[Dock-fans] using mpich2 with dock5

Kaitlyn Hwang hwang at mcs.anl.gov
Wed Oct 5 22:35:53 PDT 2005


Got dock5 working in parallel with mpich2-1.0.2p1  if I take out the
bugfixes 6 & 8 from
http://dock.compbio.ucsf.edu/DOCK_5/5.2.0/5.2.0_bugfixes.htm

Here's some dock5 errors with bugfixes 6&8 patched in and debug with gdb

94        while(c_library.get_mol(mol)) {
(gdb) n

Program received signal SIGSEGV, Segmentation fault.
0x40119dd7 in std::string::_Rep::_M_grab(std::allocator<char> const&,
std::allocator<char> const&) ()
   from /usr/lib/gcc-lib/i386-pc-linux-gnu/3.3.5/libstdc++.so.5
(gdb) up
#1  0x4011b1b7 in std::string::assign(std::string const&) ()
   from /usr/lib/gcc-lib/i386-pc-linux-gnu/3.3.5/libstdc++.so.5
(gdb) up
#2  0x4011a476 in std::string::operator=(std::string const&) ()
   from /usr/lib/gcc-lib/i386-pc-linux-gnu/3.3.5/libstdc++.so.5
(gdb) up
#3  0x080867b6 in copy_molecule(DOCKMol&, DOCKMol const&)
(target=@0xbfffcb30,
    original=@0x8168c10) at dockmol.cpp:222
222             target.current_data = original.current_data;



kaitlyn

On Mon, 3 Oct 2005, Philip Crawford wrote:

> Dock 4 doesn't use mpich for parallel execution. On a small testbed cluster
> (2 DP systems, athlon 1.8 DP plus athlon 2 DP, running RHEL 3) I had
> problems with Dock 4 parallelisation.
> Terry Lang advised that we download Dock 5.2. The MPI version compiled and
> ran on the same pair of DP systems without any problems. With MPI, one
> process (of 4 in our case) controls the parallel execution, the rest
> actually do the docking calculations. So, using "top" on a Unix system, you
> will only see n-1 processes chewing processor time if you launch n processes.
> I compiled against the latest 1.x MPI library, not the 2.x one. As I
> recall, the documentation with the source said Dock 5.2 had been written to
> that interface.
>
> Phil Crawford
> UNSW
> Sydney
>
> At 01:52 AM 2/10/2005, Kaitlyn Hwang wrote:
> >I tested ...
> >
> >mpiexec -np 2 valgrind ../../../../../dock5_mpi2/bin/dock5 -i
dock.in -o
> >dockval.out
> >
> >
> >I have a reponse from another dock-fan user that parallel works with
> >dock4. Has anyone seen or fix the below error? Or have dock4 working in
> >parallel. If so, which mpich did you use?
> >________________________________________________________________________
> >
> >==15884== Invalid read of size 4
> >==15884==    at 0x1BA04DD7:
> >std::string::_Rep::_M_grab(std::allocator<char> const&,
> >std::allocator<char> cons
> >t&) (in /usr/lib/gcc-lib/i386-pc-linux-gnu/3.3.5/libstdc++.so.5.0.7)
> >==15884==    by 0x1BA061B6: std::string::assign(std::string const&) (in
> >/usr/lib/gcc-lib/i386-pc-linux-gnu/3.
> >3.5/libstdc++.so.5.0.7)
> >==15884==    by 0x1BA05475: std::string::operator=(std::string const&) (in
> >/usr/lib/gcc-lib/i386-pc-linux-gnu
> >/3.3.5/libstdc++.so.5.0.7)
> >==15884==    by 0x80867B5: copy_molecule(DOCKMol&, DOCKMol const&)
> >(dockmol.cpp:222)
> >==15884==    by 0x8065E08: Base_MPI::get_from_send_queue(DOCKMol&)
> >(base_mpi.cpp:464)
> >==15884==    by 0x808FFC8: Library_File::get_mol(DOCKMol&)
> >(library_file.cpp:779)
> >==15884==    by 0x804B36C: main (dock.cpp:94)
> >==15884==  Address 0x4F525241 is not stack'd, malloc'd or (recently)
> >free'd
> >________________________________________________________________________
> >
> >
> >@*-*@
> >kaitlyn
> >_______________________________________________
> >Dock-fans mailing list
> >Dock-fans at docking.org
> >http://blur.compbio.ucsf.edu/mailman/listinfo/dock-fans
>
> Philip Crawford, B.Comp.Sc., MIEEE      Phone:  +61-2-93852564
> Computer Systems Officer                Fax:    +61-2-93851059
> Educational Media Unit                  Email:
> School of Medical Sciences,             P.Crawford at unsw.edu.au
> University of New South Wales,
> UNSW, NSW, 2052, Australia              Mobile: 0419-294 698
>
>



More information about the Dock-fans mailing list