[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Help-glpk] Multithreading/parallelization

From: Harley Mackenzie
Subject: Re: [Help-glpk] Multithreading/parallelization
Date: Tue, 18 Dec 2012 21:31:00 +1100


I have nearly got this working. I used the get.glpk script in the third
party directory and then used your options below to compile the CBC
program on an Ubuntu 64 bit machine.

I used this compilation command:

  ./configure CFLAGS='-m64 -O3' CC=gcc-4.7 --enable-gnu-packages
  --enable-cbc-parallel --enable-debug -C --prefix=/usr/local 
  sudo make install

I attempted to use the following command line with the compiled CBC

  cbc braemar_prices_dop.mod -randomCbcSeed 0 -randomSeed 0 -proximity
  on -threads 8 -printi csv -solve -solu braemar_prices_dop.csv

  command line - cbc braemar_prices_dop.mod -randomCbcSeed 0 -randomSeed
  0 -proximity on -threads 8 -printi csv -solve -solu
  braemar_prices_dop.csv (default strategy 1)
  At line 1 param n_revenue >= 0;
  Unknown image param n_revenue >= 0; at line 1 of file
  Coin0008I  read with -2 errors
  There were -2 errors on input
  randomCbcSeed was changed from -1 to 0
  using time of day randomSeed was changed from 1234567 to 1355826052
  Option for proximitySearch changed from off to on
  threads was changed from 0 to 8
  Option for printingOptions changed from normal to csv
  ** Current model not valid
  ** Current model not valid
  No match for braemar_prices_dop.csv - ? for list of commands
  Total time (CPU seconds):       0.00   (Wallclock seconds):       0.00

The first line is a standard mathprog command:

  param n_revenue >= 0;

So it would seem that the mathprog interface is not being used for the
model as this is pretty straight forward mathprog line and the whole
model runs on GLPK. The mod file has the structure:


All in the one file called braemar_prices_dop.mod.

There wasnt any command in the ./configure step to include glpk - is
anything required? Or is there a command line option needed for CBC to
use mathprog to process the model?

Any help or directions to any documentation would be greatly
appreciated. How did you work out how to set up CBC for mathprog?



On Tue, Dec 18, 2012, at 8:21, Noli Sicad wrote:
> Hi Harley,
> > I assume that you are using mathprog and then producing an MPS file for
> > CBC, and as you point out, you can't then use the mathprog output
> > statements and would have to process the results some other way.
> No, I run my MIP MathPog model in CBC solver. I compiled CBC with
> MathProg parser in Mac OS X. CBC has third parties "add-ons" and
> GLPK/MahtProg is one of these add-ons packages. Before you compile CBC
> get glpk first, if you like to include support of MathProg parser.
> CBC with MathProg parser support just ignores those MathProg scripts
> after the solve statement.
> Yes, you can use MPS and CPLEX LP format as input for the CBC solver.
> > Are you able to share the details on how you make this work? I presume
> > CBC is the COIN-OR MIP solver. How is easy is CBC to work with and how
> > do you handle the output processing?
> First download the CBC trunk. Yes, it is the COIN-OR MIP solver.
> CBC trunk has threads support with proximity search heuristic
> algorithm. Before compiling run the "get.glpk" script in the third
> parties folder > glpk folder. The default glpk is glpk-4.45, just use
> it. It does not interfere with your installed glpk (glpk-4.47).
> Compile with  --enable-cbc-parallel
> For linux and Mac OS X.
> #####
> ./configure CFLAGS='-m64 -O3' CC=gcc-4.7 --enable-gnu-packages
> --enable-cbc-parallel --enable-debug -C --prefix=/usr/local
> #####
> For Windows OS, it has support for various version of MSVisualStudio.
> Have a look at the MSVisualStudio folder in CBC after you download.
> Run the CBC solver with -print csv to get the solution.
> e.g.
> #####
> > cbc myMIP.mod -randomCbcSeed 0 -randomSeed 0 -proximity on -threads 8 
> > -printi csv -solve -solu solution_math_period1-10p1.csv
> #####
> Noli

  Dr. Harley Mackenzie
  HARD software

  + 61 3 5222 3435

reply via email to

[Prev in Thread] Current Thread [Next in Thread]