Building a model with more than 2 billion nonzeros: GRBnewmodel vs GRBXaddvars

回答済み

jake roth

2023年03月15日 16:29 編集済み

• 
• 
• 

I am building an LP model with the C API that I know has more than 2 billion nonzeros, and I am wondering whether the best practice is to (1) initialize all of the variables and bounds with `GRBnewmodel` or to (2) initialize an empty model with `GRBnewmodel` and then add variables with `GRBXaddvars`. Or should upper and lower bounds be added with `GRBXaddrangeconstrs`?

関連

• C
• Memory

0

• 

  Silke Horn

  • Gurobi Staff

  2023年03月20日 08:52

  You can either

  • initialize all variables and bounds at the beginning in GRBnewmodel and then add the constraints using GRBXaddconstrs or
  • create an empty model in GRBnewmodel and then add the variables using GRBXaddvars.

  I would not expect a massive performance difference between the two approaches.

  In any case, I would recommend defining the variable bounds when adding the variables and not as range constraints.

  1
• 

  jake roth

  2023年03月20日 16:47

  • 
  • 
  • 

  Thanks!

  0

サインインしてコメントを残してください。