Benchmarks

This section presents computational experiments on the creation of Integer Programming models using different mathematical modelling packages. Gurobi is the default Gurobi ® Python interface, which currently only supports the Python interpreter CPython. Pulp supports both CPython and also the just-in-time compiler Pypy. MIP also suports both. JuMP [DHL17I] is the Mathematical Programming package of the Julia programming language. Both Jump and Pulp use intermediate data structures to store the mathematical programming model before flushing it to the solver, so that the selected solver does not impacts on the model creation times. MIP does not stores the model itself, directly calling problem creation/modification routines on the solver engine.

Since MIP communicates every problem modification directly to the solver engine, the engine must handle efficiently many small modification request to avoid potentially expensive resize/move operations in the constraint matrix. Gurobi automatically buffers problem modification requests and has an update method to flush these request. CBC did not had an equivalent mechanism, so we implemented an automatic buffering/flushing mechanism in the CBC C Interface. Our computational results already consider this updated CBC version.

Computational experiments executed on a ThinkPad ® X1 notebook with an Intel ® Core™ i7-7600U processor and 8 Gb of RAM using the Linux operating system. The following software releases were used: CPython 3.7.3, Pypy 7.1.1, Julia 1.1.1, JuMP 0.19 and Gurobi 8.1 and CBC svn 2563.

n-Queens

These are binary programming models. The largest model has 1 million variables and roughly 6000 constraints and 4 million of non-zero entries in the constraint matrix.

\(n\)

Gurobi

CPython

Pulp

Python-MIP

JuMP

Gurobi

CBC

CPython

Pypy

CPython

Pypy

CPython

Pypy

100

0.24

0.27

0.18

0.65

0.97

0.30

0.45

2.38

200

1.43

1.73

0.43

1.60

0.18

1.47

0.19

0.25

300

4.97

5.80

0.92

5.48

0.37

5.12

0.38

0.47

400

12.37

14.29

1.55

13.58

0.72

13.24

0.74

1.11

500

24.70

32.62

2.62

27.25

1.25

26.30

1.23

2.09

600

43.88

49.92

4.10

47.75

2.02

46.23

1.99

2.86

700

69.77

79.57

5.84

75.97

3.04

74.47

2.94

3.64

800

105.04

119.07

8.19

114.86

4.33

112.10

4.26

5.58

900

150.89

169.92

10.84

163.36

5.95

160.67

5.83

8.08

1000

206.63

232.32

14.26

220.56

8.02

222.09

7.76

10.02