Survey of Paint-by-Number Puzzle Solvers

Old Results

Jan Wolter

1. Introduction

Up until the November 2009 version of my Survey of Paint-by-Number Puzzle Solvers all run time tests were done on an older model of computer with less memory. In December of 2009 I replaced that computer with a newer one and re-ran all tests on the new machine. This document preserves the older results.

The specifications of this older computer are as follows:

Computer A
CPU:2.4GHz AMD Athlon 64 4000+ single processor
Memory:1 Gb
OS:OpenSuse 10.3
Compiler:gcc 4.2
Thus it had a single core CPU instead of a quad core, and each of the cores was somewhat slower. It had substantially less memory and and older version of OpenSuse with an older compiler.

This older compiler version had a bug which caused it to fail if Simpson's solver was compiled with the -O2 flag. Compiler optimization makes a huge difference to all of these solvers. They run perhaps two or three times slower without it. So to be fair, all solvers were built without optimization, if possible. Otherwise, all tests were done the same way as in the newer results.

1. Sample Results

COMPUTER A: Single Processor, 1Gb memory, compiler optimization disabled
puzzle size Unique
Solution?		 Line
Solvable?		 Blank
Lines?		 Wolter Olšák Simp-
son		 Lager-
kvist/
Gecode		 Kjeller-
strand/
Gecode		 Kjeller-
strand/
Lazyfd		 Wilk Ishi-
gaki		 Ware-
ham		 Bosch/
glpk
#1: Dancer * 5 x 10 Yes Yes No 0 0 0 0.02s 0.03s 0.07s 0 0.04s 0.07s 0.01s
#6: Cat * 20 x 20 Yes Yes No 0 0 0 0.02s 0.04s 0.81s 0 0.35s 0.18s 3.5s
#21: Skid * 14 x 25 Yes Yes Yes 0 0 0 0.02s 0.04s 1.1s 0 Fail X X
#27: Bucks * 27 x 23 Yes No Yes 0 0 0 0.02s 0.05s 2.0s 0 Fail X X
#23: Edge * 10 x 11 Yes No No 0 0 0 0.03s 0.04s 0.12s 0 Fail Fail 0.24s
#2413: Smoke 20 x 20 Yes No No 0 0 0 0.03s 0.04s 1.1s 0 1.6s + 12.2m
#16: Knot * 34 x 34 Yes Yes No 0 0 0 0.04s 0.07s 10.1s 0.02s 25.0s + Fail
#529: Swing * 45 x 45 Yes Yes No 0 0 0.01 0.07s 0.12s 23.8s 0.11s 5.7s - -
#65: Mum * 34 x 40 Yes No No 0.02s 0 0.05s 0.10s 0.15s 19.9s 1.18s + - -
#1694: Tragic 45 x 50 Yes No No 0.18s 0.09s 6.2s 0.60s 18.0m 59.6s 2.6s + - -
#1611: Merka 55 x 60 Yes No Yes 0.03s 0.01s 0.04s 0.13s + 1.9m 20.7m + X X
#436: Petro * 40 x 35 Yes No No 0.38s 40.1s 0.54s 6.4m 7.3s 11.5s + + - -
#4645: M&M 50 x 70 Yes No Yes 0.38s 0.28s 0.12s 4.4s 1.3s 84.7s + - X X
#3541: Signed 60 x 50 Yes No No 0.16s 3.1s + 2.4s 2.9s 1.8m + - - -
#803: Light 50 x 45 Yes No No 3.9s + 0.11s + + 9.1s + - - -
#6574: Forever * 25 x 25 Yes No No + 4.9s 90.1s 20.8s 12.0s 3.3s + - - +
#2040: Hot 55 x 60 Yes No No 4.6s + 7.5m + + 4.5m + - - -
#6739: Karate 40 x 40 No No No 26.3s 42.1s + 3.9m 5.0m 25.6s + - - -
#2556: Flag 65 x 45 No No Yes + 4.1s 0 11.5s + 7.7s + - X X
#2712: Lion 47 x 47 No No No 6.7m + + + + + + - - -

A still older version of this table including obsolete pbnsolve versions is available.

3. Full Results

2,491 Black & White Puzzles on Computer A
Solver 0 to
0.09 sec		 0.10 to
0.49 sec		 0.50 to
1.99 sec		 2.00 to
9.99 sec		 10.00 to
29.99 sec		 30 sec to
2 min		 2 min
or more
Wolter 2402 64 16 5 1 0 3
Simpson 2409 30 14 12 4 3 17
Olšák 2402 26 13 20 3 8 19
Kjellerstrand/Gecode 1755 571 51 27 16 12 59
Wilk 1979 174 105 85 34 24 90

3,232 Multicolor Puzzles on Computer A
Solver 0 to
0.09 sec		 0.10 to
0.49 sec		 0.50 to
1.99 sec		 2.00 to
9.99 sec		 10.00 to
29.99 sec		 30 sec to
2 min		 2 min
or more
Wolter 3168 41 10 3 0 2 8
Olšák 3168 19 8 8 8 2 19
Last Update: Fri Dec 4 14:55:31 PST 2009