We Keep Coding

Find the minimal absolute value of a sum of two elements

I am solving the Codility problem provided below,
Let A be a non-empty array consisting of N integers.
The abs sum of two for a pair of indices (P, Q) is the absolute value |A[P] + A[Q]|, for 0 ≤ P ≤ Q < N.
For example, the following array A:
A[0] = 1 A1 = 4 A[2] = -3 has pairs of indices (0, 0), (0,
1), (0, 2), (1, 1), (1, 2), (2, 2). The abs sum of two for the pair
(0, 0) is A[0] + A[0] = |1 + 1| = 2. The abs sum of two for the pair
(0, 1) is A[0] + A1 = |1 + 4| = 5. The abs sum of two for the pair
(0, 2) is A[0] + A[2] = |1 + (−3)| = 2. The abs sum of two for the
pair (1, 1) is A1 + A1 = |4 + 4| = 8. The abs sum of two for the
pair (1, 2) is A1 + A[2] = |4 + (−3)| = 1. The abs sum of two for
the pair (2, 2) is A[2] + A[2] = |(−3) + (−3)| = 6.`
Write a function:
class Solution { public int solution(int[] A); }
that, given a non-empty array A consisting of N integers, returns the minimal abs sum of two for any pair of indices in this array.
For example, given the following array A:
A[0] = 1 A1 = 4 A[2] = -3 the function should return 1, as
explained above.
Given array A:
A[0] = -8 A1 = 4 A[2] = 5 A[3] =-10 A[4] = 3 the function
should return |(−8) + 5| = 3.
Write an efficient algorithm for the following assumptions:
N is an integer within the range [1..100,000]; each element of array A is an integer within the range [−1,000,000,000..1,000,000,000].
I write the solution provided below.
public static int solution(int[] A) {
int N = A.length;
Arrays.sort(A);
if (A[0] >= 0) {
return 2 * A[0];
}
int i = 0;
int j = N - 1;
int sum = Math.abs((A[i] + A[j]));
// -10, -8, 3, 4, 5
while (i <= j) {
if (Math.abs(A[i + 1] + A[j]) < sum) {
++i;
sum = Math.abs(A[i] + A[j]);
} else if (Math.abs(A[i] + A[j - 1]) < sum) {
--j;
sum = Math.abs(A[i] + A[j]);
} else {
i++;
j--;
}
}
return sum;
}
The solution gets hanged in the online judge and seems it enters in a forever loop. Is there a possibility that the code can enter a non-ending loop?
UPDATE
After I update the solution with the all negatives check, the code passed the online judge and provides a good performance.
if(A[N-1] <=0){
return 2* Math.abs(A[N-1]);
}

For input arrays e.g({-1, -2, -3}, {-1, -2}, {-1} your algorithm throws ArrayIndexOutOfBoundsException, so arrays when there are only negative numbers and there is no repeats
There is no chance to reach endless loop because either i or j change only + or - 1

I would like to explain the algorithm that I have implemented and then the implementation in C++.
Sort the array because otherwise we would need to check any two arbitrary indices. That brute-force solution would result in O(N ^ 2) runtime complexity.
We initialise the min abs sum to something higher than the possible value in the arrays.
Apply the caterpillar method by having front and back indices.
In every iteration, update the min abs sum as and when needed.
There two special cases:
a. all values are zero or positive. We could return A[front] * 2 early.
b. all values are negative or zero. We could return A[back] * 2 early.
In both cases, we could return early, however, it would result in a bit more code, so I personally avoid that. In the above cases, we can still go through the array without degrading the overall runtime complexity. In these cases, it also does not matter how we go through the array with regards to the result, so I just go through the array in one way in both cases. But the code will be shared with the third case.
In the third case, where the array, and thereby the sorted array, contains both negative and positive values, we try to keep the sum of front and back to zero since this when the abs sum value is minimised. In other words, we try to keep the balance between the negative and positive numbers by keeping their distance to the minimum.
Therefore, if the sum of front and back are less than zero, then we know that the negative front value is greater than the positive back value. As a direct consequence of that, we need to advance the front index.
If the sum of front and back are equal to zero, then we found the smallest min abs sum that is ever possible. The absolute value cannot be less than zero. Whilst I could return at this point in the code for some constant optimisation, I do not do so to keep the code the minimal and also functional.
If the sum of front and back are greater than zero, then we know that the negative front value is less than the positive back value. As a direct consequence of that, we need to decrease the index.
We loop until the front and back indices meet, but we handle the case when they are equal since according to the task specification, the same index can be used twice for the absolute sum.
The runtime complexity of this solution is O(N * logN) because the sorting of the array is O(N * logN) and the loop is O(N) since we go through every element. Therefore, the sorting runtime complexity dominates the loop.
The space complexity is O(1) because we only allocate constant space independently from the number of inputs. The sorting is done in place.
int solution(vector<int> &A)
{
const int N = A.size();
sort(A.begin(), A.end());
int min_abs_sum_of_two = INT_MAX;
for (int front = 0, back = N - 1; front <= back;) {
const int ef = A[front];
const int eb = A[back];
min_abs_sum_of_two = min(min_abs_sum_of_two, abs(ef + eb));
if (ef + eb < 0) ++front; else --back;
}
return min_abs_sum_of_two;
}

I got 100% for following code ( java). Slightly modified version of https://www.techiedelight.com/find-pair-array-minimum-absolute-sum/
import java.util.*;
// you can write to stdout for debugging purposes, e.g.
// System.out.println("this is a debug message");
class Solution {
public int solution(int[] A) {
// sort the array if it is unsorted
Arrays.sort(A);
int low = 0;
int high = A.length - 1;
if (A[0] >= 0) {
return 2 * A[0];
}
if (A[high] <= 0) {
return -2 * A[high];
}
// maintain two indexes pointing to endpoints of the array
// `min` stores the minimum absolute difference
int min = Integer.MAX_VALUE;
int i = 0, j = 0;
// reduce the search space `A[low…high]` at each iteration of the loop
int sum = 0;
// loop if `low` is less than `high`
while (low < high)
{
// update the minimum if the current absolute sum is less.
sum = A[high] + A[low];
if (Math.abs(sum) < min)
{
min = Math.abs(sum);
i = low;
j = high;
}
// optimization: pair with zero-sum is found
if (min == 0) {
break;
}
// increment `low` index if the total is less than 0;
// decrement `high` index if the total is more than 0
if (sum < 0) {
low++;
}
else {
high--;
}
}
return min;
}
}

The Answer is late but I hope to help anyone have the same question
//============================================================================
// Author: Hamdy Abd El Fattah
// Code is like humor. When you have to explain it, it’s bad.
//============================================================================
#include <bits/stdc++.h>
#define FIO cin.tie(0), cin.sync_with_stdio(0)
#define ll long long
#define INF 0x3f3f3f3f
using namespace std;
int main() {
FIO;
ll n , m = INF , x;
vector<ll> positive,negative;
cin>>n;
while(n--){
cin>>x;
if(x < 0)
negative.push_back(x * -1);
else if(x > 0)
positive.push_back(x);
else
m = 0;
}
if(m == 0){
cout<<"0\n";
}else{
sort(positive.begin(), positive.end());
sort(negative.begin(), negative.end());
int i= 0, j = 0;
int positive_size = positive.size(), negative_size =negative.size();
while(i < positive_size || j < negative_size){
if(abs(positive[i] - negative[j]) < m){
m=abs(positive[i] - negative[j]);
m=min(min(m,positive[i]*2),negative[j] * 2);
}
if((i < positive_size && positive[i] <= negative[j]) || j == negative_size)
i++;
else if((j < negative_size && positive[i] > negative[j]) || i == positive_size)
j++;
}
cout<<m<<endl;
}
return 0;
}

Looking through different combinations through matrix using just visited int variable?

I am looking at this topcoder problem here:
http://community.topcoder.com/tc?module=ProblemDetail&rd=4725&pm=2288
Under the java section there is this code :
public class KiloManX {
boolean ddd = false;
int[] s2ia(String s) {
int[] r = new int[s.length()];
for (int i = 0; i < s.length(); i++) {
r[i] = s.charAt(i) - '0' ;
}
return r;
}
public int leastShots(String[] damageChart, int[] bossHealth) {
int i, j, k;
int n = damageChart.length;
int[][] dc = new int[n][];
int[] cost = new int[1 << n];
for (i = 0; i < n; i++) {
dc[i] = s2ia(damageChart[i]) ;
}
for (i = 1; i < 1 << n; i++) {
cost[i] = 65536 * 30000;
for (j = 0; j < n; j++) {
int pre = i - (1 << j);
if ((i & (1 << j)) != 0) {
cost[i] = Math.min(cost[i], cost[pre] + bossHealth[j]) ;
for (k = 0; k < n; k++) {
if ((i & (1 << k)) != 0 && k != j && dc[k][j] > 0) {
cost[i] = Math.min(cost[i],
cost[pre] + (bossHealth[j] + dc[k][j] - 1) / dc[k][j]);
}
}
}
}
}
return cost[(1 << n) - 1] ;
}
static void pp(Object o) {
System.out.println(o);
}
}
I am trying to understand what he is been done. So what I understand is :
i - keeps track of the visited nodes somehow(this is the most baffling part of the code)
j - is the monster we want to defeat
k - is the previous monster's weapon we are using to defeat j
dc is the input array of string into a matrix
cost, keep cost at each step, some sort of dynamic programming? I don't understand how cost[1 << n] can give the result?
What I understand is they are going through all the possible sets / combinations. What I am baffled by (even after executing and starring at this for more than a week) is:
how do they keep track of all the combinations?
I understand pre - is the cost of the previous monster defeated (i.e. how much cost we incurred there), but I don't understand how you get it from just (i - 1 << j).
I have executed the program(debugger), stared at it for more than a week, and tried to decode it, but I am baffled by the bit-manipulation part of the code. Can someone please shed light on this?

cost, keep cost at each step, some sort of dynamic programming?
They are partial costs, yes, but characterizing them as per-step costs misses the most important significance of the indices into this array. More below.
I don't understand how cost[1 << n] can give the result?
That doesn't give any result by itself, of course. It just declares an array with 2n elements.
how do they keep track of all the combinations?
See below. This is closely related to why the cost array is declared the size it is.
I understand pre - is the cost of the previous monster defeated (i.e. how much cost we incurred there), but I don't understand how you get it from just (i - 1 << j).
Surely pre is not itself a cost. It is, however, used conditionally as an index into the cost array. Now consider the condition:
if ((i & (1 << j)) != 0) {
The expression i & (1 << j) tests whether bit j of the value of i is set. When it is, i - (1 << j) (i.e. pre) evaluates to the the result of turning off bit j of the value of i. That should clue you in that the indices of cost are bit masks. The size of that array (1 << n) is another clue: it is the number of distinct n-bit bitmasks.
The trick here is a relatively common one, and a good one to know. Suppose you have a set of N objects, and you want somehow to represent all of its subsets (== all the distinct combinations of its elements). Each subset is characterized by whether each of the N objects is an element or not. You can represent that as N numbers, each either 0 or 1 -- i.e. N bits. Now suppose you string those bits together into N-bit numbers. Every integer from 0 (inclusive) to 2N (exclusive) has a distinct pattern of its least-significant N bits, so each corresponds to different subset.
The code presented uses exactly this sort of correspondence to encode the different subsets of the set of bosses as different indices into the cost array -- which answers your other question of how it keeps track of combinations. Given one such index i that represents a subset containing boss j, the index i - (1 << j) represents the set obtained from it by removing boss j.
Roughly speaking, then, the program proceeds by optimizing the cost of each non-empty subset by checking all the ways to form it from a subset with one element fewer. (1 << n) - 1 is the index corresponding to the whole set, so at the end, that element of cost contains the overall optimized value.

Longest path in unweighted undirected graph

Having this graph as reference let's say i want the longest path between 0 and 5.
That would be: 0->1->3->2->4->6->5
Is there any good algorithm for this? I've searched and haven't found anything that i was able to understand.
I've found plenty algorithms for the shortest path (0->1->2->4->6->5) and i've implemented them successfully.
Maybe i'm the problem, but i would like to think otherwise :)
Any help would be welcome

This problem is NP-Hard (there is a simple reduction from a Hamiltonian path to your problem, and a Hamiltonian path search is known to be NP-hard). It means that there is no polynomial solution for this problem (unless P = NP).
If you need an exact solution, you can use dynamic programming (with exponential number of states): the state is (mask of visited vertices, last_vertex), the value is true or false. A transition is adding a new vertex which is not in the mask if there an edge between the last_vertex and the new vertex. It has O(2^n * n^2) time complexity, which is still better than O(n!) backtracking.
Here is pseudo code of a dynamic programming solution:
f = array of (2 ^ n) * n size filled with false values
f(1 << start, start) = true
for mask = 0 ... (1 << n) - 1:
for last = 0 ... n - 1:
for new = 0 ... n - 1:
if there is an edge between last and new and mask & (1 << new) == 0:
f(mask | (1 << new), new) |= f(mask, last)
res = 0
for mask = 0 ... (1 << n) - 1:
if f(mask, end):
res = max(res, countBits(mask))
return res
And a little bit more about reduction from Hamiltonian path to this problem:
def hamiltonianPathExists():
found = false
for i = 0 ... n - 1:
for j = 0 ... n - 1:
if i != j:
path = getLongestPath(i, j) // calls a function that solves this problem
if length(path) == n:
found = true
return found
Here is a Java implementation (I did not test properly, so it can contain bugs):
/**
* Finds the longest path between two specified vertices in a specified graph.
* #param from The start vertex.
* #param to The end vertex.
* #param graph The graph represented as an adjacency matrix.
* #return The length of the longest path between from and to.
*/
public int getLongestPath(int from, int to, boolean[][] graph) {
int n = graph.length;
boolean[][] hasPath = new boolean[1 << n][n];
hasPath[1 << from][from] = true;
for (int mask = 0; mask < (1 << n); mask++)
for (int last = 0; last < n; last++)
for (int curr = 0; curr < n; curr++)
if (graph[last][curr] && (mask & (1 << curr)) == 0)
hasPath[mask | (1 << curr)][curr] |= hasPath[mask][last];
int result = 0;
for (int mask = 0; mask < (1 << n); mask++)
if (hasPath[mask][to])
result = Math.max(result, Integer.bitCount(mask));
return result;
}

represent an integer using binary in java language

Here is the problem:
You're given 2 32-bit numbers, N & M, and two bit positions, i & j. write a method to set all bits between i and j in N equal to M (e.g. M becomes a substring of N at locating at i
and starting at j)
For example:
input:
int N = 10000000000, M = 10101, i = 2, j = 6;
output:
int N = 10001010100
My solution:
step 1: compose one mask to clear sets from i to j in N
mask= ( ( ( ((1<<(31-j))-1) << (j-i+1) ) + 1 ) << i ) - 1
for the example, we have
mask= 11...10000011
step 2:
(N & mask) | (M<<i)
Question:
what is the convenient data type to implement the algorithm? for example
we have int n = 0x100000 in C, so that we can apply bitwise operators on n.
in Java, we have BitSet class, it has clear, set method, but doesnt support
left/right shift operator; if we use int, it supports left/right shift, but
doesnt have binary representation (I am not talking about binary string representation)
what is the best way to implement this?
Code in java (after reading all comments):
int x = Integer.parseInt("10000000000",2);
int x = Integer.parseInt("10101",2);
int i = 2, j = 6;
public static int F(int x, int y, int i, int j){
int mask = (-1<<(j+1)) | (-1>>>(32-i));
return (mask & x ) | (y<<i);
}

the bit-wise operators |, &, ^ and ~ and the hex literal (0x1010) are all available in java
32 bit numbers are ints if that constraint remains int will be a valid data type
btw
mask = (-1<<j)|(-1>>>(32-i));
is a slightly clearer construction of the mask

Java's int has all the operations you need. I did not totally understand your question (too tired now), so I'll not give you a complete answer, just some hints. (I'll revise it later, if needed.)
Here are j ones in a row: (1 << j)-1.
Here are j ones in a row, followed by i zeros: ((1 << j) - 1) << i.
Here is a bitmask which masks out j positions in the middle of x: x & ~(((1 << j) - 1) << i).
Try these with Integer.toBinaryString() to see the results. (They might also give strange results for negative or too big values.)

I think you're misunderstanding how Java works. All values are represented as 'a series of bits' under the hood, ints and longs are included in that.
Based on your question, a rough solution is:
public static int applyBits(int N, int M, int i, int j) {
M = M << i; // Will truncate left-most bits if too big
// Assuming j > i
for(int loopVar = i; loopVar < j; loopVar++) {
int bitToApply = 1 << loopVar;
// Set the bit in N to 0
N = N & ~bitToApply;
// Apply the bit if M has it set.
N = (M & bitToApply) | N;
}
return N;
}
My assumptions are:
i is the right-most (least-significant) bit that is being set in N.
M's right-most bit maps to N's ith bit from the right.
That premature optimization is the root of all evil - this is O(j-i). If you used a complicated mask like you did in the question you can do it in O(1), but it won't be as readable, and readable code is 97% of the time more important than efficient code.

Fastest way to determine if an integer's square root is an integer

I'm looking for the fastest way to determine if a long value is a perfect square (i.e. its square root is another integer):
I've done it the easy way, by using the built-in Math.sqrt()
function, but I'm wondering if there is a way to do it faster by
restricting yourself to integer-only domain.
Maintaining a lookup table is impractical (since there are about
231.5 integers whose square is less than 263).
Here is the very simple and straightforward way I'm doing it now:
public final static boolean isPerfectSquare(long n)
{
if (n < 0)
return false;
long tst = (long)(Math.sqrt(n) + 0.5);
return tst*tst == n;
}
Note: I'm using this function in many Project Euler problems. So no one else will ever have to maintain this code. And this kind of micro-optimization could actually make a difference, since part of the challenge is to do every algorithm in less than a minute, and this function will need to be called millions of times in some problems.
I've tried the different solutions to the problem:
After exhaustive testing, I found that adding 0.5 to the result of Math.sqrt() is not necessary, at least not on my machine.
The fast inverse square root was faster, but it gave incorrect results for n >= 410881. However, as suggested by BobbyShaftoe, we can use the FISR hack for n < 410881.
Newton's method was a good bit slower than Math.sqrt(). This is probably because Math.sqrt() uses something similar to Newton's Method, but implemented in the hardware so it's much faster than in Java. Also, Newton's Method still required use of doubles.
A modified Newton's method, which used a few tricks so that only integer math was involved, required some hacks to avoid overflow (I want this function to work with all positive 64-bit signed integers), and it was still slower than Math.sqrt().
Binary chop was even slower. This makes sense because the binary chop will on average require 16 passes to find the square root of a 64-bit number.
According to John's tests, using or statements is faster in C++ than using a switch, but in Java and C# there appears to be no difference between or and switch.
I also tried making a lookup table (as a private static array of 64 boolean values). Then instead of either switch or or statement, I would just say if(lookup[(int)(n&0x3F)]) { test } else return false;. To my surprise, this was (just slightly) slower. This is because array bounds are checked in Java.

I figured out a method that works ~35% faster than your 6bits+Carmack+sqrt code, at least with my CPU (x86) and programming language (C/C++). Your results may vary, especially because I don't know how the Java factor will play out.
My approach is threefold:
First, filter out obvious answers. This includes negative numbers and looking at the last 4 bits. (I found looking at the last six didn't help.) I also answer yes for 0. (In reading the code below, note that my input is int64 x.)
if( x < 0 || (x&2) || ((x & 7) == 5) || ((x & 11) == 8) )
return false;
if( x == 0 )
return true;
Next, check if it's a square modulo 255 = 3 * 5 * 17. Because that's a product of three distinct primes, only about 1/8 of the residues mod 255 are squares. However, in my experience, calling the modulo operator (%) costs more than the benefit one gets, so I use bit tricks involving 255 = 2^8-1 to compute the residue. (For better or worse, I am not using the trick of reading individual bytes out of a word, only bitwise-and and shifts.)
int64 y = x;
y = (y & 4294967295LL) + (y >> 32);
y = (y & 65535) + (y >> 16);
y = (y & 255) + ((y >> 8) & 255) + (y >> 16);
// At this point, y is between 0 and 511. More code can reduce it farther.
To actually check if the residue is a square, I look up the answer in a precomputed table.
if( bad255[y] )
return false;
// However, I just use a table of size 512
Finally, try to compute the square root using a method similar to Hensel's lemma. (I don't think it's applicable directly, but it works with some modifications.) Before doing that, I divide out all powers of 2 with a binary search:
if((x & 4294967295LL) == 0)
x >>= 32;
if((x & 65535) == 0)
x >>= 16;
if((x & 255) == 0)
x >>= 8;
if((x & 15) == 0)
x >>= 4;
if((x & 3) == 0)
x >>= 2;
At this point, for our number to be a square, it must be 1 mod 8.
if((x & 7) != 1)
return false;
The basic structure of Hensel's lemma is the following. (Note: untested code; if it doesn't work, try t=2 or 8.)
int64 t = 4, r = 1;
t <<= 1; r += ((x - r * r) & t) >> 1;
t <<= 1; r += ((x - r * r) & t) >> 1;
t <<= 1; r += ((x - r * r) & t) >> 1;
// Repeat until t is 2^33 or so. Use a loop if you want.
The idea is that at each iteration, you add one bit onto r, the "current" square root of x; each square root is accurate modulo a larger and larger power of 2, namely t/2. At the end, r and t/2-r will be square roots of x modulo t/2. (Note that if r is a square root of x, then so is -r. This is true even modulo numbers, but beware, modulo some numbers, things can have even more than 2 square roots; notably, this includes powers of 2.) Because our actual square root is less than 2^32, at that point we can actually just check if r or t/2-r are real square roots. In my actual code, I use the following modified loop:
int64 r, t, z;
r = start[(x >> 3) & 1023];
do {
z = x - r * r;
if( z == 0 )
return true;
if( z < 0 )
return false;
t = z & (-z);
r += (z & t) >> 1;
if( r > (t >> 1) )
r = t - r;
} while( t <= (1LL << 33) );
The speedup here is obtained in three ways: precomputed start value (equivalent to ~10 iterations of the loop), earlier exit of the loop, and skipping some t values. For the last part, I look at z = r - x * x, and set t to be the largest power of 2 dividing z with a bit trick. This allows me to skip t values that wouldn't have affected the value of r anyway. The precomputed start value in my case picks out the "smallest positive" square root modulo 8192.
Even if this code doesn't work faster for you, I hope you enjoy some of the ideas it contains. Complete, tested code follows, including the precomputed tables.
typedef signed long long int int64;
int start[1024] =
{1,3,1769,5,1937,1741,7,1451,479,157,9,91,945,659,1817,11,
1983,707,1321,1211,1071,13,1479,405,415,1501,1609,741,15,339,1703,203,
129,1411,873,1669,17,1715,1145,1835,351,1251,887,1573,975,19,1127,395,
1855,1981,425,453,1105,653,327,21,287,93,713,1691,1935,301,551,587,
257,1277,23,763,1903,1075,1799,1877,223,1437,1783,859,1201,621,25,779,
1727,573,471,1979,815,1293,825,363,159,1315,183,27,241,941,601,971,
385,131,919,901,273,435,647,1493,95,29,1417,805,719,1261,1177,1163,
1599,835,1367,315,1361,1933,1977,747,31,1373,1079,1637,1679,1581,1753,1355,
513,1539,1815,1531,1647,205,505,1109,33,1379,521,1627,1457,1901,1767,1547,
1471,1853,1833,1349,559,1523,967,1131,97,35,1975,795,497,1875,1191,1739,
641,1149,1385,133,529,845,1657,725,161,1309,375,37,463,1555,615,1931,
1343,445,937,1083,1617,883,185,1515,225,1443,1225,869,1423,1235,39,1973,
769,259,489,1797,1391,1485,1287,341,289,99,1271,1701,1713,915,537,1781,
1215,963,41,581,303,243,1337,1899,353,1245,329,1563,753,595,1113,1589,
897,1667,407,635,785,1971,135,43,417,1507,1929,731,207,275,1689,1397,
1087,1725,855,1851,1873,397,1607,1813,481,163,567,101,1167,45,1831,1205,
1025,1021,1303,1029,1135,1331,1017,427,545,1181,1033,933,1969,365,1255,1013,
959,317,1751,187,47,1037,455,1429,609,1571,1463,1765,1009,685,679,821,
1153,387,1897,1403,1041,691,1927,811,673,227,137,1499,49,1005,103,629,
831,1091,1449,1477,1967,1677,697,1045,737,1117,1737,667,911,1325,473,437,
1281,1795,1001,261,879,51,775,1195,801,1635,759,165,1871,1645,1049,245,
703,1597,553,955,209,1779,1849,661,865,291,841,997,1265,1965,1625,53,
1409,893,105,1925,1297,589,377,1579,929,1053,1655,1829,305,1811,1895,139,
575,189,343,709,1711,1139,1095,277,993,1699,55,1435,655,1491,1319,331,
1537,515,791,507,623,1229,1529,1963,1057,355,1545,603,1615,1171,743,523,
447,1219,1239,1723,465,499,57,107,1121,989,951,229,1521,851,167,715,
1665,1923,1687,1157,1553,1869,1415,1749,1185,1763,649,1061,561,531,409,907,
319,1469,1961,59,1455,141,1209,491,1249,419,1847,1893,399,211,985,1099,
1793,765,1513,1275,367,1587,263,1365,1313,925,247,1371,1359,109,1561,1291,
191,61,1065,1605,721,781,1735,875,1377,1827,1353,539,1777,429,1959,1483,
1921,643,617,389,1809,947,889,981,1441,483,1143,293,817,749,1383,1675,
63,1347,169,827,1199,1421,583,1259,1505,861,457,1125,143,1069,807,1867,
2047,2045,279,2043,111,307,2041,597,1569,1891,2039,1957,1103,1389,231,2037,
65,1341,727,837,977,2035,569,1643,1633,547,439,1307,2033,1709,345,1845,
1919,637,1175,379,2031,333,903,213,1697,797,1161,475,1073,2029,921,1653,
193,67,1623,1595,943,1395,1721,2027,1761,1955,1335,357,113,1747,1497,1461,
1791,771,2025,1285,145,973,249,171,1825,611,265,1189,847,1427,2023,1269,
321,1475,1577,69,1233,755,1223,1685,1889,733,1865,2021,1807,1107,1447,1077,
1663,1917,1129,1147,1775,1613,1401,555,1953,2019,631,1243,1329,787,871,885,
449,1213,681,1733,687,115,71,1301,2017,675,969,411,369,467,295,693,
1535,509,233,517,401,1843,1543,939,2015,669,1527,421,591,147,281,501,
577,195,215,699,1489,525,1081,917,1951,2013,73,1253,1551,173,857,309,
1407,899,663,1915,1519,1203,391,1323,1887,739,1673,2011,1585,493,1433,117,
705,1603,1111,965,431,1165,1863,533,1823,605,823,1179,625,813,2009,75,
1279,1789,1559,251,657,563,761,1707,1759,1949,777,347,335,1133,1511,267,
833,1085,2007,1467,1745,1805,711,149,1695,803,1719,485,1295,1453,935,459,
1151,381,1641,1413,1263,77,1913,2005,1631,541,119,1317,1841,1773,359,651,
961,323,1193,197,175,1651,441,235,1567,1885,1481,1947,881,2003,217,843,
1023,1027,745,1019,913,717,1031,1621,1503,867,1015,1115,79,1683,793,1035,
1089,1731,297,1861,2001,1011,1593,619,1439,477,585,283,1039,1363,1369,1227,
895,1661,151,645,1007,1357,121,1237,1375,1821,1911,549,1999,1043,1945,1419,
1217,957,599,571,81,371,1351,1003,1311,931,311,1381,1137,723,1575,1611,
767,253,1047,1787,1169,1997,1273,853,1247,413,1289,1883,177,403,999,1803,
1345,451,1495,1093,1839,269,199,1387,1183,1757,1207,1051,783,83,423,1995,
639,1155,1943,123,751,1459,1671,469,1119,995,393,219,1743,237,153,1909,
1473,1859,1705,1339,337,909,953,1771,1055,349,1993,613,1393,557,729,1717,
511,1533,1257,1541,1425,819,519,85,991,1693,503,1445,433,877,1305,1525,
1601,829,809,325,1583,1549,1991,1941,927,1059,1097,1819,527,1197,1881,1333,
383,125,361,891,495,179,633,299,863,285,1399,987,1487,1517,1639,1141,
1729,579,87,1989,593,1907,839,1557,799,1629,201,155,1649,1837,1063,949,
255,1283,535,773,1681,461,1785,683,735,1123,1801,677,689,1939,487,757,
1857,1987,983,443,1327,1267,313,1173,671,221,695,1509,271,1619,89,565,
127,1405,1431,1659,239,1101,1159,1067,607,1565,905,1755,1231,1299,665,373,
1985,701,1879,1221,849,627,1465,789,543,1187,1591,923,1905,979,1241,181};
bool bad255[512] =
{0,0,1,1,0,1,1,1,1,0,1,1,1,1,1,0,0,1,1,0,1,0,1,1,1,0,1,1,1,1,0,1,
1,1,0,1,0,1,1,1,1,1,1,1,1,1,1,1,1,0,1,0,1,1,1,0,1,1,1,1,0,1,1,1,
0,1,0,1,1,0,0,1,1,1,1,1,0,1,1,1,1,0,1,1,0,0,1,1,1,1,1,1,1,1,0,1,
1,1,1,1,0,1,1,1,1,1,0,1,1,1,1,0,1,1,1,0,1,1,1,1,0,0,1,1,1,1,1,1,
1,1,1,1,1,1,1,0,0,1,1,1,1,1,1,1,0,0,1,1,1,1,1,0,1,1,0,1,1,1,1,1,
1,1,1,1,1,1,0,1,1,0,1,0,1,1,0,1,1,1,1,1,1,1,1,1,1,1,0,1,1,0,1,1,
1,1,1,0,0,1,1,1,1,1,1,1,0,0,1,1,1,1,1,1,1,1,1,1,1,1,1,0,0,1,1,1,
1,0,1,1,1,0,1,1,1,1,0,1,1,1,1,1,0,1,1,1,1,1,0,1,1,1,1,1,1,1,1,
0,0,1,1,0,1,1,1,1,0,1,1,1,1,1,0,0,1,1,0,1,0,1,1,1,0,1,1,1,1,0,1,
1,1,0,1,0,1,1,1,1,1,1,1,1,1,1,1,1,0,1,0,1,1,1,0,1,1,1,1,0,1,1,1,
0,1,0,1,1,0,0,1,1,1,1,1,0,1,1,1,1,0,1,1,0,0,1,1,1,1,1,1,1,1,0,1,
1,1,1,1,0,1,1,1,1,1,0,1,1,1,1,0,1,1,1,0,1,1,1,1,0,0,1,1,1,1,1,1,
1,1,1,1,1,1,1,0,0,1,1,1,1,1,1,1,0,0,1,1,1,1,1,0,1,1,0,1,1,1,1,1,
1,1,1,1,1,1,0,1,1,0,1,0,1,1,0,1,1,1,1,1,1,1,1,1,1,1,0,1,1,0,1,1,
1,1,1,0,0,1,1,1,1,1,1,1,0,0,1,1,1,1,1,1,1,1,1,1,1,1,1,0,0,1,1,1,
1,0,1,1,1,0,1,1,1,1,0,1,1,1,1,1,0,1,1,1,1,1,0,1,1,1,1,1,1,1,1,
0,0};
inline bool square( int64 x ) {
// Quickfail
if( x < 0 || (x&2) || ((x & 7) == 5) || ((x & 11) == 8) )
return false;
if( x == 0 )
return true;
// Check mod 255 = 3 * 5 * 17, for fun
int64 y = x;
y = (y & 4294967295LL) + (y >> 32);
y = (y & 65535) + (y >> 16);
y = (y & 255) + ((y >> 8) & 255) + (y >> 16);
if( bad255[y] )
return false;
// Divide out powers of 4 using binary search
if((x & 4294967295LL) == 0)
x >>= 32;
if((x & 65535) == 0)
x >>= 16;
if((x & 255) == 0)
x >>= 8;
if((x & 15) == 0)
x >>= 4;
if((x & 3) == 0)
x >>= 2;
if((x & 7) != 1)
return false;
// Compute sqrt using something like Hensel's lemma
int64 r, t, z;
r = start[(x >> 3) & 1023];
do {
z = x - r * r;
if( z == 0 )
return true;
if( z < 0 )
return false;
t = z & (-z);
r += (z & t) >> 1;
if( r > (t >> 1) )
r = t - r;
} while( t <= (1LL << 33) );
return false;
}

I'm pretty late to the party, but I hope to provide a better answer; shorter and (assuming my benchmark is correct) also much faster.
long goodMask; // 0xC840C04048404040 computed below
{
for (int i=0; i<64; ++i) goodMask |= Long.MIN_VALUE >>> (i*i);
}
public boolean isSquare(long x) {
// This tests if the 6 least significant bits are right.
// Moving the to be tested bit to the highest position saves us masking.
if (goodMask << x >= 0) return false;
final int numberOfTrailingZeros = Long.numberOfTrailingZeros(x);
// Each square ends with an even number of zeros.
if ((numberOfTrailingZeros & 1) != 0) return false;
x >>= numberOfTrailingZeros;
// Now x is either 0 or odd.
// In binary each odd square ends with 001.
// Postpone the sign test until now; handle zero in the branch.
if ((x&7) != 1 | x <= 0) return x == 0;
// Do it in the classical way.
// The correctness is not trivial as the conversion from long to double is lossy!
final long tst = (long) Math.sqrt(x);
return tst * tst == x;
}
The first test catches most non-squares quickly. It uses a 64-item table packed in a long, so there's no array access cost (indirection and bounds checks). For a uniformly random long, there's a 81.25% probability of ending here.
The second test catches all numbers having an odd number of twos in their factorization. The method Long.numberOfTrailingZeros is very fast as it gets JIT-ed into a single i86 instruction.
After dropping the trailing zeros, the third test handles numbers ending with 011, 101, or 111 in binary, which are no perfect squares. It also cares about negative numbers and also handles 0.
The final test falls back to double arithmetic. As double has only 53 bits mantissa,
the conversion from long to double includes rounding for big values. Nonetheless, the test is correct (unless the proof is wrong).
Trying to incorporate the mod255 idea wasn't successful.

You'll have to do some benchmarking. The best algorithm will depend on the distribution of your inputs.
Your algorithm may be nearly optimal, but you might want to do a quick check to rule out some possibilities before calling your square root routine. For example, look at the last digit of your number in hex by doing a bit-wise "and." Perfect squares can only end in 0, 1, 4, or 9 in base 16, So for 75% of your inputs (assuming they are uniformly distributed) you can avoid a call to the square root in exchange for some very fast bit twiddling.
Kip benchmarked the following code implementing the hex trick. When testing numbers 1 through 100,000,000, this code ran twice as fast as the original.
public final static boolean isPerfectSquare(long n)
{
if (n < 0)
return false;
switch((int)(n & 0xF))
{
case 0: case 1: case 4: case 9:
long tst = (long)Math.sqrt(n);
return tst*tst == n;
default:
return false;
}
}
When I tested the analogous code in C++, it actually ran slower than the original. However, when I eliminated the switch statement, the hex trick once again make the code twice as fast.
int isPerfectSquare(int n)
{
int h = n & 0xF; // h is the last hex "digit"
if (h > 9)
return 0;
// Use lazy evaluation to jump out of the if statement as soon as possible
if (h != 2 && h != 3 && h != 5 && h != 6 && h != 7 && h != 8)
{
int t = (int) floor( sqrt((double) n) + 0.5 );
return t*t == n;
}
return 0;
}
Eliminating the switch statement had little effect on the C# code.

I was thinking about the horrible times I've spent in Numerical Analysis course.
And then I remember, there was this function circling around the 'net from the Quake Source code:
float Q_rsqrt( float number )
{
long i;
float x2, y;
const float threehalfs = 1.5F;
x2 = number * 0.5F;
y = number;
i = * ( long * ) &y; // evil floating point bit level hacking
i = 0x5f3759df - ( i >> 1 ); // wtf?
y = * ( float * ) &i;
y = y * ( threehalfs - ( x2 * y * y ) ); // 1st iteration
// y = y * ( threehalfs - ( x2 * y * y ) ); // 2nd iteration, this can be removed
#ifndef Q3_VM
#ifdef __linux__
assert( !isnan(y) ); // bk010122 - FPE?
#endif
#endif
return y;
}
Which basically calculates a square root, using Newton's approximation function (cant remember the exact name).
It should be usable and might even be faster, it's from one of the phenomenal id software's game!
It's written in C++ but it should not be too hard to reuse the same technique in Java once you get the idea:
I originally found it at: http://www.codemaestro.com/reviews/9
Newton's method explained at wikipedia: http://en.wikipedia.org/wiki/Newton%27s_method
You can follow the link for more explanation of how it works, but if you don't care much, then this is roughly what I remember from reading the blog and from taking the Numerical Analysis course:
the * (long*) &y is basically a fast convert-to-long function so integer operations can be applied on the raw bytes.
the 0x5f3759df - (i >> 1); line is a pre-calculated seed value for the approximation function.
the * (float*) &i converts the value back to floating point.
the y = y * ( threehalfs - ( x2 * y * y ) ) line bascially iterates the value over the function again.
The approximation function gives more precise values the more you iterate the function over the result. In Quake's case, one iteration is "good enough", but if it wasn't for you... then you could add as much iteration as you need.
This should be faster because it reduces the number of division operations done in naive square rooting down to a simple divide by 2 (actually a * 0.5F multiply operation) and replace it with a few fixed number of multiplication operations instead.

I'm not sure if it would be faster, or even accurate, but you could use John Carmack's Magical Square Root, algorithm to solve the square root faster. You could probably easily test this for all possible 32 bit integers, and validate that you actually got correct results, as it's only an appoximation. However, now that I think about it, using doubles is approximating also, so I'm not sure how that would come into play.

If you do a binary chop to try to find the "right" square root, you can fairly easily detect if the value you've got is close enough to tell:
(n+1)^2 = n^2 + 2n + 1
(n-1)^2 = n^2 - 2n + 1
So having calculated n^2, the options are:
n^2 = target: done, return true
n^2 + 2n + 1 > target > n^2 : you're close, but it's not perfect: return false
n^2 - 2n + 1 < target < n^2 : ditto
target < n^2 - 2n + 1 : binary chop on a lower n
target > n^2 + 2n + 1 : binary chop on a higher n
(Sorry, this uses n as your current guess, and target for the parameter. Apologise for the confusion!)
I don't know whether this will be faster or not, but it's worth a try.
EDIT: The binary chop doesn't have to take in the whole range of integers, either (2^x)^2 = 2^(2x), so once you've found the top set bit in your target (which can be done with a bit-twiddling trick; I forget exactly how) you can quickly get a range of potential answers. Mind you, a naive binary chop is still only going to take up to 31 or 32 iterations.

I ran my own analysis of several of the algorithms in this thread and came up with some new results. You can see those old results in the edit history of this answer, but they're not accurate, as I made a mistake, and wasted time analyzing several algorithms which aren't close. However, pulling lessons from several different answers, I now have two algorithms that crush the "winner" of this thread. Here's the core thing I do differently than everyone else:
// This is faster because a number is divisible by 2^4 or more only 6% of the time
// and more than that a vanishingly small percentage.
while((x & 0x3) == 0) x >>= 2;
// This is effectively the same as the switch-case statement used in the original
// answer.
if((x & 0x7) != 1) return false;
However, this simple line, which most of the time adds one or two very fast instructions, greatly simplifies the switch-case statement into one if statement. However, it can add to the runtime if many of the tested numbers have significant power-of-two factors.
The algorithms below are as follows:
Internet - Kip's posted answer
Durron - My modified answer using the one-pass answer as a base
DurronTwo - My modified answer using the two-pass answer (by #JohnnyHeggheim), with some other slight modifications.
Here is a sample runtime if the numbers are generated using Math.abs(java.util.Random.nextLong())
0% Scenario{vm=java, trial=0, benchmark=Internet} 39673.40 ns; ?=378.78 ns # 3 trials
33% Scenario{vm=java, trial=0, benchmark=Durron} 37785.75 ns; ?=478.86 ns # 10 trials
67% Scenario{vm=java, trial=0, benchmark=DurronTwo} 35978.10 ns; ?=734.10 ns # 10 trials
benchmark us linear runtime
Internet 39.7 ==============================
Durron 37.8 ============================
DurronTwo 36.0 ===========================
vm: java
trial: 0
And here is a sample runtime if it's run on the first million longs only:
0% Scenario{vm=java, trial=0, benchmark=Internet} 2933380.84 ns; ?=56939.84 ns # 10 trials
33% Scenario{vm=java, trial=0, benchmark=Durron} 2243266.81 ns; ?=50537.62 ns # 10 trials
67% Scenario{vm=java, trial=0, benchmark=DurronTwo} 3159227.68 ns; ?=10766.22 ns # 3 trials
benchmark ms linear runtime
Internet 2.93 ===========================
Durron 2.24 =====================
DurronTwo 3.16 ==============================
vm: java
trial: 0
As you can see, DurronTwo does better for large inputs, because it gets to use the magic trick very very often, but gets clobbered compared to the first algorithm and Math.sqrt because the numbers are so much smaller. Meanwhile, the simpler Durron is a huge winner because it never has to divide by 4 many many times in the first million numbers.
Here's Durron:
public final static boolean isPerfectSquareDurron(long n) {
if(n < 0) return false;
if(n == 0) return true;
long x = n;
// This is faster because a number is divisible by 16 only 6% of the time
// and more than that a vanishingly small percentage.
while((x & 0x3) == 0) x >>= 2;
// This is effectively the same as the switch-case statement used in the original
// answer.
if((x & 0x7) == 1) {
long sqrt;
if(x < 410881L)
{
int i;
float x2, y;
x2 = x * 0.5F;
y = x;
i = Float.floatToRawIntBits(y);
i = 0x5f3759df - ( i >> 1 );
y = Float.intBitsToFloat(i);
y = y * ( 1.5F - ( x2 * y * y ) );
sqrt = (long)(1.0F/y);
} else {
sqrt = (long) Math.sqrt(x);
}
return sqrt*sqrt == x;
}
return false;
}
And DurronTwo
public final static boolean isPerfectSquareDurronTwo(long n) {
if(n < 0) return false;
// Needed to prevent infinite loop
if(n == 0) return true;
long x = n;
while((x & 0x3) == 0) x >>= 2;
if((x & 0x7) == 1) {
long sqrt;
if (x < 41529141369L) {
int i;
float x2, y;
x2 = x * 0.5F;
y = x;
i = Float.floatToRawIntBits(y);
//using the magic number from
//http://www.lomont.org/Math/Papers/2003/InvSqrt.pdf
//since it more accurate
i = 0x5f375a86 - (i >> 1);
y = Float.intBitsToFloat(i);
y = y * (1.5F - (x2 * y * y));
y = y * (1.5F - (x2 * y * y)); //Newton iteration, more accurate
sqrt = (long) ((1.0F/y) + 0.2);
} else {
//Carmack hack gives incorrect answer for n >= 41529141369.
sqrt = (long) Math.sqrt(x);
}
return sqrt*sqrt == x;
}
return false;
}
And my benchmark harness: (Requires Google caliper 0.1-rc5)
public class SquareRootBenchmark {
public static class Benchmark1 extends SimpleBenchmark {
private static final int ARRAY_SIZE = 10000;
long[] trials = new long[ARRAY_SIZE];
#Override
protected void setUp() throws Exception {
Random r = new Random();
for (int i = 0; i < ARRAY_SIZE; i++) {
trials[i] = Math.abs(r.nextLong());
}
}
public int timeInternet(int reps) {
int trues = 0;
for(int i = 0; i < reps; i++) {
for(int j = 0; j < ARRAY_SIZE; j++) {
if(SquareRootAlgs.isPerfectSquareInternet(trials[j])) trues++;
}
}
return trues;
}
public int timeDurron(int reps) {
int trues = 0;
for(int i = 0; i < reps; i++) {
for(int j = 0; j < ARRAY_SIZE; j++) {
if(SquareRootAlgs.isPerfectSquareDurron(trials[j])) trues++;
}
}
return trues;
}
public int timeDurronTwo(int reps) {
int trues = 0;
for(int i = 0; i < reps; i++) {
for(int j = 0; j < ARRAY_SIZE; j++) {
if(SquareRootAlgs.isPerfectSquareDurronTwo(trials[j])) trues++;
}
}
return trues;
}
}
public static void main(String... args) {
Runner.main(Benchmark1.class, args);
}
}
UPDATE: I've made a new algorithm that is faster in some scenarios, slower in others, I've gotten different benchmarks based on different inputs. If we calculate modulo 0xFFFFFF = 3 x 3 x 5 x 7 x 13 x 17 x 241, we can eliminate 97.82% of numbers that cannot be squares. This can be (sort of) done in one line, with 5 bitwise operations:
if (!goodLookupSquares[(int) ((n & 0xFFFFFFl) + ((n >> 24) & 0xFFFFFFl) + (n >> 48))]) return false;
The resulting index is either 1) the residue, 2) the residue + 0xFFFFFF, or 3) the residue + 0x1FFFFFE. Of course, we need to have a lookup table for residues modulo 0xFFFFFF, which is about a 3mb file (in this case stored as ascii text decimal numbers, not optimal but clearly improvable with a ByteBuffer and so forth. But since that is precalculation it doesn't matter so much. You can find the file here (or generate it yourself):
public final static boolean isPerfectSquareDurronThree(long n) {
if(n < 0) return false;
if(n == 0) return true;
long x = n;
while((x & 0x3) == 0) x >>= 2;
if((x & 0x7) == 1) {
if (!goodLookupSquares[(int) ((n & 0xFFFFFFl) + ((n >> 24) & 0xFFFFFFl) + (n >> 48))]) return false;
long sqrt;
if(x < 410881L)
{
int i;
float x2, y;
x2 = x * 0.5F;
y = x;
i = Float.floatToRawIntBits(y);
i = 0x5f3759df - ( i >> 1 );
y = Float.intBitsToFloat(i);
y = y * ( 1.5F - ( x2 * y * y ) );
sqrt = (long)(1.0F/y);
} else {
sqrt = (long) Math.sqrt(x);
}
return sqrt*sqrt == x;
}
return false;
}
I load it into a boolean array like this:
private static boolean[] goodLookupSquares = null;
public static void initGoodLookupSquares() throws Exception {
Scanner s = new Scanner(new File("24residues_squares.txt"));
goodLookupSquares = new boolean[0x1FFFFFE];
while(s.hasNextLine()) {
int residue = Integer.valueOf(s.nextLine());
goodLookupSquares[residue] = true;
goodLookupSquares[residue + 0xFFFFFF] = true;
goodLookupSquares[residue + 0x1FFFFFE] = true;
}
s.close();
}
Example runtime. It beat Durron (version one) in every trial I ran.
0% Scenario{vm=java, trial=0, benchmark=Internet} 40665.77 ns; ?=566.71 ns # 10 trials
33% Scenario{vm=java, trial=0, benchmark=Durron} 38397.60 ns; ?=784.30 ns # 10 trials
67% Scenario{vm=java, trial=0, benchmark=DurronThree} 36171.46 ns; ?=693.02 ns # 10 trials
benchmark us linear runtime
Internet 40.7 ==============================
Durron 38.4 ============================
DurronThree 36.2 ==========================
vm: java
trial: 0

It should be much faster to use Newton's method to calculate the Integer Square Root, then square this number and check, as you do in your current solution. Newton's method is the basis for the Carmack solution mentioned in some other answers. You should be able to get a faster answer since you're only interested in the integer part of the root, allowing you to stop the approximation algorithm sooner.
Another optimization that you can try: If the Digital Root of a number doesn't end in
1, 4, 7, or 9 the number is not a perfect square. This can be used as a quick way to eliminate 60% of your inputs before applying the slower square root algorithm.

I want this function to work with all
positive 64-bit signed integers
Math.sqrt() works with doubles as input parameters, so you won't get accurate results for integers bigger than 2^53.

An integer problem deserves an integer solution. Thus
Do binary search on the (non-negative) integers to find the greatest integer t such that t**2 <= n. Then test whether r**2 = n exactly. This takes time O(log n).
If you don't know how to binary search the positive integers because the set is unbounded, it's easy. You starting by computing your increasing function f (above f(t) = t**2 - n) on powers of two. When you see it turn positive, you've found an upper bound. Then you can do standard binary search.

Just for the record, another approach is to use the prime decomposition. If every factor of the decomposition is even, then the number is a perfect square. So what you want is to see if a number can be decomposed as a product of squares of prime numbers. Of course, you don't need to obtain such a decomposition, just to see if it exists.
First build a table of squares of prime numbers which are lower than 2^32. This is far smaller than a table of all integers up to this limit.
A solution would then be like this:
boolean isPerfectSquare(long number)
{
if (number < 0) return false;
if (number < 2) return true;
for (int i = 0; ; i++)
{
long square = squareTable[i];
if (square > number) return false;
while (number % square == 0)
{
number /= square;
}
if (number == 1) return true;
}
}
I guess it's a bit cryptic. What it does is checking in every step that the square of a prime number divide the input number. If it does then it divides the number by the square as long as it is possible, to remove this square from the prime decomposition.
If by this process, we came to 1, then the input number was a decomposition of square of prime numbers. If the square becomes larger than the number itself, then there is no way this square, or any larger squares, can divide it, so the number can not be a decomposition of squares of prime numbers.
Given nowadays' sqrt done in hardware and the need to compute prime numbers here, I guess this solution is way slower. But it should give better results than solution with sqrt which won't work over 2^54, as says mrzl in his answer.

It's been pointed out that the last d digits of a perfect square can only take on certain values. The last d digits (in base b) of a number n is the same as the remainder when n is divided by bd, ie. in C notation n % pow(b, d).
This can be generalized to any modulus m, ie. n % m can be used to rule out some percentage of numbers from being perfect squares. The modulus you are currently using is 64, which allows 12, ie. 19% of remainders, as possible squares. With a little coding I found the modulus 110880, which allows only 2016, ie. 1.8% of remainders as possible squares. So depending on the cost of a modulus operation (ie. division) and a table lookup versus a square root on your machine, using this modulus might be faster.
By the way if Java has a way to store a packed array of bits for the lookup table, don't use it. 110880 32-bit words is not much RAM these days and fetching a machine word is going to be faster than fetching a single bit.

The following simplification of maaartinus's solution appears to shave a few percentage points off the runtime, but I'm not good enough at benchmarking to produce a benchmark I can trust:
long goodMask; // 0xC840C04048404040 computed below
{
for (int i=0; i<64; ++i) goodMask |= Long.MIN_VALUE >>> (i*i);
}
public boolean isSquare(long x) {
// This tests if the 6 least significant bits are right.
// Moving the to be tested bit to the highest position saves us masking.
if (goodMask << x >= 0) return false;
// Remove an even number of trailing zeros, leaving at most one.
x >>= (Long.numberOfTrailingZeros(x) & (-2);
// Repeat the test on the 6 least significant remaining bits.
if (goodMask << x >= 0 | x <= 0) return x == 0;
// Do it in the classical way.
// The correctness is not trivial as the conversion from long to double is lossy!
final long tst = (long) Math.sqrt(x);
return tst * tst == x;
}
It would be worth checking how omitting the first test,
if (goodMask << x >= 0) return false;
would affect performance.

For performance, you very often have to do some compromsies. Others have expressed various methods, however, you noted Carmack's hack was faster up to certain values of N. Then, you should check the "n" and if it is less than that number N, use Carmack's hack, else use some other method described in the answers here.

This is the fastest Java implementation I could come up with, using a combination of techniques suggested by others in this thread.
Mod-256 test
Inexact mod-3465 test (avoids integer division at the cost of some false positives)
Floating-point square root, round and compare with input value
I also experimented with these modifications but they did not help performance:
Additional mod-255 test
Dividing the input value by powers of 4
Fast Inverse Square Root (to work for high values of N it needs 3 iterations, enough to make it slower than the hardware square root function.)
public class SquareTester {
public static boolean isPerfectSquare(long n) {
if (n < 0) {
return false;
} else {
switch ((byte) n) {
case -128: case -127: case -124: case -119: case -112:
case -111: case -103: case -95: case -92: case -87:
case -79: case -71: case -64: case -63: case -60:
case -55: case -47: case -39: case -31: case -28:
case -23: case -15: case -7: case 0: case 1:
case 4: case 9: case 16: case 17: case 25:
case 33: case 36: case 41: case 49: case 57:
case 64: case 65: case 68: case 73: case 81:
case 89: case 97: case 100: case 105: case 113:
case 121:
long i = (n * INV3465) >>> 52;
if (! good3465[(int) i]) {
return false;
} else {
long r = round(Math.sqrt(n));
return r*r == n;
}
default:
return false;
}
}
}
private static int round(double x) {
return (int) Double.doubleToRawLongBits(x + (double) (1L << 52));
}
/** 3465<sup>-1</sup> modulo 2<sup>64</sup> */
private static final long INV3465 = 0x8ffed161732e78b9L;
private static final boolean[] good3465 =
new boolean[0x1000];
static {
for (int r = 0; r < 3465; ++ r) {
int i = (int) ((r * r * INV3465) >>> 52);
good3465[i] = good3465[i+1] = true;
}
}
}

You should get rid of the 2-power part of N right from the start.
2nd Edit
The magical expression for m below should be
m = N - (N & (N-1));
and not as written
End of 2nd edit
m = N & (N-1); // the lawest bit of N
N /= m;
byte = N & 0x0F;
if ((m % 2) || (byte !=1 && byte !=9))
return false;
1st Edit:
Minor improvement:
m = N & (N-1); // the lawest bit of N
N /= m;
if ((m % 2) || (N & 0x07 != 1))
return false;
End of 1st edit
Now continue as usual. This way, by the time you get to the floating point part, you already got rid of all the numbers whose 2-power part is odd (about half), and then you only consider 1/8 of whats left. I.e. you run the floating point part on 6% of the numbers.

Project Euler is mentioned in the tags and many of the problems in it require checking numbers >> 2^64. Most of the optimizations mentioned above don't work easily when you are working with an 80 byte buffer.
I used java BigInteger and a slightly modified version of Newton's method, one that works better with integers. The problem was that exact squares n^2 converged to (n-1) instead of n because n^2-1 = (n-1)(n+1) and the final error was just one step below the final divisor and the algorithm terminated. It was easy to fix by adding one to the original argument before computing the error. (Add two for cube roots, etc.)
One nice attribute of this algorithm is that you can immediately tell if the number is a perfect square - the final error (not correction) in Newton's method will be zero. A simple modification also lets you quickly calculate floor(sqrt(x)) instead of the closest integer. This is handy with several Euler problems.

The sqrt call is not perfectly accurate, as has been mentioned, but it's interesting and instructive that it doesn't blow away the other answers in terms of speed. After all, the sequence of assembly language instructions for a sqrt is tiny. Intel has a hardware instruction, which isn't used by Java I believe because it doesn't conform to IEEE.
So why is it slow? Because Java is actually calling a C routine through JNI, and it's actually slower to do so than to call a Java subroutine, which itself is slower than doing it inline. This is very annoying, and Java should have come up with a better solution, ie building in floating point library calls if necessary. Oh well.
In C++, I suspect all the complex alternatives would lose on speed, but I haven't checked them all.
What I did, and what Java people will find usefull, is a simple hack, an extension of the special case testing suggested by A. Rex. Use a single long value as a bit array, which isn't bounds checked. That way, you have 64 bit boolean lookup.
typedef unsigned long long UVLONG
UVLONG pp1,pp2;
void init2() {
for (int i = 0; i < 64; i++) {
for (int j = 0; j < 64; j++)
if (isPerfectSquare(i * 64 + j)) {
pp1 |= (1 << j);
pp2 |= (1 << i);
break;
}
}
cout << "pp1=" << pp1 << "," << pp2 << "\n";
}
inline bool isPerfectSquare5(UVLONG x) {
return pp1 & (1 << (x & 0x3F)) ? isPerfectSquare(x) : false;
}
The routine isPerfectSquare5 runs in about 1/3 the time on my core2 duo machine. I suspect that further tweaks along the same lines could reduce the time further on average, but every time you check, you are trading off more testing for more eliminating, so you can't go too much farther on that road.
Certainly, rather than having a separate test for negative, you could check the high 6 bits the same way.
Note that all I'm doing is eliminating possible squares, but when I have a potential case I have to call the original, inlined isPerfectSquare.
The init2 routine is called once to initialize the static values of pp1 and pp2.
Note that in my implementation in C++, I'm using unsigned long long, so since you're signed, you'd have to use the >>> operator.
There is no intrinsic need to bounds check the array, but Java's optimizer has to figure this stuff out pretty quickly, so I don't blame them for that.

I like the idea to use an almost correct method on some of the input. Here is a version with a higher "offset". The code seems to work and passes my simple test case.
Just replace your:
if(n < 410881L){...}
code with this one:
if (n < 11043908100L) {
//John Carmack hack, converted to Java.
// See: http://www.codemaestro.com/reviews/9
int i;
float x2, y;
x2 = n * 0.5F;
y = n;
i = Float.floatToRawIntBits(y);
//using the magic number from
//http://www.lomont.org/Math/Papers/2003/InvSqrt.pdf
//since it more accurate
i = 0x5f375a86 - (i >> 1);
y = Float.intBitsToFloat(i);
y = y * (1.5F - (x2 * y * y));
y = y * (1.5F - (x2 * y * y)); //Newton iteration, more accurate
sqrt = Math.round(1.0F / y);
} else {
//Carmack hack gives incorrect answer for n >= 11043908100.
sqrt = (long) Math.sqrt(n);
}

Considering for general bit length (though I have used specific type here), I tried to design simplistic algo as below. Simple and obvious check for 0,1,2 or <0 is required initially.
Following is simple in sense that it doesn't try to use any existing maths functions. Most of the operator can be replaced with bit-wise operators. I haven't tested with any bench mark data though. I'm neither expert at maths or computer algorithm design in particular, I would love to see you pointing out problem. I know there is lots of improvement chances there.
int main()
{
unsigned int c1=0 ,c2 = 0;
unsigned int x = 0;
unsigned int p = 0;
int k1 = 0;
scanf("%d",&p);
if(p % 2 == 0) {
x = p/2;
}
else {
x = (p/2) +1;
}
while(x)
{
if((x*x) > p) {
c1 = x;
x = x/2;
}else {
c2 = x;
break;
}
}
if((p%2) != 0)
c2++;
while(c2 < c1)
{
if((c2 * c2 ) == p) {
k1 = 1;
break;
}
c2++;
}
if(k1)
printf("\n Perfect square for %d", c2);
else
printf("\n Not perfect but nearest to :%d :", c2);
return 0;
}

This a rework from decimal to binary of the old Marchant calculator algorithm (sorry, I don't have a reference), in Ruby, adapted specifically for this question:
def isexactsqrt(v)
value = v.abs
residue = value
root = 0
onebit = 1
onebit <<= 8 while (onebit < residue)
onebit >>= 2 while (onebit > residue)
while (onebit > 0)
x = root + onebit
if (residue >= x) then
residue -= x
root = x + onebit
end
root >>= 1
onebit >>= 2
end
return (residue == 0)
end
Here's a workup of something similar (there may be coding style/smells or clunky O/O - it's the algorithm that counts, and C++ is not my home language). In this case, we're looking for residue == 0:
#include <iostream>
using namespace std;
typedef unsigned long long int llint;
class ISqrt { // Integer Square Root
llint value; // Integer whose square root is required
llint root; // Result: floor(sqrt(value))
llint residue; // Result: value-root*root
llint onebit, x; // Working bit, working value
public:
ISqrt(llint v = 2) { // Constructor
Root(v); // Take the root
};
llint Root(llint r) { // Resets and calculates new square root
value = r; // Store input
residue = value; // Initialise for subtracting down
root = 0; // Clear root accumulator
onebit = 1; // Calculate start value of counter
onebit <<= (8*sizeof(llint)-2); // Set up counter bit as greatest odd power of 2
while (onebit > residue) {onebit >>= 2; }; // Shift down until just < value
while (onebit > 0) {
x = root ^ onebit; // Will check root+1bit (root bit corresponding to onebit is always zero)
if (residue >= x) { // Room to subtract?
residue -= x; // Yes - deduct from residue
root = x + onebit; // and step root
};
root >>= 1;
onebit >>= 2;
};
return root;
};
llint Residue() { // Returns residue from last calculation
return residue;
};
};
int main() {
llint big, i, q, r, v, delta;
big = 0; big = (big-1); // Kludge for "big number"
ISqrt b; // Make q sqrt generator
for ( i = big; i > 0 ; i /= 7 ) { // for several numbers
q = b.Root(i); // Get the square root
r = b.Residue(); // Get the residue
v = q*q+r; // Recalc original value
delta = v-i; // And diff, hopefully 0
cout << i << ": " << q << " ++ " << r << " V: " << v << " Delta: " << delta << "\n";
};
return 0;
};

I checked all of the possible results when the last n bits of a square is observed. By successively examining more bits, up to 5/6th of inputs can be eliminated. I actually designed this to implement Fermat's Factorization algorithm, and it is very fast there.
public static boolean isSquare(final long val) {
if ((val & 2) == 2 || (val & 7) == 5) {
return false;
}
if ((val & 11) == 8 || (val & 31) == 20) {
return false;
}
if ((val & 47) == 32 || (val & 127) == 80) {
return false;
}
if ((val & 191) == 128 || (val & 511) == 320) {
return false;
}
// if((val & a == b) || (val & c == d){
// return false;
// }
if (!modSq[(int) (val % modSq.length)]) {
return false;
}
final long root = (long) Math.sqrt(val);
return root * root == val;
}
The last bit of pseudocode can be used to extend the tests to eliminate more values. The tests above are for k = 0, 1, 2, 3
a is of the form (3 << 2k) - 1
b is of the form (2 << 2k)
c is of the form (2 << 2k + 2) - 1
d is of the form (2 << 2k - 1) * 10
It first tests whether it has a square residual with moduli of power of two, then it tests based on a final modulus, then it uses the Math.sqrt to do a final test. I came up with the idea from the top post, and attempted to extend upon it. I appreciate any comments or suggestions.
Update: Using the test by a modulus, (modSq) and a modulus base of 44352, my test runs in 96% of the time of the one in the OP's update for numbers up to 1,000,000,000.

Here is a divide and conquer solution.
If the square root of a natural number (number) is a natural number (solution), you can easily determine a range for solution based on the number of digits of number:
number has 1 digit: solution in range = 1 - 4
number has 2 digits: solution in range = 3 - 10
number has 3 digits: solution in range = 10 - 40
number has 4 digits: solution in range = 30 - 100
number has 5 digits: solution in range = 100 - 400
Notice the repetition?
You can use this range in a binary search approach to see if there is a solution for which:
number == solution * solution
Here is the code
Here is my class SquareRootChecker
public class SquareRootChecker {
private long number;
private long initialLow;
private long initialHigh;
public SquareRootChecker(long number) {
this.number = number;
initialLow = 1;
initialHigh = 4;
if (Long.toString(number).length() % 2 == 0) {
initialLow = 3;
initialHigh = 10;
}
for (long i = 0; i < Long.toString(number).length() / 2; i++) {
initialLow *= 10;
initialHigh *= 10;
}
if (Long.toString(number).length() % 2 == 0) {
initialLow /= 10;
initialHigh /=10;
}
}
public boolean checkSquareRoot() {
return findSquareRoot(initialLow, initialHigh, number);
}
private boolean findSquareRoot(long low, long high, long number) {
long check = low + (high - low) / 2;
if (high >= low) {
if (number == check * check) {
return true;
}
else if (number < check * check) {
high = check - 1;
return findSquareRoot(low, high, number);
}
else {
low = check + 1;
return findSquareRoot(low, high, number);
}
}
return false;
}
}
And here is an example on how to use it.
long number = 1234567;
long square = number * number;
SquareRootChecker squareRootChecker = new SquareRootChecker(square);
System.out.println(square + ": " + squareRootChecker.checkSquareRoot()); //Prints "1524155677489: true"
long notSquare = square + 1;
squareRootChecker = new SquareRootChecker(notSquare);
System.out.println(notSquare + ": " + squareRootChecker.checkSquareRoot()); //Prints "1524155677490: false"

Newton's Method with integer arithmetic
If you wish to avoid non-integer operations you could use the method below. It basically uses Newton's Method modified for integer arithmetic.
/**
* Test if the given number is a perfect square.
* #param n Must be greater than 0 and less
* than Long.MAX_VALUE.
* #return <code>true</code> if n is a perfect
* square, or <code>false</code> otherwise.
*/
public static boolean isSquare(long n)
{
long x1 = n;
long x2 = 1L;
while (x1 > x2)
{
x1 = (x1 + x2) / 2L;
x2 = n / x1;
}
return x1 == x2 && n % x1 == 0L;
}
This implementation can not compete with solutions that use Math.sqrt. However, its performance can be improved by using the filtering mechanisms described in some of the other posts.

Square Root of a number, given that the number is a perfect square.
The complexity is log(n)
/**
* Calculate square root if the given number is a perfect square.
*
* Approach: Sum of n odd numbers is equals to the square root of n*n, given
* that n is a perfect square.
*
* #param number
* #return squareRoot
*/
public static int calculateSquareRoot(int number) {
int sum=1;
int count =1;
int squareRoot=1;
while(sum<number) {
count+=2;
sum+=count;
squareRoot++;
}
return squareRoot;
}

Here is the simplest and most concise way, although I do not know how it compares in terms of CPU cycles. This works great if you only wish to know if the root is a whole number. If you really care if it is an integer, you can also figure that out. Here is a simple (and pure) function:
private static final MathContext precision = new MathContext(20);
private static final Function<Long, Boolean> isRootWhole = (n) -> {
long digit = n % 10;
if (digit == 2 || digit == 3 || digit == 7 || digit == 8) {
return false;
}
return new BigDecimal(n).sqrt(precision).scale() == 0;
};
If you do not need micro-optimization, this answer is better in terms of simplicity and maintainability. If you will be calculating negative numbers, you will need to handle that accordingly, and send the absolute value into the function. I have included a minor optimization because no perfect squares have a tens digit of 2, 3, 7, or 8 due to quadratic residues mod 10.
On my CPU, a run of this algorithm on 0 - 10,000,000 took an average of 1000 - 1100 nanoseconds per calculation.
If you are performing a lesser number of calculations, the earlier calculations take a bit longer.
I had a negative comment that my previous edit did not work for large numbers. The OP mentioned Longs, and the largest perfect square that is a Long is 9223372030926249001, so this method works for all Longs.

This question got me wondering, so I did some simple coding and I'm presenting it here because I think it's interesting, relevant, but I don't know how useful. There's a simple algorithm
a_n+1 = (a_n + x/a_n)/2
for calculating square roots, but it's meant to be used for decimals. I wondered what would happen if I just coded the same algorithm using integer maths. Would it even converge on the right answer? I didn't know, so I wrote a program...
#include <stdio.h>
#include <stdint.h>
#include <stdlib.h>
#include <math.h>
_Bool isperfectsquare(uint64_t x, uint64_t *isqrtx) {
// NOTE: isqrtx approximate for non-squares. (benchmarked at 162ns 3GHz i5)
uint32_t i;
uint64_t ai;
ai = 1 + ((x & 0xffff000000000000) >> 32) + ((x & 0xffff00000000) >> 24) + ((x & 0xffff0000) >> 16);
ai = (ai + x/ai)/2;
ai = (ai + x/ai)/2;
ai = (ai + x/ai)/2;
ai = (ai + x/ai)/2;
ai = (ai + x/ai)/2;
ai = (ai + x/ai)/2;
ai = (ai + x/ai)/2;
ai = (ai + x/ai)/2;
ai = (ai + x/ai)/2;
ai = (ai + x/ai)/2;
ai = (ai + x/ai)/2;
ai = (ai + x/ai)/2;
ai = ai & 0xffffffff;
if (isqrtx != NULL) isqrtx[0] = ai;
return ai*ai == x;
}
void main() {
uint64_t x, isqrtx;
uint64_t i;
for (i=1; i<0x100000000; i++) {
if (!isperfectsquare(i*i, &isqrtx)) {
printf("Failed at %li", i);
exit(1);
}
}
printf("All OK.\n");
}
So, it turns out that 12 iterations of the formula is enough to give correct results for all 64 bit unsigned longs that are perfect squares, and of course, non-squares will return false.
simon#simon-Inspiron-N5040:~$ time ./isqrt.bin
All OK.
real 11m37.096s
user 11m35.053s
sys 0m0.272s
So 697s/2^32 is approx 162ns. As it is, the function will have the same runtime for all inputs. Some of the measures detailed elsewhere in the discussion could speed it up for non-squares by checking the last four bits etc. Hope someone finds this interesting as I did.

If speed is a concern, why not partition off the most commonly used set of inputs and their values to a lookup table and then do whatever optimized magic algorithm you have come up with for the exceptional cases?

"I'm looking for the fastest way to determine if a long value is a perfect square (i.e. its square root is another integer)."
The answers are impressive, but I failed to see a simple check :
check whether the first number on the right of the long it a member of the set (0,1,4,5,6,9) . If it is not, then it cannot possibly be a 'perfect square' .
eg.
4567 - cannot be a perfect square.

It ought to be possible to pack the 'cannot be a perfect square if the last X digits are N' much more efficiently than that! I'll use java 32 bit ints, and produce enough data to check the last 16 bits of the number - that's 2048 hexadecimal int values.
...
Ok. Either I have run into some number theory that is a little beyond me, or there is a bug in my code. In any case, here is the code:
public static void main(String[] args) {
final int BITS = 16;
BitSet foo = new BitSet();
for(int i = 0; i< (1<<BITS); i++) {
int sq = (i*i);
sq = sq & ((1<<BITS)-1);
foo.set(sq);
}
System.out.println("int[] mayBeASquare = {");
for(int i = 0; i< 1<<(BITS-5); i++) {
int kk = 0;
for(int j = 0; j<32; j++) {
if(foo.get((i << 5) | j)) {
kk |= 1<<j;
}
}
System.out.print("0x" + Integer.toHexString(kk) + ", ");
if(i%8 == 7) System.out.println();
}
System.out.println("};");
}
and here are the results:
(ed: elided for poor performance in prettify.js; view revision history to see.)