[Cbc] Parallell CBC

Thu Jul 14 06:12:08 EDT 2016

It seems the binary provided by AMPL works in parallel:
http://ampl.com/dl/open/cbc/cbc-win64.zip

That one is at version 2.9.7. Otherwise, it is easy to build a Cbc binary
on Windows using MinGW by modifying the instructions here:

https://projects.coin-or.org/SYMPHONY/wiki/BuildAndInstall#BUILDINGWITHMSys2orCYGWIN

The instructions are for SYMPHONY, but it should be clear how to modify
them for Cbc. You also need to add the configuration option
--enable-parallel, as Haroldo mentioned.

By the way, I also got an automated build and test setup for Windows
working via Appveyor a while back and it should be relatively easy to
upload the built binaries from there to somewhere you can download:

https://ci.appveyor.com/project/tkralphs/cbc

I'm thinking about how to make that work. I'm also working on a docker
image with pre-built binaries, which will be pushed soon.

Cheers,

Ted

On Wed, Jul 13, 2016 at 4:45 PM, Bjørn Sigurd Johansen (Spider Solutions
AS) <bjorn.sigurd.johansen at spidersolutions.no> wrote:

> Thank you, but the build from https://projects.coin-or.org/Cmpl  =>
> http://www.coliop.org/download.html  seems to be even older:
>
> C:\Users\bjorn_000>C:\tmp\cbc.exe
> Welcome to the CBC MILP Solver
> Version: Trunk (unstable)
> Build Date: Mar  9 2015
> Revision Number: 2156
>
> Any other suggestions on how to get 2.9.8 cbc.exe for Win64 ?
>
>
> -----Original Message-----
> From: Haroldo Gambini Santos [mailto:haroldo.santos at gmail.com]
> Sent: 13. juli 2016 16:06
> To: Bjørn Sigurd Johansen (Spider Solutions AS) <
> bjorn.sigurd.johansen at spidersolutions.no>; cbc at list.coin-or.org
> Subject: Re: [Cbc] Parallell CBC
>
> Hi,
>
> AFAIK the visual studio build does not includes pthread and is not
> parallel.
>
> It is possible to build in windows a parallel version using MINGW
> (http://www.mingw.org/) by adding --enable-cbc-parallel .
>
> I strongly suggest that you update to version 2.9.8, many bugs were solved.
>
> I think that CMPL provides the 2.9.8 CBC version for windows bundled.
> https://projects.coin-or.org/Cmpl
>
>
> Cheers,
>
>
>
> Em 13-07-2016 03:54, Bjørn Sigurd Johansen (Spider Solutions AS) escreveu:
> > Dear CBC,
> >
> >
> >
> > I am using CBC 2.9.4 for MILP problem, but is not able to run in
> > parallel.  Using the option "-threads <some number>" makes CBC hang
> > after a few seconds.
> >
> > Is there a new Win64 cbc.exe that properly handles the "-threads"
> > option?  (And possibly with improved MILP performance?)
> >
> >
> >
> >
> >
> > *CBC stops after the following print:*
> >
> > C:\Users\bjorn_000>C:\Users\bjorn_000\AppData\Roaming\Microsoft\AddIns
> > \Solvers\win64\cbc.exe -threads 2 -directory
> > "C:\Dropbox\kh\Optimize\ModelsAndSolutions"
> > -import "C:\Dropbox\kh\Optimize\ModelsAndSolutions\ModelForGurobi.lp"
> > -seconds 20 -maxIterations 50000 -ratioGap 0 -solve -solution
> > "C:\Dropbox\kh\Optimize\ModelsAndSolutions\Solution.txt"
> >
> > Welcome to the CBC MILP Solver
> >
> > Version: 2.9.4
> >
> > Build Date: May 18 2015
> >
> >
> >
> > command line -
> > C:\Users\bjorn_000\AppData\Roaming\Microsoft\AddIns\Solvers\win64\cbc.
> > exe -threads
> > 2 -directory C:\Dropbox\kh\Optimize\ModelsAndSolutions -import
> > C:\Dropbox\kh\Optimize\ModelsAndSolutions\ModelForGurobi.lp -seconds
> > 20 -maxIterations 50000 -ratioGap 0 -solve -solution
> > C:\Dropbox\kh\Optimize\ModelsAndSolutions\Solution.txt (default
> > strategy 1)
> >
> > threads was changed from 0 to 2
> >
> > seconds was changed from 1e+100 to 20
> >
> > maxIterations was changed from 2147483647 to 50000
> >
> > ratioGap was changed from 0 to 0
> >
> > Continuous objective value is 178.167 - 0.09 seconds
> >
> > Cgl0003I 0 fixed, 0 tightened bounds, 1 strengthened rows, 5
> > substitutions
> >
> > Cgl0003I 0 fixed, 0 tightened bounds, 1 strengthened rows, 0
> > substitutions
> >
> > Cgl0003I 0 fixed, 0 tightened bounds, 1 strengthened rows, 0
> > substitutions
> >
> > Cgl0003I 0 fixed, 0 tightened bounds, 1 strengthened rows, 0
> > substitutions
> >
> > Cgl0003I 0 fixed, 0 tightened bounds, 1 strengthened rows, 0
> > substitutions
> >
> > Cgl0003I 0 fixed, 0 tightened bounds, 1 strengthened rows, 0
> > substitutions
> >
> > Cgl0003I 0 fixed, 0 tightened bounds, 1 strengthened rows, 0
> > substitutions
> >
> > Cgl0003I 0 fixed, 0 tightened bounds, 1 strengthened rows, 0
> > substitutions
> >
> > Cgl0003I 0 fixed, 0 tightened bounds, 1 strengthened rows, 0
> > substitutions
> >
> > Cgl0004I processed model has 2400 rows, 2976 columns (2973 integer
> > (1497 of which binary)) and 43145 elements
> >
> > Cbc0038I Initial state - 16 integers unsatisfied sum - 3.95151
> >
> > Cbc0038I Pass   1: suminf.    1.27858 (7) obj. 786.122 iterations 1458
> >
> > Cbc0038I Pass   2: suminf.    0.82530 (3) obj. 1498.29 iterations 288
> >
> > Cbc0038I Solution found of 1498.29
> >
> > Cbc0038I Branch and bound needed to clear up 3 general integers
> >
> > Cbc0038I Full problem 2400 rows 2976 columns, reduced to 19 rows 17
> > columns
> >
> > Cbc0038I Cleaned solution of 1092.61
> >
> > Cbc0038I Before mini branch and bound, 2948 integers at bound fixed
> > and
> > 1 continuous of which 1 were internal integer and 0 internal
> > continuous
> >
> > Cbc0038I Full problem 2400 rows 2976 columns, reduced to 32 rows 24
> > columns
> >
> > Cbc0038I Mini branch and bound did not improve solution (1.10 seconds)
> >
> > Cbc0038I Freeing continuous variables gives a solution of 1092.61
> >
> > Cbc0038I Round again with cutoff of 1001.22
> >
> > Cbc0038I Reduced cost fixing fixed 31 variables on major pass 2
> >
> > Cbc0038I Pass   3: suminf.    1.27858 (7) obj. 786.122 iterations 0
> >
> > Cbc0038I Pass   4: suminf.    1.24256 (7) obj. 1001.22 iterations 961
> >
> > Cbc0038I Pass   5: suminf.    0.73804 (7) obj. 1001.22 iterations 5
> >
> > Cbc0038I Pass   6: suminf.    1.20089 (5) obj. 1001.22 iterations 1597
> >
> > Cbc0038I Pass   7: suminf.    1.24057 (5) obj. 1001.22 iterations 622
> >
> > Cbc0038I Solution found of 1001.22
> >
> > Cbc0038I Branch and bound needed to clear up 5 general integers
> >
> > Cbc0038I Mini branch and bound could not fix general integers
> >
> > Cbc0038I No solution found this major pass
> >
> > Cbc0038I Before mini branch and bound, 2941 integers at bound fixed
> > and
> > 1 continuous
> >
> > Cbc0038I Full problem 2400 rows 2976 columns, reduced to 42 rows 32
> > columns
> >
> > Cbc0038I Mini branch and bound did not improve solution (1.69 seconds)
> >
> > Cbc0038I After 1.69 seconds - Feasibility pump exiting with objective
> > of
> > 1092.61 - took 0.94 seconds
> >
> > Cbc0012I Integer solution of 1092.6135 found by feasibility pump after
> > 0 iterations and 0 nodes (1.69 seconds)
> >
> > Cbc0038I Full problem 2400 rows 2976 columns, reduced to 32 rows 24
> > columns
> >
> > Cbc0012I Integer solution of 719.2773 found by RINS after 0 iterations
> > and 0 nodes (1.84 seconds)
> >
> > Cbc0031I 11 added rows had average density of 528.81818
> >
> > Cbc0013I At root node, 11 cuts changed objective from 178.71006 to
> > 179.04079 in 10 passes
> >
> > Cbc0014I Cut generator 0 (Probing) - 0 row cuts average 0.0 elements,
> > 1 column cuts (1 active)  in 0.598 seconds - new frequency is 1
> >
> > Cbc0014I Cut generator 1 (Gomory) - 12 row cuts average 1499.0
> > elements,
> > 0 column cuts (0 active)  in 0.095 seconds - new frequency is 1
> >
> > Cbc0014I Cut generator 2 (Knapsack) - 0 row cuts average 0.0 elements,
> > 0 column cuts (0 active)  in 0.054 seconds - new frequency is -100
> >
> > Cbc0014I Cut generator 3 (Clique) - 0 row cuts average 0.0 elements, 0
> > column cuts (0 active)  in 0.002 seconds - new frequency is -100
> >
> > Cbc0014I Cut generator 4 (MixedIntegerRounding2) - 14 row cuts average
> > 652.2 elements, 0 column cuts (0 active)  in 0.166 seconds - new
> > frequency is 1
> >
> > Cbc0014I Cut generator 5 (FlowCover) - 8 row cuts average 4.5
> > elements,
> > 0 column cuts (0 active)  in 0.023 seconds - new frequency is 1
> >
> > Cbc0014I Cut generator 6 (TwoMirCuts) - 2 row cuts average 6.0
> > elements,
> > 0 column cuts (0 active)  in 0.058 seconds - new frequency is -100
> >
> > Cbc0010I After 0 nodes, 1 on tree, 719.2773 best solution, best
> > possible
> > 179.04079 (3.24 seconds)
> >
> > Cbc0012I Integer solution of 266.2883 found by DiveCoefficient after
> > 104 iterations and 1 nodes (3.38 seconds)
> >
> > Cbc0012I Integer solution of 266.2883 found by heuristic after 98
> > iterations and 0 nodes (3.56 seconds)
> >
> >
> >
> > Best regards,
> >
> >
> >
> >
> >
> > _______________________________________________
> > Cbc mailing list
> > Cbc at list.coin-or.org
> > http://list.coin-or.org/mailman/listinfo/cbc
> >
>
> --
> ==================================================
> Haroldo Gambini Santos
> D.Sc, Computer Science
> Universidade Federal de Ouro Preto
> http://www.decom.ufop.br/haroldo/
>
> _______________________________________________
> Cbc mailing list
> Cbc at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/cbc
>

-- 
Dr. Ted Ralphs
Professor, Lehigh University
(610) 628-1280
ted 'at' lehigh 'dot' edu
coral.ie.lehigh.edu/~ted
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/cbc/attachments/20160714/79c52af4/attachment.html>