Thanks Jan, I am sorry, but I realised only now that I missed the last part of your previous message about the request to test the timings demo.
It works nicely. Thanks again. Best, Corrado > On 04 Apr 2015, at 14:48, Jan Blechta <[email protected]> wrote: > > The fix is in master. > > Jan > > > On Thu, 2 Apr 2015 15:09:19 +0200 > Jan Blechta <[email protected]> wrote: > >> So pushing into next. Expect it getting to master in few days. >> >> On Thu, 2 Apr 2015 14:55:38 +0200 >> Corrado Maurini <[email protected]> wrote: >> >>> Nice, it works! (or at least it compiles) >> >> You could try running timing demo in parallel. >> >> Jan >> >>> >>> Thanks a lot. >>> >>> Corrado >>> >>> Corrado Maurini >>> [email protected] >>> >>> >>> >>>> On 02 Apr 2015, at 14:27, Jan Blechta <[email protected]> >>>> wrote: >>>> >>>> Please, try again with DOLFIN >>>> 94ba11b4f63c83590d1dce2583754fe85ae6cb08 >>>> >>>> Jan >>>> >>>> >>>> On Thu, 2 Apr 2015 11:46:12 +0200 >>>> Jan Blechta <[email protected]> wrote: >>>> >>>>> Hi Corrado, >>>>> >>>>> I'm looking into this and it seems to me that there is something >>>>> non-standard in MPI_Op definition in your MPI implementation. >>>>> Compiler understands MPI_MAX's type as <anonymous enum> while it >>>>> should be MPI_Op so that >>>>> >>>>> template<> >>>>> Table dolfin::MPI::all_reduce(MPI_Comm, const Table&, MPI_Op); >>>>> >>>>> specialization is used rather than >>>>> >>>>> template<typename T, typename X> >>>>> T dolfin::MPI::all_reduce(MPI_Comm comm, const T& value, X op) >>>>> >>>>> which is expected to fail for T=Table, X=MPI_Op. >>>>> >>>>> Could you point me to the headers with declaration of MPI_Op, >>>>> MPI_MAX? >>>>> >>>>> Jan >>>>> >>>>> >>>>> On Wed, 1 Apr 2015 22:50:12 +0200 >>>>> Corrado Maurini <[email protected]> wrote: >>>>> >>>>>> Hi all, >>>>>> >>>>>> I got the error below when building dolfin-dev >>>>>> (git:05296b363ece99fca64bd1b9312a5c2ecefe0777) on a SGI shared >>>>>> memory machine using mpt/2.11 as mpi library: >>>>>> >>>>>> You find here the full build.log: http://pastebin.com/jhEW1WHd >>>>>> <http://pastebin.com/jhEW1WHd> >>>>>> >>>>>> FYI, I use hashdist to build, but I think this is completely >>>>>> transparent. I can build without problem fenics 1.5. >>>>>> >>>>>> Can someone help? I think that the recent changes introduced >>>>>> some compatibility issues. >>>>>> >>>>>> >>>>>> 2015/04/01 22:30:03 - INFO: [package:run_job] [ 65%] Building >>>>>> CXX object dolfin/CMakeFiles/dolfin.dir/la/uBLASVector.cpp.o >>>>>> 2015/04/01 22:30:08 - INFO: [package:run_job] [ 65%] Building >>>>>> CXX object dolfin/CMakeFiles/dolfin.dir/log/Event.cpp.o >>>>>> 2015/04/01 22:30:08 - INFO: [package:run_job] [ 65%] Building >>>>>> CXX object dolfin/CMakeFiles/dolfin.dir/log/LogManager.cpp.o >>>>>> 2015/04/01 22:30:09 >>>>>> - INFO: [package:run_job] [ 67%] Building CXX object >>>>>> dolfin/CMakeFiles/dolfin.dir/log/LogStream.cpp.o 2015/04/01 >>>>>> 22:30:12 >>>>>> - INFO: [package:run_job] [ 67%] Building CXX object >>>>>> dolfin/CMakeFiles/dolfin.dir/log/Logger.cpp.o 2015/04/01 >>>>>> 22:30:15 >>>>>> - INFO: [package:run_job] In file included >>>>>> from >>>>>> /opt/dev/libs/fenics/hashdist-builds/tmp/dolfin-jt5cc6fobjg7/dolfin/log/Logger.cpp:45:0: >>>>>> 2015/04/01 22:30:15 - INFO: >>>>>> [package:run_job] >>>>>> /opt/dev/libs/fenics/hashdist-builds/tmp/dolfin-jt5cc6fobjg7/dolfin/common/MPI.h: >>>>>> In instantiation of 'static MPI_Datatype dolfin::MPI::mpi_type() >>>>>> [with T = dol fin::Table; MPI_Datatype = unsigned int]': >>>>>> 2015/04/01 22:30:15 - INFO: >>>>>> [package:run_job] >>>>>> /opt/dev/libs/fenics/hashdist-builds/tmp/dolfin-jt5cc6fobjg7/dolfin/common/MPI.h:613:64: >>>>>> required from 'static T dolfin::MPI::all_reduce(MPI_Comm, const >>>>>> T&, X ) [with T = dolfin::Table; X = <anonymous enum>; MPI_Comm >>>>>> = unsigned int]' 2015/04/01 22:30:15 - INFO: >>>>>> [package:run_job] >>>>>> /opt/dev/libs/fenics/hashdist-builds/tmp/dolfin-jt5cc6fobjg7/dolfin/common/MPI.h:623:43: >>>>>> required from 'static T dolfin::MPI::max(MPI_Comm, const T&) >>>>>> [with T = dolfin::Table; MPI_Comm = unsigned int]' 2015/04/01 >>>>>> 22:30:15 - INFO: >>>>>> [package:run_job] >>>>>> /opt/dev/libs/fenics/hashdist-builds/tmp/dolfin-jt5cc6fobjg7/dolfin/log/Logger.cpp:327:43: >>>>>> required from here 2015/04/01 22:30:15 - INFO: >>>>>> [package:run_job] >>>>>> /opt/dev/libs/fenics/hashdist-builds/tmp/dolfin-jt5cc6fobjg7/dolfin/common/MPI.h:230:7: >>>>>> error: static assertion failed: Unknown MPI type 2015/04/01 >>>>>> 22:30:15 >>>>>> - INFO: [package:run_job] >>>>>> static_assert(dependent_false<T>::value, "Unknown MPI type"); >>>>>> 2015/04/01 22:30:15 - INFO: [package:run_job] ^ >>>>>> 2015/04/01 22:30:16 - INFO: [package:run_job] gmake[2]: *** >>>>>> [dolfin/CMakeFiles/dolfin.dir/log/Logger.cpp.o] Error 1 >>>>>> 2015/04/01 22:30:16 - INFO: [package:run_job] gmake[1]: *** >>>>>> [dolfin/CMakeFiles/dolfin.dir/all] Error 2 2015/04/01 22:30:16 - >>>>>> INFO: [package:run_job] gmake: *** [all] Error 2 2015/04/01 >>>>>> 22:30:16 >>>>>> - ERROR: [package:run_job] Command '[u'/bin/bash', >>>>>> '_hashdist/build.sh']' returned non-zero exit status 2 >>>>>> 2015/04/01 22:30:16 - ERROR: [package:run_job] command failed >>>>>> (code=2); raising >>>>>> >>>>>> Corrado >>>>>> >>>>>> >>>>>> >>>>> >>>>> _______________________________________________ >>>>> fenics-support mailing list >>>>> [email protected] >>>>> http://fenicsproject.org/mailman/listinfo/fenics-support >>>> >> >> _______________________________________________ >> fenics-support mailing list >> [email protected] >> http://fenicsproject.org/mailman/listinfo/fenics-support > _______________________________________________ fenics-support mailing list [email protected] http://fenicsproject.org/mailman/listinfo/fenics-support
