<< problem 47 - Distinct primes factors | Prime permutations - problem 49 >> |

# Problem 48: Self powers

(see projecteuler.net/problem=48)

The series, 1^1 + 2^2 + 3^3 + ... + 10^10 = 10405071317.

Find the last ten digits of the series, 1^1 + 2^2 + 3^3 + ... + 1000^1000.

# Algorithm

This problem is pretty easy for languages with support for BigIntegers (such as Java and Python).

I don't want to use external libraries and thus have to write some code in C++ ...

welcome to the world of modular arithmetic !

There are two functions:

- `mulmod`

is the same as `(a*b)%modulo`

where `a*b`

is allowed to exceed 64 bits (`unsigned long long`

).

- `powmod`

is the same as `(base^exponent)%modulo`

where `base^exponent`

is allowed to exceed 64 bits (`unsigned long long`

).

`mulmod`

has two paths:

- if both factors are small (fit in 32 bits) then there will be no overflow and the original formula `(a * b) % modulo`

is executed, which is pretty fast

- else `a`

and `b`

are multiplied bitwise (similar to written multiplication), which is much slower

`powmod`

employs the tricks of en.wikipedia.org/wiki/Exponentiation_by_squaring

## Alternative Approaches

GCC supports `__int128`

which allows to use the fast path of `mulmod`

all the time.

See my toolbox for more alternatives.

# My code

… was written in C++ and can be compiled with G++, Clang++, Visual C++. You can download it, too.

#include <iostream>
// return (a*b) % modulo
// note: NOT USED because mulmod() is faster (see below)

unsigned long long mulmodBitwise(unsigned long long a, unsigned long long b, unsigned long long modulo)
{
// (a * b) % modulo = (a % modulo) * (b % modulo) % modulo
a %= modulo;
b %= modulo;
// fast path
if (a <= 0xFFFFFFF && b <= 0xFFFFFFF)
return (a * b) % modulo;
// we might encounter overflows (slow path)
// the number of loops depends on b, therefore try to minimize b
if (b > a)
std::swap(a, b);
// bitwise multiplication
unsigned long long result = 0;
while (a > 0 && b > 0)
{
// b is odd ? a*b = a + a*(b-1)
if (b & 1)
{
result += a;
result %= modulo;
// skip b-- because the bit-shift at the end will remove the lowest bit anyway
}
// b is even ? a*b = (2*a)*(b/2)
a <<= 1;
a %= modulo;
// next bit
b >>= 1;
}
return result;
}
// return (a*b) % modulo
// very similar to mulmodBitWise, but multiple bits are processed at once (instead of just 1 bit per iteration)

unsigned long long mulmod(unsigned long long a, unsigned long long b, unsigned long long modulo)
{
// (a * b) % modulo = (a % modulo) * (b % modulo) % modulo
a %= modulo;
b %= modulo;
// fast path
if (a <= 0xFFFFFFF && b <= 0xFFFFFFF)
return (a * b) % modulo;
// count leading zero bits of modulo
unsigned int leadingZeroes = 0;
unsigned long long m = modulo;
while ((m & 0x8000000000000000ULL) == 0)
{
leadingZeroes++;
m <<= 1;
}
// cover all bits of modulo
unsigned long long mask = (1 << leadingZeroes) - 1;
// blockwise multiplication
unsigned long long result = 0;
while (a > 0 && b > 0)
{
result += (b & mask) * a;
result %= modulo;
// next bits
b >>= leadingZeroes;
a <<= leadingZeroes;
a %= modulo;
}
return result;
}
// return (base^exponent) % modulo

unsigned long long powmod(unsigned long long base, unsigned long long exponent, unsigned long long modulo)
{
unsigned long long result = 1;
while (exponent > 0)
{
// fast exponentation:
// odd exponent ? a^b = a*a^(b-1)
if (exponent & 1)
result = mulmod(result, base, modulo);
// even exponent ? a^b = (a*a)^(b/2)
base = mulmod(base, base, modulo);
exponent >>= 1;
}
return result;
}
int main()
{
// sum from 1^1 to x^x
unsigned int x;
std::cin >> x;
// keep the last 10 digits
const unsigned long long TenDigits = 10000000000ULL;
unsigned long long sum = 0;
// add all parts and don't forget the modulo ...
for (unsigned int i = 1; i <= x; i++)
sum += powmod(i, i, TenDigits);
std::cout << (sum % TenDigits) << std::endl;
return 0;
}

This solution contains 19 empty lines, 26 comments and 1 preprocessor command.

# Interactive test

You can submit your own input to my program and it will be instantly processed at my server:

This is equivalent to`echo 10 | ./48`

Output:

*Note:* the original problem's input `1000`

__cannot__ be entered

because just copying results is a soft skill reserved for idiots.

*(this interactive test is still under development, computations will be aborted after one second)*

# Benchmark

The correct solution to the original Project Euler problem was found in **less than 0.01** seconds on a Intel® Core™ i7-2600K CPU @ 3.40GHz.

(compiled for x86_64 / Linux, GCC flags: `-O3 -march=native -fno-exceptions -fno-rtti -std=c++11 -DORIGINAL`

)

See here for a comparison of all solutions.

**Note:** interactive tests run on a weaker (=slower) computer. Some interactive tests are compiled without `-DORIGINAL`

.

# Changelog

February 26, 2017 submitted solution

April 19, 2017 added comments

# Hackerrank

see https://www.hackerrank.com/contests/projecteuler/challenges/euler048

My code solved **6** out of **6** test cases (score: **100%**)

# Difficulty

Project Euler ranks this problem at **5%** (out of 100%).

Hackerrank describes this problem as **easy**.

*Note:*

Hackerrank has strict execution time limits (typically 2 seconds for C++ code) and often a much wider input range than the original problem.

In my opinion, Hackerrank's modified problems are usually a lot harder to solve. As a rule thumb: brute-force is never an option.

# Links

projecteuler.net/thread=48 - **the** best forum on the subject (*note:* you have to submit the correct solution first)

Code in various languages:

Python: www.mathblog.dk/project-euler-48-last-ten-digits/ (written by Kristian Edlund)

Haskell: github.com/nayuki/Project-Euler-solutions/blob/master/haskell/p048.hs (written by Nayuki)

Java: github.com/nayuki/Project-Euler-solutions/blob/master/java/p048.java (written by Nayuki)

Mathematica: github.com/nayuki/Project-Euler-solutions/blob/master/mathematica/p048.mathematica (written by Nayuki)

C: github.com/eagletmt/project-euler-c/blob/master/40-49/problem48.c (written by eagletmt)

Javascript: github.com/dsernst/ProjectEuler/blob/master/48 Self powers.js (written by David Ernst)

Scala: github.com/samskivert/euler-scala/blob/master/Euler048.scala (written by Michael Bayne)

# Heatmap

green problems solve the original Project Euler problem and have a perfect score of 100% at Hackerrank, too.

yellow problems score less than 100% at Hackerrank (but still solve the original problem).

gray problems are already solved but I haven't published my solution yet.

blue problems are already solved and there wasn't a Hackerrank version of it (at the time I solved it) or I didn't care about it because it differed too much.

*Please click on a problem's number to open my solution to that problem:*

1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 |

26 | 27 | 28 | 29 | 30 | 31 | 32 | 33 | 34 | 35 | 36 | 37 | 38 | 39 | 40 | 41 | 42 | 43 | 44 | 45 | 46 | 47 | 48 | 49 | 50 |

51 | 52 | 53 | 54 | 55 | 56 | 57 | 58 | 59 | 60 | 61 | 62 | 63 | 64 | 65 | 66 | 67 | 68 | 69 | 70 | 71 | 72 | 73 | 74 | 75 |

76 | 77 | 78 | 79 | 80 | 81 | 82 | 83 | 84 | 85 | 86 | 87 | 88 | 89 | 90 | 91 | 92 | 93 | 94 | 95 | 96 | 97 | 98 | 99 | 100 |

101 | 102 | 103 | 104 | 105 | 106 | 107 | 108 | 109 | 110 | 111 | 112 | 113 | 114 | 115 | 116 | 117 | 118 | 119 | 120 | 121 | 122 | 123 | 124 | 125 |

126 | 127 | 128 | 129 | 130 | 131 | 132 | 133 | 134 | 135 | 136 | 137 | 138 | 139 | 140 | 141 | 142 | 143 | 144 | 145 | 146 | 147 | 148 | 149 | 150 |

151 | 152 | 153 | 154 | 155 | 156 | 157 | 158 | 159 | 160 | 161 | 162 | 163 | 164 | 165 | 166 | 167 | 168 | 169 | 170 | 171 | 172 | 173 | 174 | 175 |

176 | 177 | 178 | 179 | 180 | 181 | 182 | 183 | 184 | 185 | 186 | 187 | 188 | 189 | 190 | 191 | 192 | 193 | 194 | 195 | 196 | 197 | 198 | 199 | 200 |

<< problem 47 - Distinct primes factors | Prime permutations - problem 49 >> |