Problem 126: Cuboid layers

(see projecteuler.net/problem=126)

The minimum number of cubes to cover every visible face on a cuboid measuring 3 x 2 x 1 is twenty-two.

cuboids

If we then add a second layer to this solid it would require forty-six cubes to cover every visible face,
the third layer would require seventy-eight cubes, and the fourth layer would require one-hundred and eighteen cubes to cover every visible face.

However, the first layer on a cuboid measuring 5 x 1 x 1 also requires twenty-two cubes; similarly the first layer on cuboids measuring 5 x 3 x 1, 7 x 2 x 1, and 11 x 1 x 1 all contain forty-six cubes.

We shall define C(n) to represent the number of cuboids that contain n cubes in one of its layers. So C(22) = 2, C(46) = 4, C(78) = 5, and C(118) = 8.

It turns out that 154 is the least value of n for which C(n) = 10.

Find the least value of n for which C(n) = 1000.

My Algorithm

The initial cuboid (without any layers) has six sides:

The total surface area is 2 * (width * height + height * depth + depth * width).
Each part of the surface is covered by the first layer. That means:
C(1) = 2 * (width * height + height * depth + depth * width).

For the 3 x 2 x 1 cuboid: C(1) = 2 * (3 * 2 + 2 * 1 + 1 * 3) = 2 * (6 + 2 + 3) = 2 * 11 = 22

The computation of the second, third, fourth, ... layer wasn't obvious to me. My drawing skills are sub-par at best.
So I decided to write some code: start with a shape and move that shape one unit left, right, up, down, forward and backward
(start each move at the initial position). The difference between the old shape and the shape created by these six movements is the volume of the new layer.
My function naive() (last lines of my code) performs exactly these steps. Starting with a specified cuboid it finds the volume of each additional layer and prints it.

Playing around with several cuboid sizes I saw a pattern:
C(2) = C(1) + 4 * (width + height + depth)
C(3) = C(2) + 4 * (width + height + depth) + 8
C(4) = C(3) + 4 * (width + height + depth) + 16
C(5) = C(4) + 4 * (width + height + depth) + 24

A bit more general:
C(1) = 2 * (width * height + height * depth + depth * width).
C(n) = C(n-1) + 4 * (width + height + depth) + 8 * (n - 2)

You can convert the second formula into a closed form but then the equation looks a bit messy.
The C++ compiler won't generate much faster code, too. The function Cuboid::calculate evaluates these two equations.

In order to solve C(n) = x for arbitrary x (it's x = 1000 in the original problem) I wrote code based on std::map:
Starting with the most basic cuboid (1,1,1) with only 1 layer I repeat the following:
Adding a cuboid is encapsulated in the add() function. It rejects cuboids where height > width and depth > height to avoid
having the same (rotated) cuboid multiple times in the container. Adding exactly the same container twice is prevented, too.
On top of that, add() keeps track of the volume of the most recent layer (in count[]).

That algorithm stops when the smallest cuboid has exactly 1000 "siblings" (→ the count of the current volume is 1000).
It shows the correct result after about 4 seconds.

However, my profiler showed that a few million cuboids where added (and removed) from the container.
The overhead of memory allocation was responsible for the majority of the execution time.

A much faster approach is to guess an upper limit of the result and then create all cuboids (and layers) that fit into that limit.
Keeping track of just the various values of C() instead of the cuboids can be done with a small std::vector.
The program becomes about 50x faster and has almost no memory consumption. See my function fast().

The main problem is to find a good guess for the upper limit.
fast() returns 0 if no result was found, that's when the initial guess has to be incremented and another call to fast() is required.
Starting with a very high guess will find the result but might waste lots of time computing cuboids with too large volumes.
Currently, I use a StepSize = 10000 ... yes, I know it's not optimal (it doesn't find the result in its first iteration)
but I deliberately keep it this way.
If StepSize = 50 then fast() isn't faster than my original code.
If StepSize = 150000 then fast() isn't faster than my original code.
That means for the original problem, fast() is only truly faster if 50 < StepSize < 150000.

Note

Even though the code of fast() is shorter (and can be faster), I have a strange feeling about it:
somehow it needs a good guess but I haven't found a way to come up with a straightforward way to generate that guess.

I was stuck for a while because initially I stopped when I found a layer volume with at least 1000 matching cuboids (instead of exactly 1000).
There are a few volumes below the correct solution with more than 1000 cuboids.

Kristian Edlund visualized a nice way to enumerate the cubes added in each layer. His blog can found in my links.

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 10 | ./126

Output:

(please click 'Go !')

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)

My code

… was written in C++11 and can be compiled with G++, Clang++, Visual C++. You can download it, too. Or just jump to my GitHub repository.

#include <iostream>
#include <vector>
#include <set>
#include <map>
 
// a cuboid with 1+ added layers
struct Cuboid
{
// cuboid size
const unsigned short width;
const unsigned short height;
const unsigned short depth;
// wrapped several times
const unsigned short layers;
// compute volume with calculated()
const unsigned int hullVolume;
 
// create new cuboid
Cuboid(unsigned short width_, unsigned short height_, unsigned short depth_, unsigned short layers_)
: width(width_), height(height_), depth(depth_), layers(layers_),
hullVolume(calculate())
{}
 
// calculate volume of the most recent / outermost layer
unsigned int calculate() const
{
// innermost hull
auto result = 2 * ((unsigned int)width * height +
(unsigned int)height * depth +
(unsigned int)depth * width);
 
// need just one layer ?
if (layers == 1)
return result;
 
// add more layers
auto increment = 4 * (width + height + depth);
for (unsigned short current = 2; current <= layers; current++)
{
result += increment;
increment += 8;
}
 
return result;
}
 
// for std::set
bool operator<(const Cuboid& other) const
{
// volume
if (hullVolume < other.hullVolume) return true;
if (hullVolume > other.hullVolume) return false;
// the following lines are arbitrary => they only need to be consistent
// sides
if (width < other.width) return true;
if (width > other.width) return false;
if (height < other.height) return true;
if (height > other.height) return false;
if (depth < other.depth) return true;
if (depth > other.depth) return false;
// layers
return layers < other.layers;
}
};
 
// list of cuboids, sorted by volume of their outermost layer
std::set<Cuboid> todo;
// count how many cuboid layers have a certain volume (looking for 1000)
std::map<unsigned int, unsigned int> count; // could be replaced by std::vector if you can estimate the upper limit
// note: this information is implicitely stored in "todo" as well: count the consecutive cuboids with the same hullVolume
 
// add a cuboid with a certain amount of layers, return number of cuboids with the same hull volume
bool add(unsigned short width, unsigned short height, unsigned short depth, unsigned short layers)
{
// accept only width >= height >= depth
if (width < height)
return false;
if (height < depth)
return false;
 
// create new cuboid
Cuboid cuboid(width, height, depth, layers);
 
// already queued ?
if (todo.count(cuboid) != 0)
return false;
 
// add cuboid
todo.insert(cuboid);
count[cuboid.hullVolume]++;
 
// successfully added
return true;
}
 
const unsigned int Failed = 0;
// find the correct result if you know an upper bound maxVolume, if not found return Failed
unsigned int fast(unsigned int stopIf, unsigned int maxVolume)
{
// allocate enough memory
std::vector<unsigned int> count(maxVolume + 1, 0);
 
// just four nested loops to generate all cuboids with hulls <= maxVolume
// ensure that x >= y >= z to avoid enumerating some volume multiple times
for (unsigned short x = 1; Cuboid(x,1,1,1).hullVolume <= maxVolume; x++)
for (unsigned short y = 1; y <= x && Cuboid(x,y,1,1).hullVolume <= maxVolume; y++)
for (unsigned short z = 1; z <= y && Cuboid(x,y,z,1).hullVolume <= maxVolume; z++)
for (unsigned short layer = 1; ; layer++)
{
auto current = Cuboid(x,y,z,layer).hullVolume;
if (current > maxVolume)
break;
 
count[current]++;
}
 
// find smallest hull volume that appears exactly 1000 times
for (size_t i = 0; i < count.size(); i++)
if (count[i] == stopIf)
return i;
 
// failed
return Failed;
}
 
// forward declaration of my old helper code
//unsigned int naive(unsigned int x, unsigned int y, unsigned int z, unsigned int layers);
 
int main()
{
// start old code to analyze a single cuboid
//naive(3,2,1, 5); return 0;
 
// look for that number of cuboids with the same volume
unsigned int stopIf = 1000;
std::cin >> stopIf;
 
// ---------- fast enumeration ----------
 
// find solution if an upper limit of the result is known
#define FAST_ENUMERATION
#ifdef FAST_ENUMERATION
const unsigned int StepSize = 10000; // I deliberately choose not the optimal value !
// assume that the result is below 10000, then 20000, ... until fast() returns > 0
for (unsigned int maxVolume = StepSize; ; maxVolume += StepSize)
{
auto current = fast(stopIf, maxVolume); // returns 0 if not found
// found ?
if (current != Failed)
{
std::cout << current << std::endl;
return 0;
}
}
#endif
 
// ---------- slower enumeration ----------
 
// note: I found the correct result with this approach but later created the faster enumeration you find above
 
// initial cuboid: one unit wide, high and deep with one layer around it
add(1,1,1,1);
unsigned int volume = 0;
while (count[volume] != stopIf)
{
// pick cuboid with smallest volume and remove it
auto current = *(todo.begin());
todo.erase(todo.begin());
 
volume = current.hullVolume;
 
// enlarge cuboid
add(current.width + 1, current.height, current.depth, current.layers);
add(current.width, current.height + 1, current.depth, current.layers);
add(current.width, current.height, current.depth + 1, current.layers);
// add one layer
add(current.width, current.height, current.depth, current.layers + 1);
}
 
// display result
std::cout << volume << std::endl;
return 0;
}
 
// ---------- compute hull volume, slow approach ----------
 
// a single cube that is part of an cuboid (more or less a coordinate)
struct Cube
{
// coordinates
const int x, y, z;
 
// create new cube
Cube(int xx, int yy, int zz)
: x(xx), y(yy), z(zz)
{}
 
// for std::set
bool operator<(const Cube& other) const
{
if (x < other.x) return true;
if (x > other.x) return false;
if (y < other.y) return true;
if (y > other.y) return false;
return z < other.z;
}
};
 
unsigned int naive(unsigned int x, unsigned int y, unsigned int z, unsigned int layers)
{
// keep track of all cubes belonging to the shape
std::set<Cube> cuboid;
// cubes added in the most recent layer
std::vector<Cube> hull;
 
// initial shape
for (int xx = 0; xx < (int)x; xx++)
for (int yy = 0; yy < (int)y; yy++)
for (int zz = 0; zz < (int)z; zz++)
{
Cube current(xx, yy, zz);
cuboid.insert(current);
hull.push_back(current);
}
 
// add desired numbers of layers
for (unsigned int layer = 1; layer <= layers; layer++)
{
std::vector<Cube> next;
for (auto current : hull)
{
// grow in each direction x,y,z
Cube left (current.x - 1, current.y, current.z);
Cube right(current.x + 1, current.y, current.z);
Cube down (current.x, current.y - 1, current.z);
Cube up (current.x, current.y + 1, current.z);
Cube back (current.x, current.y, current.z - 1);
Cube front(current.x, current.y, current.z + 1);
 
// add cube (if not already existing)
if (cuboid.count(left) == 0)
{
cuboid.insert(left);
next.push_back(left);
}
if (cuboid.count(right) == 0)
{
cuboid.insert(right);
next.push_back(right);
}
if (cuboid.count(down) == 0)
{
cuboid.insert(down);
next.push_back(down);
}
if (cuboid.count(up) == 0)
{
cuboid.insert(up);
next.push_back(up);
}
if (cuboid.count(back) == 0)
{
cuboid.insert(back);
next.push_back(back);
}
if (cuboid.count(front) == 0)
{
cuboid.insert(front);
next.push_back(front);
}
}
 
std::cout << "layer " << layer << ": "
<< "C(" << x << "," << y << "," << z << ")=" << next.size();
if (layer > 1)
std::cout << " delta=" << next.size() - hull.size();
std::cout << std::endl;
 
// next iteration
hull = std::move(next);
}
 
return hull.size();
}

This solution contains 40 empty lines, 57 comments and 7 preprocessor commands.

Benchmark

The correct solution to the original Project Euler problem was found in 0.04 seconds on an Intel® Core™ i7-2600K CPU @ 3.40GHz.
(compiled for x86_64 / Linux, GCC flags: -O3 -march=native -fno-exceptions -fno-rtti -std=gnu++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 18, 2017 submitted solution
August 18, 2017 added comments

Difficulty

55% Project Euler ranks this problem at 55% (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
black problems are solved but access to the solution is blocked for a few days until the next problem is published
[new] the flashing problem is the one I solved most recently

I stopped working on Project Euler problems around the time they released 617.
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
576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600
601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625
626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650
651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675
676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700
701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725
726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750
751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775
776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800
801 802 803 804 805 806
The 310 solved problems (that's level 12) had an average difficulty of 32.6% at Project Euler and
I scored 13526 points (out of 15700 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 !