Choco Performance Report

Benchmark Description

The benchmarck is done on a Pentium 4 3Gh, 512 Mo with the sun jdk1.4.2_03 in server mode (-server). The stack size used is the default jvm size (no -X options).

Note
By default, the mindom heuristic for choosing variable and the increasing domain heuristic for choosing value are used. Ties between variables are broken with the order where variables were created (The first minimum domain variable created is selected).

N Queens

The problem is modeled with N integer variables of domain 1 .. N corresponding to each line of the board, and only inequality constraints are used.

Magic Square

The problem is modeled with N * N integer variables of domain 1 .. N * N corresponding to each cell of the square, and linear combinations are posted for each line, each column and the two diagonals (the magic number is given a priori).

Golomb Rules

N variable for all distances (all pairs of mark) are used. Differences between all distances are stated with a BoundAllDiff constraint. Moreover, a global redondant constraint is added. The variable selection is done in the order of distances (from 0 to 1,2,3, then from 1 to 2,3,4, etc ...) : the first non instantiated distance is choosen.

Latin Square

The problem is modeled with N * N integer variables of domain 1 .. N corresponding to each cell of the square, and uses simple inequalities between variables of the same line and column.

Last Version Performance (in ms)

Performance report on version '1_0b001'
Problems MAC CBJ MACDBT DR
MagicSquare 4 all 45641 241781 N/A N/A
MagicSquare 5 one 688 1656 16578 6281
NQueens 12 all 11532 52235 427922 N/A
NQueens 13 all 59688 298015 N/A N/A
NQueens 14 all 352704 N/A N/A N/A
GolombProblem 11 one 101078 N/A N/A N/A
GolombProblem 12 one 970484 N/A N/A N/A
LatinSquare 11 one N/A 47 78 78

Performance evolution

MACDBT

MACDBT

DR

DR

CBJ

CBJ

MAC

MAC