[Cbc] Threadsafe usage of Cbc

[Ted Ralphs]

We just recently ran into exactly the same issue calling Cbc to solve
individual subproblems in parallel within our decomposition framework
(DIP). Rather than calling CbcMain*, we are now just directly calling the
branchAndBound of CbcModel and it seems to be working fine, though I
haven't looked into how the performance compares. John can probably say
what the difference is between the two methods of calling Cbc. For example,
see the code starting here:

https://projects.coin-or.org/Dip/browser/trunk/Dip/src/DecompModel.cpp#L234

This is also how Cbc is called within Couenne (and probably Bonmin as
well). I believe Couenne was also successfully using Cbc in parallel.
Couenne probably does this a bit more smartly (in terms of setting
parameters, etc.). You can poke around this file:

https://projects.coin-or.org/Couenne/browser/trunk/Couenne/src/main/CouenneBab.cpp

for ideas.

Cheers,

Ted


On Tue, Dec 17, 2013 at 10:34 AM, John Forrest
<john.forrest at fastercoin.com>wrote:

>  Cornelius,
>
> static variables were used just because that was the way a main program
> became callable.  Some work was started a few years ago to improve - I will
> have another look to see what minimum work needs to be done to make
> CbcMain0/1 calls threadsafe.
>
> John Forrest
>
>
> On 17/12/13 15:11, Cornelius Riemenschneider wrote:
>
>  Hello,
>
> we use CBC via googles or-tools from java in our application.
>
> Compared to scientific problems, our problems are small and solved by Cbc
> in <1s (which is great!).
>
> As we process requests from users concurrently, we’d also like to call Cbc
> concurrently.
>
> Currently, this is impossible, because Cbc uses the static parameters
> array in CbcMain0 and CbcMain1 and thus crashes or reads incorrect
> parameters.
>
> One backtrace, for example is:
>
>
>
> #2 0x00007fb3435bf52b in __libc_message (do_abort=<optimized out>,
> fmt=<optimized out>) at ../sysdeps/unix/sysv/linux/libc_fatal.c:189
> #3 0x00007fb3435c8d76 in malloc_printerr (action=3, str=0x7fb3436a1248
> "double free or corruption (!prev)", ptr=<optimized out>) at malloc.c:6283
> #4 0x00007fb3435cdaac in *__GI___libc_free (mem=<optimized out>) at
> malloc.c:3738
> #5 0x00007fb342733f06 in std::string::assign(std::string const&) () from
> /usr/lib/x86_64-linux-gnu/libstdc++.so.6
> #6 0x00007fb2d652cb7b in establishParams(int&, CbcOrClpParam*) () from
> libjnilinearsolver.so
> #7 0x00007fb2d64edc3b in CbcMain0(CbcModel&) () from libjnilinearsolver.so
> #8 0x00007fb2d6511bec in callCbc(char const*, CbcModel&) () from
> libjnilinearsolver.so
> #9 0x00007fb2d64e3fce in
> operations_research::CBCInterface::Solve(operations_research::MPSolverParameters
> const&) () from libjnilinearsolver.so
> #10 0x00007fb2d64cfe45 in
> operations_research::MPSolver::Solve(operations_research::MPSolverParameters
> const&) () from libjnilinearsolver.so
> #11 0x00007fb2d64cff3b in operations_research::MPSolver::Solve() () from
> libjnilinearsolver.so
> #12 0x00007fb2d64c91cf in
> Java_com_google_ortools_linearsolver_mainJNI_MPSolver_1solve_1_1SWIG_10 ()
> from libjnilinearsolver.so
>
> # from here on it’s just the JVM
>
>
>
> As I don’t know the code and it’s hard to read, I can’t really fix the
> problem by myself,
>
> but wouldn’t it be better to move parameters inside CbcModel instead of
> having a static array laying around?
>
>
>
> Andi f that’s done (or another solution is found), do you know of any
> thread-safety issues in cbc?
>
> Keep in mind, we don’t want Cbc to solve our problem in multiple threads,
> we just want to be able to solve different problems simultaneous.
>
>
>
> Regards,
>
> Cornelius Riemenschneider
>
> --
>
> ITscope GmbH
>
> Ludwig-Erhard-Allee 20
>
> 76131 Karlsruhe
>
> Email: cornelius.riemenschneider at itscope.de
>
> https://www.itscope.com
>
> Handelsregister: AG Mannheim, HRB 232782
>
> Sitz der Gesellschaft: Karlsruhe
>
> Geschäftsführer: Alexander Münkel, Benjamin Mund, Stefan Reger
>
>
>
>
> _______________________________________________
> Cbc mailing listCbc at list.coin-or.orghttp://list.coin-or.org/mailman/listinfo/cbc
>
>
>
> _______________________________________________
> Cbc mailing list
> Cbc at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/cbc
>
>


-- 
Dr. Ted Ralphs
Associate 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/20131217/cb9f670e/attachment-0001.html>