Problem 401: Sum of squares of divisors

(see projecteuler.net/problem=401)

The divisors of 6 are 1,2,3 and 6.
The sum of the squares of these numbers is 1+4+9+36=50.

Let sigma2(n) represent the sum of the squares of the divisors of n. Thus sigma2(6)=50.
Let SIGMA2 represent the summatory function of sigma2, that is SIGMA2(n)=sum{sigma2(i)} for i=1 to n.
The first 6 values of SIGMA2 are: 1,6,16,37,63 and 113.

Find SIGMA2(10^15) mod 10^9.

My Algorithm

Any number i has \lfloor frac{n}{i} \rfloor multiples (integer division).
When calculating SIGMA2(6):
- 1 has \lfloor frac{6}{1} \rfloor = 6 multiples, add 6 * 1^2 = 6 * 1 = 6
- 2 has \lfloor frac{6}{2} \rfloor = 3 multiples, add 3 * 2^2 = 3 * 4 = 24
- 3 has \lfloor frac{6}{3} \rfloor = 2 multiples, add 2 * 3^2 = 2 * 9 = 18
- 4 has \lfloor frac{6}{4} \rfloor = 1 multiples, add 1 * 4^2 = 1 * 16 = 16
- 5 has \lfloor frac{6}{5} \rfloor = 1 multiples, add 1 * 5^2 = 1 * 25 = 25
- 6 has \lfloor frac{6}{6} \rfloor = 1 multiples, add 1 * 6^2 = 1 * 36 = 36
SIGMA2(6) = 6 + 24 + 18 + 16 + 25 + 36 = 113

Unfortunately this requires n iterations, which is no problem for n = 6 but next to impossible for n = 15.
Looking closer at the example reveals that half of the divisors appears only once (4, 5 and 6).
In general, a few number appear very often and the majority just a few times.
Because i appears \lfloor frac{n}{i} \rfloor times, i <= \lfloor frac{n}{i} \rfloor if i <= sqrt{n}.
If i > sqrt{n} then several squares appear the same time. As mentioned before, there are three squares that appear once.

My program applies the basic algorithm for all numbers up to sqrt{n}, which needs sqrt{10^15} approx 3.2 * 10^7 iterations.
Then it determines which squares appear once, twice, etc. going backwards from 10^15 to sqrt{n}.
This can be quite efficient because the sum of a series of squares is (see en.wikipedia.org/wiki/Square pyramidal number):
P(n) = sum_{k=1...n}{k^2} = dfrac{n (n + 1) (2n + 1)}{6}

The largest square pyramidal number that appears exactly m times is \lfloor frac{n}{m} \rfloor.
To obtain the sum of all square that appear m times I calculate P(\lfloor frac{n}{m} \rfloor) - P(\lfloor frac{n}{m+1} \rfloor).

pyramidal(n) returns the n-th square pyramidal number (mod 10^9). This function is rather tricky because I hit the limits of C++'s 64 bit integers.
Luckily, G++ has routines for 128 bit integers to work around this issue - but therefore this code doesn't compile with Visual C++.

The code is scattered with mod statement. Maybe a few are redundant.

Note

Apparently you can implement pyramidal without __int128.

Interactive test

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

Input data (separated by spaces or newlines):

This is equivalent to
echo 6 | ./401

Output:

(please click 'Go !')

Note: the original problem's input 1000000000000000 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)

My code

… was written in C++11 and can be compiled with G++, Clang++. You can download it, too.

#include <iostream>
#include <cmath>
 
// return sum of 1^2 + 2^2 + 3^2 + 4^2 + ... + n^2 mod modulo
// note: needs GCC's 128 bit integers (not available on Visual C++)
unsigned int pyramidal(unsigned long long n, unsigned int modulo)
{
// inspired by https://stackoverflow.com/questions/39826483/square-pyramidal-number-of-int32-modulo-some-m-using-only-int64-intermediates
unsigned __int128 nn = n;
// solve n * (n + 1) * (2n + 1) / 6
// either nn or nn+1 is even => their product must be divisible by 2
auto x = nn * (nn + 1) / 2;
auto y = 2 * n + 1;
// now equation becomes x * y / 3 (mod modulo)
 
// is x a multiple of 3 ?
auto div3 = x / 3;
if (div3 * 3 == x)
return ((div3 % modulo) * y) % modulo;
 
// no, y must be a multiple of 3
return (x % modulo) * (y / 3) % modulo;
}
 
// count all divisors
unsigned int solve(unsigned long long n, unsigned int modulo)
{
unsigned long long sum = 0;
 
// for each number i there is a second divisor n/i (integer division)
// for i < sqrt(n) those second divisors are "unique"
// for i > sqrt(n) several i share the same second divisor n/i
unsigned int threshold = sqrt(n);
 
// look at small divisors
for (unsigned long long i = 1; i <= threshold; i++)
{
// each divisor-th number is a multiple of divisor
auto count = (n / i) % modulo;
// square (reduce with modulo)
auto square = (i * i) % modulo;
 
// add
sum += (count * square) % modulo;
}
 
// apply the same logic in reverse order:
// for each divisor i there is a divisor j = n / i
for (unsigned int j = 1; j <= n / (threshold + 1); j++)
{
auto from = n / j;
auto to = n / (j + 1);
std::cout << from << " " << to << std::endl;
 
unsigned long long sumOfConsecutiveSquares = 0;
// note: from > to, therefore decrement instead of increment
/*for (unsigned long long k = from; k > to; k--)
{
auto square = (k * k) % modulo;
sumOfConsecutiveSquares += square;
sumOfConsecutiveSquares %= modulo;
}*/
 
// same as the for-loop above
auto squareSumFrom = pyramidal(from, modulo); // sum of squares <= from
auto squareSumTo = pyramidal(to, modulo); // sum of squares <= to
// add modulo to prevent negative numbers
if (squareSumFrom < squareSumTo)
squareSumFrom += modulo;
sumOfConsecutiveSquares = squareSumFrom - squareSumTo;
 
// those squares can be found "j" times
sum += (j * sumOfConsecutiveSquares) % modulo;
}
 
// and again, take care of modulo
return sum % modulo;
}
 
int main()
{
// 10^9
const auto Modulo = 1000000000;
// 10^15
auto n = 1000000000000000ULL;
std::cin >> n;
std::cout << solve(n, Modulo) << std::endl;
return 0;
}

This solution contains 13 empty lines, 25 comments and 2 preprocessor commands.

Benchmark

The correct solution to the original Project Euler problem was found in 3.8 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

August 22, 2017 submitted solution
August 22, 2017 added comments

Difficulty

25% Project Euler ranks this problem at 25% (out of 100%).

Heatmap

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

green   solutions solve the original Project Euler problem and have a perfect score of 100% at Hackerrank, too
yellow solutions score less than 100% at Hackerrank (but still solve the original problem easily)
gray problems are already solved but I haven't published my solution yet
blue solutions are relevant for Project Euler only: there wasn't a Hackerrank version of it (at the time I solved it) or it differed too much
orange problems are solved but exceed the time limit of one minute or the memory limit of 256 MByte
red problems are not solved yet but I wrote a simulation to approximate the result or verified at least the given example - usually I sketched a few ideas, too

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
201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225
226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250
251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275
276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300
301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325
326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350
351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375
376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400
401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425
426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450
451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475
476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500
501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525
526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550
551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575
The 270 solved problems (level 10) had an average difficulty of 31.3% at Project Euler and
I scored 13,386 points (out of 15600 possible points, top rank was 17 out of ≈60000 in August 2017) at Hackerrank's Project Euler+.

My username at Project Euler is stephanbrumme while it's stbrumme at Hackerrank.

Look at my progress and performance pages to get more details.

more about me can be found on my homepage, especially in my coding blog.
some names mentioned on this site may be trademarks of their respective owners.
thanks to the KaTeX team for their great typesetting library !