[CHiPPS] Questions About Blis And example of Alps

[Matthew Saltzman]

On Fri, 2012-01-27 at 18:25 -0500, Ted Ralphs wrote: 
> Just to follow up on this, Abc is not very well maintained and is
> there as an example of how to use Alps. I would not be surprised if it
> has bugs.
> 
> As for why Blis solves the knapsack problem you referred to in one
> node, whereas the knapsack application took many nodes, this is not
> surprising. The knapsack application uses a very naive bounding
> procedure and has no sophisticated methodology, whereas Blis uses
> cutting planes and other sophisticated procedures to do the bounding.
> It would not be unusual for a knapsack problem to take many nodes with
> the naive approach, but only one with the more sophisticated approach.
> Hope this helps clarify things.

In fact, the ALPS knapsack solver was designed as a scalability proof of
concept.  It was intentionally made naive so that it would generate lots
of nodes and dispatch them cheaply.  It was never intended to be
competitive at solving knapsack problems.

> 
> Cheers,
> 
> Ted
> 
> 2011/12/4 Yan Xu <Yan.Xu at sas.com>:
> > If you have a TotalView debugger, can try to run through the debugger. It is
> > hard to tell if it is a system error or bug in ALPS.
> >
> >
> >
> > From: chipps-bounces at list.coin-or.org
> > [mailto:chipps-bounces at list.coin-or.org] On Behalf Of ???
> > Sent: 2011年11月29日 8:47 上午
> > To: chipps at list.coin-or.org
> > Cc: geogrid at gmail.com
> > Subject: [CHiPPS] Questions About Blis And example of Alps
> >
> >
> >
> > Dear Sir,
> >
> > I am now trying to using Alps solve Ip problems.
> >
> > I have two questions.
> >
> > First, I solved the knap3(Alps/Alps/Abc/data/knap3.mps) problem using Abc in
> > parallel.
> >
> > It generate good log file. I can see the work each process has done.
> >
> > And we can see that a few hundred nodes are searched.
> >
> > But when I use Blis to solve the same the knap3 problem.
> >
> > When it is solved, the information shows that only on node searched.
> >
> > The depth of the search tree is zero.
> >
> > Why is this?
> >
> > Second, I use Abc to solve a MIP problem(MPS file: 72MB, 75861 rows, 115173
> > columns, 1417123 elements) in parallel using 3(2,4 are tried) machines.
> >
> > But the program goes wrong after more or less 2000 seconds with the put
> > out::
> >
> > Fatal error in MPI_Test: Other MPI error, error stack:
> >
> > 0: MPI_Test(153)..................: MPI_Test(request=0x7fffe1cdbdb4,
> > flag=0x7fffe1cdbdb0, status=0x7fffe1cdbd50) failed
> >
> > 0: MPIDI_CH3I_Progress(150).......:
> >
> > 0: MPID_nem_mpich2_test_recv(800).:
> >
> > 0: MPID_nem_tcp_connpoll(1720)....:
> >
> > 0: state_commrdy_handler(1556)....:
> >
> > 0: MPID_nem_tcp_recv_handler(1446): socket closed
> >
> > rank 0 in job 5  vm3-1_33710   caused collective abort of all ranks
> >
> >   exit status of rank 0: return code 1
> >
> > Could you tell me why?
> >
> >
> >
> > Thanks a lot!
> >
> >
> > _______________________________________________
> > CHiPPS mailing list
> > CHiPPS at list.coin-or.org
> > http://list.coin-or.org/mailman/listinfo/chipps
> >
> 
> 
> 

-- 
Matthew Saltzman
Clemson University Mathematical Sciences
mjs AT clemson DOT edu