.at(j)) + carry;
this->storage.push_back(val % 10);
carry = val / 10;
}
}
return *this;
}
It looks like the logic in the carry is flawed, but i'm not sure how to fix it.
I'm not sure how you're trying to do this, but here is a walkthrough of why you're getting the result 1212 instead of 132:
BigInt operator*(int x)// x is 4
{
// Let's say storage holds 33, that's
// {3, 3} in your char vector;
int extra = 0;
int dec_mod = pow(10, this->storage.size()); // dec_mod may be 100
for (auto & g : storage)
{
g = g * x + extra; // same as g = 3 * 4 + 0, g = 12
int mod_g = g % dec_mod; // same as mod_g = 12 % 100 = 12
extra = g / dec_mod; // same as 12 / 100 = 0
g = mod_g; // g = 12
}
// Exact same thing happens on the second iteration, your storage vector
// ends up as {12, 12};
// That's why your result is 1212
while (extra > 0) {
storage.push_back(extra % dec_mod);
extra /= dec_mod;
}
return *this;
}
I'm not sure how you are trying to do it, but here's my attempt, it's just as one would do it on paper:
#include <iostream>
#include <string>
#include <vector>
using namespace std;
struct BigInt
{
BigInt(std::string num) { for (auto &i : num) storage.push_back(i - 48); }
BigInt(std::vector<char> numVect) : storage(numVect) {}
vector<char> storage;
string getAsString()
{ string str; for (auto& i : storage) str += i + 48; return str; }
// Add 48 to turn 0 - 9 to ascii string.
vector<char> add(vector<char>& lhs, vector<char>& rhs)
// Add function only needed if number is multiplied by more than one digit.
{
// Fill with zeros to make both vectors same length.
int sizeDiff = (int)lhs.size() - (int)rhs.size();
if (sizeDiff < 0)
lhs.insert(lhs.begin(), abs(sizeDiff), 0);
else if (sizeDiff > 0)
rhs.insert(rhs.begin(), abs(sizeDiff), 0);
vector<char> resultVect;
int carry = 0;
for (int i = lhs.size() - 1; i >= 0; --i)
{
int result = lhs[i] + rhs[i] + carry;
carry = result / 10;
result %= 10;
resultVect.insert(resultVect.begin(), result);
}
if (carry != 0) resultVect.insert(resultVect.begin(), carry);
return resultVect;
}
BigInt operator*(BigInt rhs)
{
int unitPlace = 0; // Keeps track of how many zeros to add in subsequent results
vector<char> totalVect; // Accumulated value after each addition
vector<char> resultVect; // Result of this particular multiplication
for (int i = rhs.storage.size() - 1; i >= 0; --i, unitPlace++)
{
int carry = 0;
for (int k = 0; k < unitPlace; ++k) resultVect.push_back(0);
for (int j = storage.size() - 1; j >= 0; j--)
{
int result = rhs.storage[i] * storage[j] + carry;
carry = result / 10;
result %= 10;
resultVect.insert(resultVect.begin(), result);
}
resultVect.insert(resultVect.begin(), carry);
totalVect = add(totalVect, resultVect); // Add sub-result
resultVect.clear();
}
// Strip leading zeros
for (int i = 0; i < totalVect.size(); ++i) {
if (totalVect[i] == 0) totalVect.erase(totalVect.begin() + i);
else break;
}
return BigInt{ totalVect };
}
};
int main()
{
BigInt a{ "335467" };
BigInt b{ "1019737" };
BigInt c = a * b;
std::cout << c.getAsString() << '\n';
cin.ignore();
return 0;
}

How to convert a decimal string to binary string?

I have a decimal string like this (length < 5000):
std::string decimalString = "555";
Is there a standard way to convert this string to binary representation? Like this:
std::string binaryString = "1000101011";
Update.
This post helps me.
As the number is very large, you can use a big integer library (boost, maybe?), or write the necessary functions yourself.
If you decide to implement the functions yourself, one way is to implement the old pencil-and-paper long division method in your code, where you'll need to divide the decimal number repeatedly by 2 and accumulate the remainders in another string. May be a little cumbersome, but division by 2 should not be so hard.
Since 10 is not a power of two (or the other way round), you're out of luck. You will have to implement arithmetics in base-10. You need the following two operations:
Integer division by 2
Checking the remainder after division by 2
Both can be computed by the same algorithm.
Alternatively, you can use one of the various big integer libraries for C++, such as GNU MP or Boost.Multiprecision.
I tried to do it.. I don't think my answer is right but here is the IDEA behind what I was trying to do..
Lets say we have 2 decimals:
100 and 200..
To concatenate these, we can use the formula:
a * CalcPower(b) + b where CalcPower is defined below..
Knowing this, I tried to split the very long decimal string into chunks of 4. I convert each string to binary and store them in a vector..
Finally, I go through each string and apply the formula above to concatenate each binary string into one massive one..
I didn't get it working but here is the code.. maybe someone else see where I went wrong.. BinaryAdd, BinaryMulDec, CalcPower works perfectly fine.. the problem is actually in ToBinary
#include <iostream>
#include <bitset>
#include <limits>
#include <algorithm>
std::string BinaryAdd(std::string First, std::string Second)
{
int Carry = 0;
std::string Result;
while(Second.size() > First.size())
First.insert(0, "0");
while(First.size() > Second.size())
Second.insert(0, "0");
for (int I = First.size() - 1; I >= 0; --I)
{
int FirstBit = First[I] - 0x30;
int SecondBit = Second[I] - 0x30;
Result += static_cast<char>((FirstBit ^ SecondBit ^ Carry) + 0x30);
Carry = (FirstBit & SecondBit) | (SecondBit & Carry) | (FirstBit & Carry);
}
if (Carry)
Result += 0x31;
std::reverse(Result.begin(), Result.end());
return Result;
}
std::string BinaryMulDec(std::string value, int amount)
{
if (amount == 0)
{
for (auto &s : value)
{
s = 0x30;
}
return value;
}
std::string result = value;
for (int I = 0; I < amount - 1; ++I)
result = BinaryAdd(result, value);
return result;
}
std::int64_t CalcPowers(std::int64_t value)
{
std::int64_t t = 1;
while(t < value)
t *= 10;
return t;
}
std::string ToBinary(const std::string &value)
{
std::vector<std::string> sets;
std::vector<int> multipliers;
int Len = 0;
int Rem = value.size() % 4;
for (auto it = value.end(), jt = value.end(); it != value.begin() - 1; --it)
{
if (Len++ == 4)
{
std::string t = std::string(it, jt);
sets.push_back(std::bitset<16>(std::stoull(t)).to_string());
multipliers.push_back(CalcPowers(std::stoull(t)));
jt = it;
Len = 1;
}
}
if (Rem != 0 && Rem != value.size())
{
sets.push_back(std::bitset<16>(std::stoull(std::string(value.begin(), value.begin() + Rem))).to_string());
}
auto formula = [](std::string a, std::string b, int mul) -> std::string
{
return BinaryAdd(BinaryMulDec(a, mul), b);
};
std::reverse(sets.begin(), sets.end());
std::reverse(multipliers.begin(), multipliers.end());
std::string result = sets[0];
for (std::size_t i = 1; i < sets.size(); ++i)
{
result = formula(result, sets[i], multipliers[i]);
}
return result;
}
void ConcatenateDecimals(std::int64_t* arr, int size)
{
auto formula = [](std::int64_t a, std::int64_t b) -> std::int64_t
{
return (a * CalcPowers(b)) + b;
};
std::int64_t val = arr[0];
for (int i = 1; i < size; ++i)
{
val = formula(val, arr[i]);
}
std::cout<<val;
}
int main()
{
std::string decimal = "64497387062899840145";
//6449738706289984014 = 0101100110000010000100110010111001100010100000001000001000001110
/*
std::int64_t arr[] = {644, 9738, 7062, 8998, 4014};
ConcatenateDecimals(arr, 5);*/
std::cout<<ToBinary(decimal);
return 0;
}
I found my old code that solve sport programming task:
ai -> aj
2 <= i,j <= 36; 0 <= a <= 10^1000
time limit: 1sec
Execution time was ~0,039 in worst case. Multiplication, addition and division algorithms is very fast because of using 10^9 as numeration system, but implementation can be optimized very well I think.
source link
#include <iostream>
#include <string>
#include <vector>
using namespace std;
#define sz(x) (int((x).size()))
typedef vector<int> vi;
typedef long long llong;
int DigToNumber(char c) {
if( c <= '9' && c >= '0' )
return c-'0';
return c-'A'+10;
}
char NumberToDig(int n) {
if( n < 10 )
return '0'+n;
return n-10+'A';
}
const int base = 1000*1000*1000;
void mulint(vi& a, int b) { //a*= b
for(int i = 0, carry = 0; i < sz(a) || carry; i++) {
if( i == sz(a) )
a.push_back(0);
llong cur = carry + a[i] * 1LL * b;
a[i] = int(cur%base);
carry = int(cur/base);
}
while( sz(a) > 1 && a.back() == 0 )
a.pop_back();
}
int divint(vi& a, int d) { // carry = a%d; a /= d; return carry;
int carry = 0;
for(int i = sz(a)-1; i >= 0; i--) {
llong cur = a[i] + carry * 1LL * base;
a[i] = int(cur/d);
carry = int(cur%d);
}
while( sz(a) > 1 && a.back() == 0 )
a.pop_back();
return carry;
}
void add(vi& a, vi& b) { // a += b
for(int i = 0, c = 0, l = max(sz(a),sz(b)); i < l || c; i++) {
if( i == sz(a) )
a.push_back(0);
a[i] += ((i<sz(b))?b[i]:0) + c;
c = a[i] >= base;
if( c ) a[i] -= base;
}
}
int main() {
ios_base::sync_with_stdio(0);
cin.tie(0);
int from, to; cin >> from >> to;
string s; cin >> s;
vi res(1,0); vi m(1,1); vi tmp;
for(int i = sz(s)-1; i >= 0; i--) {
tmp.assign(m.begin(), m.end());
mulint(tmp,DigToNumber(s[i]));
add(res,tmp); mulint(m,from);
}
vi ans;
while( sz(res) > 1 || res.back() != 0 )
ans.push_back(divint(res,to));
if( sz(ans) == 0 )
ans.push_back(0);
for(int i = sz(ans)-1; i >= 0; i--)
cout << NumberToDig(ans[i]);
cout << "\n";
return 0;
}
How "from -> to" works for string "s":
accumulate Big Number (vector< int >) "res" with s[i]*from^(|s|-i-1), i = |s|-1..0
compute digits by dividing "res" by "to" until res > 0 and save them to another vector
send it to output digit-by-digit (you can use ostringstream instead)
PS I've noted that nickname of thread starter is Denis. And I think this link may be useful too.

Storing a Big Number in a Variable and Looping

How can i store a big number in a variable and use a for loop?
I have a very big number 75472202764752234070123900087933251 and i need to loop from 0 to this number!
Is it even possible to do this? how much time will it take to end?
EDIT: i am trying to solve a hard problem by brute force. its a combination problem.the bruteforcing cases may reach 470C450.
so i guess i should use a different algorithm...
This might take
0.23 x 10^23 years if C++ processed 100,000 loops per second :|
http://www.wolframalpha.com/input/?i=75472202764752234070123900087933251%2F%28100000*1*3600*24*365%29
It looks that this number fits into 128 bit. So you could use a modern system and a modern compiler that implements such numbers. This would e.g be the case for a 64bit linux system with gcc as a compiler. This has something like __uint128_t that you could use.
Obviously you can't use such a variable as a for-loop variable, others have give you the calculations. But you could use it to store some of your calculations.
Well, you would need an implementation that can handle at least a subset of the initialization, boolean, and arithmetic functions on very large integers. Something like: https://mattmccutchen.net/bigint/.
For something that would give a bit better performance than a general large integer math library, you could use specialized operations specifically to allow use of a large integer as a counter. For an example of this, see dewtell's updated answer to this question.
As for it being possible for you to loop from 0 to that number: well, yes, it is possible to write the code for it with one of the above solutions, but I think the answer is no, you personally will not be able to do it because you will not be alive to see it finish.
[edit: Yes, I would definitely recommend you find a different algorithm. :D]
If you need to loop a certain number of times, and that number is greater than 2^64, just use while(1) because your computer will break before it counts up to 2^64 anyway.
There's no need for a complete bignum package - if all you need is a loop counter, here's a simple byte counter that uses an array of bytes as a counter. It stops when the byte array wraps around to all zeros again. If you wanted to count to some other value than 2^(bytesUsed*CHAR_BITS), you could just compute the two's complement value of the negative of the number of iterations you wanted, and let it count up to 0, keeping in mind that bytes[0] is the low-order byte (or use the positive value and count down instead of up).
#include <stdio.h>
#define MAXBYTES 20
/* Simple byte counter - note it uses argc as # of bytes to use for convenience */
int main(int argc, char **argv) {
unsigned char bytes[MAXBYTES];
const int bytesUsed = argc < MAXBYTES? argc : MAXBYTES;
int i;
unsigned long counter = (unsigned long)-1; /* to give loop something to do */
for (i = 0; i < bytesUsed; i++) bytes[i] = 0; /* Initialize bytes */
do {
for (i = 0; i < bytesUsed && !++bytes[i]; i++) ; /* NULL BODY - this is the byte counter */
counter++;
} while (i < bytesUsed);
printf("With %d bytes used, final counter value = %lu\n", bytesUsed, counter);
}
Run times for the first 4 values (under Cygwin, on a Lenovo T61):
$ time ./bytecounter
With 1 bytes used, final counter value = 255
real 0m0.078s
user 0m0.031s
sys 0m0.046s
$ time ./bytecounter a
With 2 bytes used, final counter value = 65535
real 0m0.063s
user 0m0.031s
sys 0m0.031s
$ time ./bytecounter a a
With 3 bytes used, final counter value = 16777215
real 0m0.125s
user 0m0.015s
sys 0m0.046s
$ time ./bytecounter a a a
With 4 bytes used, final counter value = 4294967295
real 0m6.578s
user 0m0.015s
sys 0m0.047s
At this rate, five bytes should take around half an hour, and six bytes should take the better part of a week. Of course the counter value will be inaccurate for those - it's mostly just there to verify the number of iterations for the smaller byte values and give the loop something to do.
Edit: And here's the time for five bytes, around half an hour as I predicted:
$ time ./bytecounter a a a a
With 5 bytes used, final counter value = 4294967295
real 27m22.184s
user 0m0.015s
sys 0m0.062s
Ok, here's code to take an arbitrary decimal number passed as the first arg and count down from it to zero. I set it up to allow the counter to use different size elements (just change the typedef for COUNTER_BASE), but it turns out that bytes are actually somewhat faster than either short or long on my system.
#include <stdio.h>
#include <limits.h> // defines CHAR_BIT
#include <ctype.h>
#include <vector>
using std::vector;
typedef unsigned char COUNTER_BASE;
typedef vector<COUNTER_BASE> COUNTER;
typedef vector<unsigned char> BYTEVEC;
const unsigned long byteMask = (~0ul) << CHAR_BIT;
const size_t MAXBYTES=20;
void mult10(BYTEVEC &val) {
// Multiply value by 10
unsigned int carry = 0;
int i;
for (i = 0; i < val.size(); i++) {
unsigned long value = val[i]*10ul+carry;
carry = (value & byteMask) >> CHAR_BIT;
val[i] = value & ~byteMask;
}
if (carry > 0) val.push_back(carry);
}
void addDigit(BYTEVEC &val, const char digit) {
// Add digit to the number in BYTEVEC.
unsigned int carry = digit - '0'; // Assumes ASCII char set
int i;
for (i = 0; i < val.size() && carry; i++) {
unsigned long value = static_cast<unsigned long>(val[i])+carry;
carry = (value & byteMask) >> CHAR_BIT;
val[i] = value & ~byteMask;
}
if (carry > 0) val.push_back(carry);
}
BYTEVEC Cstr2Bytevec(const char *str) {
// Turn a C-style string into a BYTEVEC. Only the digits in str apply,
// so that one can use commas, underscores, or other non-digits to separate
// digit groups.
BYTEVEC result;
result.reserve(MAXBYTES);
result[0]=0;
unsigned char *res=&result[0]; // For debugging
while (*str) {
if (isdigit(static_cast<int>(*str))) {
mult10(result);
addDigit(result, *str);
}
str++;
}
return result;
}
void packCounter(COUNTER &ctr, const BYTEVEC &val) {
// Pack the bytes from val into the (possibly larger) datatype of COUNTER
int i;
ctr.erase(ctr.begin(), ctr.end());
COUNTER_BASE value = 0;
for (i = 0; i < val.size(); i++) {
int pos = i%sizeof(COUNTER_BASE); // position of this byte in the value
if (i > 0 && pos == 0) {
ctr.push_back(value);
value = val[i];
} else {
value |= static_cast<COUNTER_BASE>(val[i]) << pos*CHAR_BIT;
}
}
ctr.push_back(value);
}
inline bool decrementAndTest(COUNTER &ctr) {
// decrement value in ctr and return true if old value was not all zeros
int i;
for (i = 0; i < ctr.size() && !(ctr[i]--); i++) ; // EMPTY BODY
return i < ctr.size();
}
inline bool decrementAndTest2(COUNTER_BASE *ctr, const size_t size) {
// decrement value in ctr and return true if old value was not all zeros
int i;
for (i = 0; i < size && !(ctr[i]--); i++) ; // EMPTY BODY
return i < size;
}
/* Vector counter - uses first arg (if supplied) as the count */
int main(int argc, const char *argv[]) {
BYTEVEC limit = Cstr2Bytevec(argc > 1? argv[1] : "0");
COUNTER ctr;
packCounter(ctr, limit);
COUNTER_BASE *ctr_vals = ctr.size() > 0 ? &ctr[0] : NULL;
size_t ctr_size = ctr.size();
unsigned long ul_counter = 0ul; /* to give loop something to do */
while(decrementAndTest2(ctr_vals, ctr_size)) {
ul_counter++;
};
printf("With %d bytes used, final ul_counter value = %lu\n", limit.size(), ul_counter);
return 0;
}
Examples of use:
$ time ./bigcounter 5
With 1 bytes used, final ul_counter value = 5
real 0m0.094s
user 0m0.031s
sys 0m0.047s
$ time ./bigcounter 5,000
With 2 bytes used, final ul_counter value = 5000
real 0m0.062s
user 0m0.015s
sys 0m0.062s
$ time ./bigcounter 5,000,000
With 3 bytes used, final ul_counter value = 5000000
real 0m0.093s
user 0m0.015s
sys 0m0.046s
$ time ./bigcounter 1,000,000,000
With 4 bytes used, final ul_counter value = 1000000000
real 0m2.688s
user 0m0.015s
sys 0m0.015s
$ time ./bigcounter 2,000,000,000
With 4 bytes used, final ul_counter value = 2000000000
real 0m5.125s
user 0m0.015s
sys 0m0.046s
$ time ./bigcounter 3,000,000,000
With 4 bytes used, final ul_counter value = 3000000000
real 0m7.485s
user 0m0.031s
sys 0m0.047s
$ time ./bigcounter 4,000,000,000
With 4 bytes used, final ul_counter value = 4000000000
real 0m9.875s
user 0m0.015s
sys 0m0.046s
$ time ./bigcounter 5,000,000,000
With 5 bytes used, final ul_counter value = 705032704
real 0m12.594s
user 0m0.046s
sys 0m0.015s
$ time ./bigcounter 6,000,000,000
With 5 bytes used, final ul_counter value = 1705032704
real 0m14.813s
user 0m0.015s
sys 0m0.062s
Unwrapping the counter vector into C-style data structures (i.e., using decrementAndTest2 instead of decrementAndTest) sped things up by around 20-25%, but the code is still about twice as slow as my previous C program for similar-sized examples (around 4 billion). This is with MS Visual C++ 6.0 as the compiler in release mode, optimizing for speed, on a 2GHz dual-core system, for both programs. Inlining the decrementAndTest2 function definitely makes a big difference (around 12 sec. vs. 30 for the 5 billion loop), but I'll have to see whether physically inlining the code as I did in the C program can get similar performance.
the variable in main function can Store even 100 factorial
#include <iostream>
#include <cstdio>
#include <vector>
#include <cstring>
#include <string>
#include <map>
#include <functional>
#include <algorithm>
#include <cstdlib>
#include <iomanip>
#include <stack>
#include <queue>
#include <deque>
#include <limits>
#include <cmath>
#include <numeric>
#include <set>
using namespace std;
//template for BIGINIT
// base and base_digits must be consistent
const int base = 10;
const int base_digits = 1;
struct bigint {
vector<int> a;
int sign;
bigint() :
sign(1) {
}
bigint(long long v) {
*this = v;
}
bigint(const string &s) {
read(s);
}
void operator=(const bigint &v) {
sign = v.sign;
a = v.a;
}
void operator=(long long v) {
sign = 1;
if (v < 0)
sign = -1, v = -v;
for (; v > 0; v = v / base)
a.push_back(v % base);
}
bigint operator+(const bigint &v) const {
if (sign == v.sign) {
bigint res = v;
for (int i = 0, carry = 0; i < (int) max(a.size(), v.a.size()) || carry; ++i) {
if (i == (int) res.a.size())
res.a.push_back(0);
res.a[i] += carry + (i < (int) a.size() ? a[i] : 0);
carry = res.a[i] >= base;
if (carry)
res.a[i] -= base;
}
return res;
}
return *this - (-v);
}
bigint operator-(const bigint &v) const {
if (sign == v.sign) {
if (abs() >= v.abs()) {
bigint res = *this;
for (int i = 0, carry = 0; i < (int) v.a.size() || carry; ++i) {
res.a[i] -= carry + (i < (int) v.a.size() ? v.a[i] : 0);
carry = res.a[i] < 0;
if (carry)
res.a[i] += base;
}
res.trim();
return res;
}
return -(v - *this);
}
return *this + (-v);
}
void operator*=(int v) {
if (v < 0)
sign = -sign, v = -v;
for (int i = 0, carry = 0; i < (int) a.size() || carry; ++i) {
if (i == (int) a.size())
a.push_back(0);
long long cur = a[i] * (long long) v + carry;
carry = (int) (cur / base);
a[i] = (int) (cur % base);
//asm("divl %%ecx" : "=a"(carry), "=d"(a[i]) : "A"(cur), "c"(base));
}
trim();
}
bigint operator*(int v) const {
bigint res = *this;
res *= v;
return res;
}
friend pair<bigint, bigint> divmod(const bigint &a1, const bigint &b1) {
int norm = base / (b1.a.back() + 1);
bigint a = a1.abs() * norm;
bigint b = b1.abs() * norm;
bigint q, r;
q.a.resize(a.a.size());
for (int i = a.a.size() - 1; i >= 0; i--) {
r *= base;
r += a.a[i];
int s1 = r.a.size() <= b.a.size() ? 0 : r.a[b.a.size()];
int s2 = r.a.size() <= b.a.size() - 1 ? 0 : r.a[b.a.size() - 1];
int d = ((long long) base * s1 + s2) / b.a.back();
r -= b * d;
while (r < 0)
r += b, --d;
q.a[i] = d;
}
q.sign = a1.sign * b1.sign;
r.sign = a1.sign;
q.trim();
r.trim();
return make_pair(q, r / norm);
}
bigint operator/(const bigint &v) const {
return divmod(*this, v).first;
}
bigint operator%(const bigint &v) const {
return divmod(*this, v).second;
}
void operator/=(int v) {
if (v < 0)
sign = -sign, v = -v;
for (int i = (int) a.size() - 1, rem = 0; i >= 0; --i) {
long long cur = a[i] + rem * (long long) base;
a[i] = (int) (cur / v);
rem = (int) (cur % v);
}
trim();
}
bigint operator/(int v) const {
bigint res = *this;
res /= v;
return res;
}
int operator%(int v) const {
if (v < 0)
v = -v;
int m = 0;
for (int i = a.size() - 1; i >= 0; --i)
m = (a[i] + m * (long long) base) % v;
return m * sign;
}
void operator+=(const bigint &v) {
*this = *this + v;
}
void operator-=(const bigint &v) {
*this = *this - v;
}
void operator*=(const bigint &v) {
*this = *this * v;
}
void operator/=(const bigint &v) {
*this = *this / v;
}
bool operator<(const bigint &v) const {
if (sign != v.sign)
return sign < v.sign;
if (a.size() != v.a.size())
return a.size() * sign < v.a.size() * v.sign;
for (int i = a.size() - 1; i >= 0; i--)
if (a[i] != v.a[i])
return a[i] * sign < v.a[i] * sign;
return false;
}
bool operator>(const bigint &v) const {
return v < *this;
}
bool operator<=(const bigint &v) const {
return !(v < *this);
}
bool operator>=(const bigint &v) const {
return !(*this < v);
}
bool operator==(const bigint &v) const {
return !(*this < v) && !(v < *this);
}
bool operator!=(const bigint &v) const {
return *this < v || v < *this;
}
void trim() {
while (!a.empty() && !a.back())
a.pop_back();
if (a.empty())
sign = 1;
}
bool isZero() const {
return a.empty() || (a.size() == 1 && !a[0]);
}
bigint operator-() const {
bigint res = *this;
res.sign = -sign;
return res;
}
bigint abs() const {
bigint res = *this;
res.sign *= res.sign;
return res;
}
long long longValue() const {
long long res = 0;
for (int i = a.size() - 1; i >= 0; i--)
res = res * base + a[i];
return res * sign;
}
friend bigint gcd(const bigint &a, const bigint &b) {
return b.isZero() ? a : gcd(b, a % b);
}
friend bigint lcm(const bigint &a, const bigint &b) {
return a / gcd(a, b) * b;
}
void read(const string &s) {
sign = 1;
a.clear();
int pos = 0;
while (pos < (int) s.size() && (s[pos] == '-' || s[pos] == '+')) {
if (s[pos] == '-')
sign = -sign;
++pos;
}
for (int i = s.size() - 1; i >= pos; i -= base_digits) {
int x = 0;
for (int j = max(pos, i - base_digits + 1); j <= i; j++)
x = x * 10 + s[j] - '0';
a.push_back(x);
}
trim();
}
friend istream& operator>>(istream &stream, bigint &v) {
string s;
stream >> s;
v.read(s);
return stream;
}
friend ostream& operator<<(ostream &stream, const bigint &v) {
if (v.sign == -1)
stream << '-';
stream << (v.a.empty() ? 0 : v.a.back());
for (int i = (int) v.a.size() - 2; i >= 0; --i)
stream << setw(base_digits) << setfill('0') << v.a[i];
return stream;
}
static vector<int> convert_base(const vector<int> &a, int old_digits, int new_digits) {
vector<long long> p(max(old_digits, new_digits) + 1);
p[0] = 1;
for (int i = 1; i < (int) p.size(); i++)
p[i] = p[i - 1] * 10;
vector<int> res;
long long cur = 0;
int cur_digits = 0;
for (int i = 0; i < (int) a.size(); i++) {
cur += a[i] * p[cur_digits];
cur_digits += old_digits;
while (cur_digits >= new_digits) {
res.push_back(int(cur % p[new_digits]));
cur /= p[new_digits];
cur_digits -= new_digits;
}
}
res.push_back((int) cur);
while (!res.empty() && !res.back())
res.pop_back();
return res;
}
typedef vector<long long> vll;
static vll karatsubaMultiply(const vll &a, const vll &b) {
int n = a.size();
vll res(n + n);
if (n <= 32) {
for (int i = 0; i < n; i++)
for (int j = 0; j < n; j++)
res[i + j] += a[i] * b[j];
return res;
}
int k = n >> 1;
vll a1(a.begin(), a.begin() + k);
vll a2(a.begin() + k, a.end());
vll b1(b.begin(), b.begin() + k);
vll b2(b.begin() + k, b.end());
vll a1b1 = karatsubaMultiply(a1, b1);
vll a2b2 = karatsubaMultiply(a2, b2);
for (int i = 0; i < k; i++)
a2[i] += a1[i];
for (int i = 0; i < k; i++)
b2[i] += b1[i];
vll r = karatsubaMultiply(a2, b2);
for (int i = 0; i < (int) a1b1.size(); i++)
r[i] -= a1b1[i];
for (int i = 0; i < (int) a2b2.size(); i++)
r[i] -= a2b2[i];
for (int i = 0; i < (int) r.size(); i++)
res[i + k] += r[i];
for (int i = 0; i < (int) a1b1.size(); i++)
res[i] += a1b1[i];
for (int i = 0; i < (int) a2b2.size(); i++)
res[i + n] += a2b2[i];
return res;
}
bigint operator*(const bigint &v) const {
vector<int> a6 = convert_base(this->a, base_digits, 6);
vector<int> b6 = convert_base(v.a, base_digits, 6);
vll a(a6.begin(), a6.end());
vll b(b6.begin(), b6.end());
while (a.size() < b.size())
a.push_back(0);
while (b.size() < a.size())
b.push_back(0);
while (a.size() & (a.size() - 1))
a.push_back(0), b.push_back(0);
vll c = karatsubaMultiply(a, b);
bigint res;
res.sign = sign * v.sign;
for (int i = 0, carry = 0; i < (int) c.size(); i++) {
long long cur = c[i] + carry;
res.a.push_back((int) (cur % 1000000));
carry = (int) (cur / 1000000);
}
res.a = convert_base(res.a, 6, base_digits);
res.trim();
return res;
}
};
//use : bigint var;
//template for biginit over
int main()
{
bigint var=10909000890789;
cout<<var;
return 0;
}

Modular Exponentiation for high numbers in C++

So I've been working recently on an implementation of the Miller-Rabin primality test. I am limiting it to a scope of all 32-bit numbers, because this is a just-for-fun project that I am doing to familiarize myself with c++, and I don't want to have to work with anything 64-bits for awhile. An added bonus is that the algorithm is deterministic for all 32-bit numbers, so I can significantly increase efficiency because I know exactly what witnesses to test for.
So for low numbers, the algorithm works exceptionally well. However, part of the process relies upon modular exponentiation, that is (num ^ pow) % mod. so, for example,
3 ^ 2 % 5 =
9 % 5 =
4
here is the code I have been using for this modular exponentiation:
unsigned mod_pow(unsigned num, unsigned pow, unsigned mod)
{
unsigned test;
for(test = 1; pow; pow >>= 1)
{
if (pow & 1)
test = (test * num) % mod;
num = (num * num) % mod;
}
return test;
}
As you might have already guessed, problems arise when the arguments are all exceptionally large numbers. For example, if I want to test the number 673109 for primality, I will at one point have to find:
(2 ^ 168277) % 673109
now 2 ^ 168277 is an exceptionally large number, and somewhere in the process it overflows test, which results in an incorrect evaluation.
on the reverse side, arguments such as
4000111222 ^ 3 % 1608
also evaluate incorrectly, for much the same reason.
Does anyone have suggestions for modular exponentiation in a way that can prevent this overflow and/or manipulate it to produce the correct result? (the way I see it, overflow is just another form of modulo, that is num % (UINT_MAX+1))
Exponentiation by squaring still "works" for modulo exponentiation. Your problem isn't that 2 ^ 168277 is an exceptionally large number, it's that one of your intermediate results is a fairly large number (bigger than 2^32), because 673109 is bigger than 2^16.
So I think the following will do. It's possible I've missed a detail, but the basic idea works, and this is how "real" crypto code might do large mod-exponentiation (although not with 32 and 64 bit numbers, rather with bignums that never have to get bigger than 2 * log (modulus)):
Start with exponentiation by squaring, as you have.
Perform the actual squaring in a 64-bit unsigned integer.
Reduce modulo 673109 at each step to get back within the 32-bit range, as you do.
Obviously that's a bit awkward if your C++ implementation doesn't have a 64 bit integer, although you can always fake one.
There's an example on slide 22 here: http://www.cs.princeton.edu/courses/archive/spr05/cos126/lectures/22.pdf, although it uses very small numbers (less than 2^16), so it may not illustrate anything you don't already know.
Your other example, 4000111222 ^ 3 % 1608 would work in your current code if you just reduce 4000111222 modulo 1608 before you start. 1608 is small enough that you can safely multiply any two mod-1608 numbers in a 32 bit int.
I wrote something for this recently for RSA in C++, bit messy though.
#include "BigInteger.h"
#include <iostream>
#include <sstream>
#include <stack>
BigInteger::BigInteger() {
digits.push_back(0);
negative = false;
}
BigInteger::~BigInteger() {
}
void BigInteger::addWithoutSign(BigInteger& c, const BigInteger& a, const BigInteger& b) {
int sum_n_carry = 0;
int n = (int)a.digits.size();
if (n < (int)b.digits.size()) {
n = b.digits.size();
}
c.digits.resize(n);
for (int i = 0; i < n; ++i) {
unsigned short a_digit = 0;
unsigned short b_digit = 0;
if (i < (int)a.digits.size()) {
a_digit = a.digits[i];
}
if (i < (int)b.digits.size()) {
b_digit = b.digits[i];
}
sum_n_carry += a_digit + b_digit;
c.digits[i] = (sum_n_carry & 0xFFFF);
sum_n_carry >>= 16;
}
if (sum_n_carry != 0) {
putCarryInfront(c, sum_n_carry);
}
while (c.digits.size() > 1 && c.digits.back() == 0) {
c.digits.pop_back();
}
//std::cout << a.toString() << " + " << b.toString() << " == " << c.toString() << std::endl;
}
void BigInteger::subWithoutSign(BigInteger& c, const BigInteger& a, const BigInteger& b) {
int sub_n_borrow = 0;
int n = a.digits.size();
if (n < (int)b.digits.size())
n = (int)b.digits.size();
c.digits.resize(n);
for (int i = 0; i < n; ++i) {
unsigned short a_digit = 0;
unsigned short b_digit = 0;
if (i < (int)a.digits.size())
a_digit = a.digits[i];
if (i < (int)b.digits.size())
b_digit = b.digits[i];
sub_n_borrow += a_digit - b_digit;
if (sub_n_borrow >= 0) {
c.digits[i] = sub_n_borrow;
sub_n_borrow = 0;
} else {
c.digits[i] = 0x10000 + sub_n_borrow;
sub_n_borrow = -1;
}
}
while (c.digits.size() > 1 && c.digits.back() == 0) {
c.digits.pop_back();
}
//std::cout << a.toString() << " - " << b.toString() << " == " << c.toString() << std::endl;
}
int BigInteger::cmpWithoutSign(const BigInteger& a, const BigInteger& b) {
int n = (int)a.digits.size();
if (n < (int)b.digits.size())
n = (int)b.digits.size();
//std::cout << "cmp(" << a.toString() << ", " << b.toString() << ") == ";
for (int i = n-1; i >= 0; --i) {
unsigned short a_digit = 0;
unsigned short b_digit = 0;
if (i < (int)a.digits.size())
a_digit = a.digits[i];
if (i < (int)b.digits.size())
b_digit = b.digits[i];
if (a_digit < b_digit) {
//std::cout << "-1" << std::endl;
return -1;
} else if (a_digit > b_digit) {
//std::cout << "+1" << std::endl;
return +1;
}
}
//std::cout << "0" << std::endl;
return 0;
}
void BigInteger::multByDigitWithoutSign(BigInteger& c, const BigInteger& a, unsigned short b) {
unsigned int mult_n_carry = 0;
c.digits.clear();
c.digits.resize(a.digits.size());
for (int i = 0; i < (int)a.digits.size(); ++i) {
unsigned short a_digit = 0;
unsigned short b_digit = b;
if (i < (int)a.digits.size())
a_digit = a.digits[i];
mult_n_carry += a_digit * b_digit;
c.digits[i] = (mult_n_carry & 0xFFFF);
mult_n_carry >>= 16;
}
if (mult_n_carry != 0) {
putCarryInfront(c, mult_n_carry);
}
//std::cout << a.toString() << " x " << b << " == " << c.toString() << std::endl;
}
void BigInteger::shiftLeftByBase(BigInteger& b, const BigInteger& a, int times) {
b.digits.resize(a.digits.size() + times);
for (int i = 0; i < times; ++i) {
b.digits[i] = 0;
}
for (int i = 0; i < (int)a.digits.size(); ++i) {
b.digits[i + times] = a.digits[i];
}
}
void BigInteger::shiftRight(BigInteger& a) {
//std::cout << "shr " << a.toString() << " == ";
for (int i = 0; i < (int)a.digits.size(); ++i) {
a.digits[i] >>= 1;
if (i+1 < (int)a.digits.size()) {
if ((a.digits[i+1] & 0x1) != 0) {
a.digits[i] |= 0x8000;
}
}
}
//std::cout << a.toString() << std::endl;
}
void BigInteger::shiftLeft(BigInteger& a) {
bool lastBit = false;
for (int i = 0; i < (int)a.digits.size(); ++i) {
bool bit = (a.digits[i] & 0x8000) != 0;
a.digits[i] <<= 1;
if (lastBit)
a.digits[i] |= 1;
lastBit = bit;
}
if (lastBit) {
a.digits.push_back(1);
}
}
void BigInteger::putCarryInfront(BigInteger& a, unsigned short carry) {
BigInteger b;
b.negative = a.negative;
b.digits.resize(a.digits.size() + 1);
b.digits[a.digits.size()] = carry;
for (int i = 0; i < (int)a.digits.size(); ++i) {
b.digits[i] = a.digits[i];
}
a.digits.swap(b.digits);
}
void BigInteger::divideWithoutSign(BigInteger& c, BigInteger& d, const BigInteger& a, const BigInteger& b) {
c.digits.clear();
c.digits.push_back(0);
BigInteger two("2");
BigInteger e = b;
BigInteger f("1");
BigInteger g = a;
BigInteger one("1");
while (cmpWithoutSign(g, e) >= 0) {
shiftLeft(e);
shiftLeft(f);
}
shiftRight(e);
shiftRight(f);
while (cmpWithoutSign(g, b) >= 0) {
g -= e;
c += f;
while (cmpWithoutSign(g, e) < 0) {
shiftRight(e);
shiftRight(f);
}
}
e = c;
e *= b;
f = a;
f -= e;
d = f;
}
BigInteger::BigInteger(const BigInteger& other) {
digits = other.digits;
negative = other.negative;
}
BigInteger::BigInteger(const char* other) {
digits.push_back(0);
negative = false;
BigInteger ten;
ten.digits[0] = 10;
const char* c = other;
bool make_negative = false;
if (*c == '-') {
make_negative = true;
++c;
}
while (*c != 0) {
BigInteger digit;
digit.digits[0] = *c - '0';
*this *= ten;
*this += digit;
++c;
}
negative = make_negative;
}
bool BigInteger::isOdd() const {
return (digits[0] & 0x1) != 0;
}
BigInteger& BigInteger::operator=(const BigInteger& other) {
if (this == &other) // handle self assignment
return *this;
digits = other.digits;
negative = other.negative;
return *this;
}
BigInteger& BigInteger::operator+=(const BigInteger& other) {
BigInteger result;
if (negative) {
if (other.negative) {
result.negative = true;
addWithoutSign(result, *this, other);
} else {
int a = cmpWithoutSign(*this, other);
if (a < 0) {
result.negative = false;
subWithoutSign(result, other, *this);
} else if (a > 0) {
result.negative = true;
subWithoutSign(result, *this, other);
} else {
result.negative = false;
result.digits.clear();
result.digits.push_back(0);
}
}
} else {
if (other.negative) {
int a = cmpWithoutSign(*this, other);
if (a < 0) {
result.negative = true;
subWithoutSign(result, other, *this);
} else if (a > 0) {
result.negative = false;
subWithoutSign(result, *this, other);
} else {
result.negative = false;
result.digits.clear();
result.digits.push_back(0);
}
} else {
result.negative = false;
addWithoutSign(result, *this, other);
}
}
negative = result.negative;
digits.swap(result.digits);
return *this;
}
BigInteger& BigInteger::operator-=(const BigInteger& other) {
BigInteger neg_other = other;
neg_other.negative = !neg_other.negative;
return *this += neg_other;
}
BigInteger& BigInteger::operator*=(const BigInteger& other) {
BigInteger result;
for (int i = 0; i < (int)digits.size(); ++i) {
BigInteger mult;
multByDigitWithoutSign(mult, other, digits[i]);
BigInteger shift;
shiftLeftByBase(shift, mult, i);
BigInteger add;
addWithoutSign(add, result, shift);
result = add;
}
if (negative != other.negative) {
result.negative = true;
} else {
result.negative = false;
}
//std::cout << toString() << " x " << other.toString() << " == " << result.toString() << std::endl;
negative = result.negative;
digits.swap(result.digits);
return *this;
}
BigInteger& BigInteger::operator/=(const BigInteger& other) {
BigInteger result, tmp;
divideWithoutSign(result, tmp, *this, other);
result.negative = (negative != other.negative);
negative = result.negative;
digits.swap(result.digits);
return *this;
}
BigInteger& BigInteger::operator%=(const BigInteger& other) {
BigInteger c, d;
divideWithoutSign(c, d, *this, other);
*this = d;
return *this;
}
bool BigInteger::operator>(const BigInteger& other) const {
if (negative) {
if (other.negative) {
return cmpWithoutSign(*this, other) < 0;
} else {
return false;
}
} else {
if (other.negative) {
return true;
} else {
return cmpWithoutSign(*this, other) > 0;
}
}
}
BigInteger& BigInteger::powAssignUnderMod(const BigInteger& exponent, const BigInteger& modulus) {
BigInteger zero("0");
BigInteger one("1");
BigInteger e = exponent;
BigInteger base = *this;
*this = one;
while (cmpWithoutSign(e, zero) != 0) {
//std::cout << e.toString() << " : " << toString() << " : " << base.toString() << std::endl;
if (e.isOdd()) {
*this *= base;
*this %= modulus;
}
shiftRight(e);
base *= BigInteger(base);
base %= modulus;
}
return *this;
}
std::string BigInteger::toString() const {
std::ostringstream os;
if (negative)
os << "-";
BigInteger tmp = *this;
BigInteger zero("0");
BigInteger ten("10");
tmp.negative = false;
std::stack<char> s;
while (cmpWithoutSign(tmp, zero) != 0) {
BigInteger tmp2, tmp3;
divideWithoutSign(tmp2, tmp3, tmp, ten);
s.push((char)(tmp3.digits[0] + '0'));
tmp = tmp2;
}
while (!s.empty()) {
os << s.top();
s.pop();
}
/*
for (int i = digits.size()-1; i >= 0; --i) {
os << digits[i];
if (i != 0) {
os << ",";
}
}
*/
return os.str();
And an example usage.
BigInteger a("87682374682734687"), b("435983748957348957349857345"), c("2348927349872344")
// Will Calculate pow(87682374682734687, 435983748957348957349857345) % 2348927349872344
a.powAssignUnderMod(b, c);
Its fast too, and has unlimited number of digits.
Two things:
Are you using the appropriate data type? In other words, does UINT_MAX allow you to have 673109 as an argument?
No, it does not, since at one point you have Your code does not work because at one point you have num = 2^16 and the num = ... causes overflow. Use a bigger data type to hold this intermediate value.
How about taking modulo at every possible overflow oppertunity such as:
test = ((test % mod) * (num % mod)) % mod;
Edit:
unsigned mod_pow(unsigned num, unsigned pow, unsigned mod)
{
unsigned long long test;
unsigned long long n = num;
for(test = 1; pow; pow >>= 1)
{
if (pow & 1)
test = ((test % mod) * (n % mod)) % mod;
n = ((n % mod) * (n % mod)) % mod;
}
return test; /* note this is potentially lossy */
}
int main(int argc, char* argv[])
{
/* (2 ^ 168277) % 673109 */
printf("%u\n", mod_pow(2, 168277, 673109));
return 0;
}
package playTime;
public class play {
public static long count = 0;
public static long binSlots = 10;
public static long y = 645;
public static long finalValue = 1;
public static long x = 11;
public static void main(String[] args){
int[] binArray = new int[]{0,0,1,0,0,0,0,1,0,1};
x = BME(x, count, binArray);
System.out.print("\nfinal value:"+finalValue);
}
public static long BME(long x, long count, int[] binArray){
if(count == binSlots){
return finalValue;
}
if(binArray[(int) count] == 1){
finalValue = finalValue*x%y;
}
x = (x*x)%y;
System.out.print("Array("+binArray[(int) count]+") "
+"x("+x+")" +" finalVal("+ finalValue + ")\n");
count++;
return BME(x, count,binArray);
}
}
LL is for long long int
LL power_mod(LL a, LL k) {
if (k == 0)
return 1;
LL temp = power(a, k/2);
LL res;
res = ( ( temp % P ) * (temp % P) ) % P;
if (k % 2 == 1)
res = ((a % P) * (res % P)) % P;
return res;
}
Use the above recursive function for finding the mod exp of the number. This will not result in overflow because it calculates in a bottom up manner.
Sample test run for :
a = 2 and k = 168277 shows output to be 518358 which is correct and the function runs in O(log(k)) time;
You could use following identity:
(a * b) (mod m) === (a (mod m)) * (b (mod m)) (mod m)
Try using it straightforward way and incrementally improve.
if (pow & 1)
test = ((test % mod) * (num % mod)) % mod;
num = ((num % mod) * (num % mod)) % mod;