Add automated testing of basic cases

[ppebay]

To start, use the same cases as the one that are tested in LBAF, but now using the MILP configuration. Extract the optimum and verify we obtain the same results.

[tlamonthezie]

@maxime-bfsquall please find here some precisions: We have to split ccm_milp into 2 scripts:

• First script: generate the problem (remove problem solving)
• Second script: solve the problem using a solver Solving is done using integrated pulp CBC solver. It might be better to setup the solvers to test and configure solvers in pulp

GLPK Solver: https://www.gnu.org/software/glpk/ CBC Solver: https://github.com/coin-or/Cbc

Problem solving should generate the solution file (*.sol) and CI must test that the solution file contains exactly 2.0 as the objective value

GLPK can be installed on Debian using apt install glpk-utils CBC can be installed (independent from pulp) on Debian using sudo apt install coinor-cbc

[maxime-bfsquall]

@ppebay I had a problem with the solver GLPK. The names of the variables used in the problem file are special characters and GLPK solver doesn't accept that. I had to change them to be able to try the solver.

• χ => 'CHI'
• φ => PHI
• ψ => 'PSI'

Is it a problem if I let them like that or should I do some research to see if it is possible to use them as special characters with the GLPK solver ?

[ppebay]

@maxime-bfsquall no problem with changing from Greek letters to their Latin transcriptions.

One question however: why use ALL-CAPS instead of regularly-capitalized letters, e.g., Chi for χ, etc.?

[maxime-bfsquall]

@ppebay no particular reason, I can use capitalized letters.

[ppebay]

Sorry @maxime-bfsquall, my mistake, I verified in our source paper for this work, and it turns out that we use all-lowercase Latin transliteration:

So χ => 'chi', etc.

[ppebay]

@maxime-bfsquall please add the two following cases:

• load + NO memory + homing (delta=0.1) Result: 2.5

• load + NO memory + homing (delta=0.3) Result: 4.0

[maxime-bfsquall]

@ppebay

• for is_fwmp: true, delta: 0.1 : load + NO memory + homing (delta=0.1) The result is good 2.5

• for is_fwmp: true, delta: 0.3 : load + NO memory + homing (delta=0.3) The result is 3.5 : So, not good if we expected 4.0

[ppebay]

@maxime-bfsquall I just ran again the test, and I confirm that I the expected result should be 4.0:

# Available examples:
  0) small.SmallProblem
  1) synthetic_blocks.SyntheticBlocks
  2) ccm_example_no_sub_cluster.CCMExampleNoSubCluster
  3) ccm_example_with_sub_cluster.CCMExampleWithSubCluster
  4) illustration_1.Illustration1
  5) illustration_2.Illustration2
  example index: 1

# Model configuration:
  FWMP [y/N]? y
  value of alpha [0.0]? 1
  value of beta [0.0]? 
  value of gamma [0.0]? 
  value of delta [0.0]? 0.3
  bounded memory [y/N]? 
  preserve clusters [y/N]? y

# Initializing FMWP configuration with:
  alpha = 1.0
  beta = 0.0
  gamma = 0.0
  delta = 0.3
  without rank memory upper bound
  while preserving block clusters

we then obtain:

# Solution summary:
Rank 0: L = 2.5, W = 2.5, unhomed: 0
Rank 1: L = 4.0, W = 4.0, unhomed: 0
Rank 2: L = 1.5, W = 1.5, unhomed: 0
Rank 3: L = 0.0, W = 0.0, unhomed: 0
w_max = 4.0
$assignments=[0, 0, 0, 0, 1, 1, 1, 1, 2];

Can you please share your own output for this case @maxime-bfsquall ?

[maxime-bfsquall]

My apologies, I had not set preserve_clusters at True. It's done and OK now