[Llvm-bgq-discuss] bgclang r205936-20140409 on vesta/mira
Biddiscombe, John A.
biddisco at cscs.ch
Fri Apr 11 07:42:25 CDT 2014
Hal,
I have
CC="/gpfs/bbp.cscs.ch/home/biddisco/apps/clang/bgclang/bin/bgclang"
CXX="/gpfs/bbp.cscs.ch/home/biddisco/apps/clang/bgclang/bin/bgclang++11"
after a bit of experimentation, I found that
$CC hello.c -o hello
works fine and executable runs on login and ION
$CXX hello.cxx -o hello
gives the link error
but
/gpfs/bbp.cscs.ch/home/biddisco/apps/clang/bgclang/mpi/bgclang/bin/mpic++ hello.cxx -o hello
works for a non mpi application on login and ION
and for an mpi aplication
/gpfs/bbp.cscs.ch/home/biddisco/apps/clang/bgclang/mpi/bgclang/bin/mpic++ hellompi.cxx -o hello
compiles ok, but on execution dies on login and ION
bbpbg2:~/src/c++test$ mpiexec -n 1 ./hello
/bgsys/source/srcV1R2M1.17463/comm/sys/buildtools/pami/common/bgq/BgqPersonality.h:102:
/bgsys/source/srcV1R2M1.17463/comm/sys/buildtools/pami/common/bgq/BgqPersonality.h<102>
=====================================================================================
= BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES
= EXIT CODE: 134
= CLEANING UP REMAINING PROCESSES
= YOU CAN IGNORE THE BELOW CLEANUP MESSAGES
=====================================================================================
APPLICATION TERMINATED WITH THE EXIT STRING: Aborted (signal 6)
Not sure if this helps you, but it's interesting that mpic++ links ok, when c++ does not, I use c++ compiler by default and link mkpi stuff myself (I don't use the mpi wrappers generally)
JB
> -----Original Message-----
> From: Hal Finkel [mailto:hfinkel at anl.gov]
> Sent: 11 April 2014 14:19
> To: Thomas Heller
> Cc: Biddiscombe, John A.; llvm-bgq-discuss at lists.alcf.anl.gov
> Subject: Re: [Llvm-bgq-discuss] bgclang r205936-20140409 on vesta/mira
>
> I'll look into this today.
>
> -Hal
>
> ----- Original Message -----
> > From: "Thomas Heller" <thom.heller at gmail.com>
> > To: "John A. Biddiscombe" <biddisco at cscs.ch>,
> > llvm-bgq-discuss at lists.alcf.anl.gov
> > Sent: Friday, April 11, 2014 6:19:32 AM
> > Subject: Re: [Llvm-bgq-discuss] bgclang r205936-20140409 on vesta/mira
> >
> > On 04/11/2014 12:48 PM, Biddiscombe, John A. wrote:
> > > oops, I think I pasted the wrong error into my email, the actual one
> > > I can't make go away is this
> > >
> > > /gpfs/bbp.cscs.ch/home/biddisco/apps/clang/bgclang/r205936-
> 20140409/libc++/lib/libc++.so:
> > > undefined reference to `vtable for
> > > std::nested_exception at CXXABI_1.3.5' clang: error: linker command
> > > failed with exit code 1 (use -v to see invocation)
> > >
> > > any suggestions welcome.
> >
> > I haven't seen this error before ...
> >
> > >
> > > JB
> > >
> > >
> > >> -----Original Message-----
> > >> From: llvm-bgq-discuss-bounces at lists.alcf.anl.gov
> > >> [mailto:llvm-bgq-discuss-
> > >> bounces at lists.alcf.anl.gov] On Behalf Of Thomas Heller
> > >> Sent: 11 April 2014 12:42
> > >> To: llvm-bgq-discuss at lists.alcf.anl.gov
> > >> Subject: Re: [Llvm-bgq-discuss] bgclang r205936-20140409 on
> > >> vesta/mira
> > >>
> > >> John,
> > >>
> > >> I have a dirty workaround for this in HPX ... let me commit real
> > >> quick (not sure where this comes from though)
> > >>
> > >> On 04/11/2014 12:03 PM, Biddiscombe, John A. wrote:
> > >>> Hal
> > >>>
> > >>> We have the toolchain_4.7.2 files installed in what I believe is
> > >>> the default location, recompiling boost+hpx still gives me the
> > >>> std::exception link errors, if I add
> > >>>
> > >>> /bgsys/drivers/V1R2M1/ppc64/gnu-linux-4.7.2/powerpc64-bgq-linux/li
> > >>> b/li
> > >>> bstdc++.a
> > >>>
> > >>> to the link libs, then all the errors go away except this one
> > >>>
> > >>> ../../../../lib/hpx/libhpxd.so.0.9.9: undefined reference to
> > >> `boost::re_detail::perl_matcher<std::__1::__wrap_iter<char
> > >> const*>,
> > >> std::__1::allocator<boost::sub_match<std::__1::__wrap_iter<char
> > >> const*> >
> > >>> , boost::regex_traits<char, boost::cpp_regex_traits<char> >
> > >>> >::match()'
> > >>>
> > >>> Which looks like a -stdlib=c++ problem.
> > >>>
> > >>> Can you tell me what setting you use to compile boost with? I've
> > >>> been using # user-config.jam (put this file into your home
> > >>> directory) using clang
> > >>> :
> > >>> : bgclang++11
> > >>> : <cxxflags> -std=c++0x
> > >>> ;
> > >>>
> > >>> and tried adding/removing -stdlib=libc++ or -stdlib=stdc++, but so
> > >>> far no
> > >> success.
> > >>>
> > >>> yours
> > >>>
> > >>> JB
> > >>>
> > >>>
> > >>>> -----Original Message-----
> > >>>> From: Hal Finkel [mailto:hfinkel at anl.gov]
> > >>>> Sent: 10 April 2014 19:41
> > >>>> To: Biddiscombe, John A.
> > >>>> Cc: llvm-bgq-discuss at lists.alcf.anl.gov
> > >>>> Subject: Re: [Llvm-bgq-discuss] bgclang r205936-20140409 on
> > >>>> vesta/mira
> > >>>>
> > >>>> John,
> > >>>>
> > >>>> Yes, this looks exactly like the issue that really motivated the
> > >>>> switch. There are necessary symbols that are just missing from
> > >>>> the old 4.4.6 libraries (the situation is somewhat complicated
> > >>>> because the compiler actually generated multiple version of these
> > >>>> destructors, and the older gcc did not generate all of them). As
> > >>>> I said,
> > >> please let me know if you can't get the files.
> > >>>>
> > >>>> -Hal
> > >>>>
> > >>>> ----- Original Message -----
> > >>>>> From: "John A. Biddiscombe" <biddisco at cscs.ch>
> > >>>>> To: "Hal Finkel" <hfinkel at anl.gov>
> > >>>>> Cc: llvm-bgq-discuss at lists.alcf.anl.gov
> > >>>>> Sent: Thursday, April 10, 2014 12:30:12 PM
> > >>>>> Subject: Re: [Llvm-bgq-discuss] bgclang r205936-20140409 on
> > >>>>> vesta/mira
> > >>>>>
> > >>>>> Hal
> > >>>>>
> > >>>>> Linking CXX executable ../../../../bin/network_storage
> > >>>>> CMakeFiles/network_storage_exe.dir/network_storage.cpp.o: In
> > >>>>> function
> > >>>>> `current_exception_std_exception_wrapper':
> > >>>>>
> > >>>>
> > >>
> /gpfs/bbp.cscs.ch/home/biddisco/apps/clang/boost_1_55_0/include/boo
> > >> st
> > >>>> -
> > >>>>> 1_55/
> > >>>>> boost/exception/detail/exception_ptr.hpp:220: undefined
> > >>>>> reference to `std::domain_error::~domain_error()'
> > >>>>> CMakeFiles/network_storage_exe.dir/network_storage.cpp.o: In
> > >>>>> function
> > >>>>> `~current_exception_std_exception_wrapper':
> > >>>>>
> > >>>>
> > >>
> /gpfs/bbp.cscs.ch/home/biddisco/apps/clang/boost_1_55_0/include/boo
> > >> st
> > >>>> -
> > >>>>> 1_55/
> > >>>>> boost/exception/detail/exception_ptr.hpp:224: undefined
> > >>>>> reference to `std::domain_error::~domain_error()'
> > >>>>> …. + million more
> > >>>>>
> > >>>>>
> > >>>>> I have to go out now, so I can’t play, but if those errors look
> > >>>>> like a
> > >>>>> libstdc++ issue then I might need those 4.7.2 files…
> > >>>>>
> > >>>>> Cheers
> > >>>>>
> > >>>>> JB
> > >>>>>
> > >>>>>
> > >>>>>
> > >>>>> On 10/04/14 19:06, "Hal Finkel" <hfinkel at anl.gov> wrote:
> > >>>>>
> > >>>>>> I'll also add that all V1R2M1 installs (as far as I know) use
> > >>>>>> 4.4.6
> > >>>>>> by default. The 4.7.2 files are supplementary; you (or your
> > >>>>>> system
> > >>>>>> admins)
> > >>>>>> just need to unpack them.
> > >>>>>>
> > >>>>>> -Hal
> > >>>>>>
> > >>>>>> ----- Original Message -----
> > >>>>>>> From: "Hal Finkel" <hfinkel at anl.gov>
> > >>>>>>> To: "John A. Biddiscombe" <biddisco at cscs.ch>
> > >>>>>>> Cc: llvm-bgq-discuss at lists.alcf.anl.gov
> > >>>>>>> Sent: Thursday, April 10, 2014 12:04:16 PM
> > >>>>>>> Subject: Re: [Llvm-bgq-discuss] bgclang r205936-20140409 on
> > >>>>>>> vesta/mira
> > >>>>>>>
> > >>>>>>> No problem; to be clear, the files from the 4.7.2 toolchain
> > >>>>>>> that are used by bgclang should not really be sensitive to the
> > >>>>>>> driver version you're actually using. If, however, you have
> > >>>>>>> any problems with the new build, please let me know.
> > >>>>>>>
> > >>>>>>> -Hal
> > >>>>>>>
> > >>>>>>> ----- Original Message -----
> > >>>>>>>> From: "John A. Biddiscombe" <biddisco at cscs.ch>
> > >>>>>>>> To: "John A. Biddiscombe" <biddisco at cscs.ch>, "Hal Finkel"
> > >>>>>>>> <hfinkel at anl.gov>, llvm-bgq-discuss at lists.alcf.anl.gov
> > >>>>>>>> Sent: Thursday, April 10, 2014 12:00:34 PM
> > >>>>>>>> Subject: Re: [Llvm-bgq-discuss] bgclang r205936-20140409 on
> > >>>>>>>> vesta/mira
> > >>>>>>>>
> > >>>>>>>> Sorry, rereading what you wrote the answer must be no. I was
> > >>>>>>>> confused because we have V1R2M1, but use 4.4.6
> > >>>>>>>>
> > >>>>>>>> JB
> > >>>>>>>>
> > >>>>>>>> On 10/04/14 18:55, "Biddiscombe, John A." <biddisco at cscs.ch>
> > >>>>>>>> wrote:
> > >>>>>>>>
> > >>>>>>>>> Hal,
> > >>>>>>>>>
> > >>>>>>>>>>
> > >>>>>>>>>> 1. The toolchain is now setup to use libc/libstdc++ files
> > >>>>>>>>> >from the
> > >>>>>>>>>> /bgsys/drivers/toolchain/V1R2M1_base_4.7.2 -- this gives
> > >>>>>>>>>> access to a more-recent environment for C/C++03
> > >>>>>>>>>> compilation, and furthermore, is necessary to make (at
> > >>>>>>>>>> least one part of) Boost work with
> > >>>>>>>>>> bgclang++11. You
> > >>>>>>>>>> don't need to be on a system running V1R2M1 (Mira is not
> > >>>>>>>>>> yet, for
> > >>>>>>>>>> example) to make use of this.
> > >>>>>>>>>
> > >>>>>>>>> We don’t have the toolchain 4.7.2 on our system. Does that
> > >>>>>>>>> mean I can’t use the new compiler? We have only files from
> > >>>>>>>>> 4.4.6 :(
> > >>>>>>>>>
> > >>>>>>>>> Thanks
> > >>>>>>>>>
> > >>>>>>>>> JB
> > >>>>>>>>>
> > >>>>>>>>>>
> > >>>>>>>>>> 2. There are more OpenMP 4 features available for you to
> > >>>>>>>>>> play
> > >>>>>>>>>> with:
> > >>>>>>>>>> - '#pragma omp simd' (and friends) should now work (in
> > >>>>>>>>>> the
> > >>>>>>>>>> sense
> > >>>>>>>>>> that
> > >>>>>>>>>> there is a reasonable probability that it will cause QPX to
> > >>>>>>>>>> be
> > >>>>>>>>>> generated)
> > >>>>>>>>>> - '#pragma omp task depend(...)' is supported (for the
> > >>>>>>>>>> most
> > >>>>>>>>>> part; I
> > >>>>>>>>>> think there are still some corner cases where the compiler
> > >>>>>>>>>> support is ahead of the runtime).
> > >>>>>>>>>> Please try these things out and let me know what you
> > >>>>>>>>>> find.
> > >>>>>>>>>>
> > >>>>>>>>>> For those maintaining their own installs, RPMs (and SRPMs)
> > >>>>>>>>>> are available on the trac page:
> > >>>>>>>>>> https://trac.alcf.anl.gov/projects/llvm-bgq
> > >>>>>>>>>> --
> > >>>>>>>>>> If
> > >>>>>>>>>> your
> > >>>>>>>>>> system does not have
> > >>>>>>>>>> /bgsys/drivers/toolchain/V1R2M1_base_4.7.2
> > >>>>>>>>>> unpacked,
> > >>>>>>>>>> and you can't get a copy of those files, please let me
> > >>>>>>>>>> know.
> > >>>>>>>>>>
> > >>>>>>>>>> Finally, for those strange creatures wanting to run the
> > >>>>>>>>>> compiler
> > >>>>>>>>> >from the
> > >>>>>>>>>> A2 login or I/O nodes, I believe this build should now work
> > >>>>>>>>>> for you again. ;)
> > >>>>>>>>>>
> > >>>>>>>>>> Thanks again,
> > >>>>>>>>>> Hal
> > >>>>>>>>>>
> > >>>>>>>>>> --
> > >>>>>>>>>> Hal Finkel
> > >>>>>>>>>> Assistant Computational Scientist Leadership Computing
> > >>>>>>>>>> Facility Argonne National Laboratory
> > >>>>>>>>>> _______________________________________________
> > >>>>>>>>>> llvm-bgq-discuss mailing list
> > >>>>>>>>>> llvm-bgq-discuss at lists.alcf.anl.gov
> > >>>>>>>>>> https://lists.alcf.anl.gov/mailman/listinfo/llvm-bgq-discus
> > >>>>>>>>>> s
> _______________________________________________
> > >>>>>>>>>> llvm-bgq-discuss mailing list
> > >>>>>>>>>> llvm-bgq-discuss at lists.alcf.anl.gov
> > >>>>>>>>>> https://lists.alcf.anl.gov/mailman/listinfo/llvm-bgq-discus
> > >>>>>>>>>> s
> _______________________________________________
> > >>>>>>>>>> llvm-bgq-discuss mailing list
> > >>>>>>>>>> llvm-bgq-discuss at lists.alcf.anl.gov
> > >>>>>>>>>> https://lists.alcf.anl.gov/mailman/listinfo/llvm-bgq-discus
> > >>>>>>>>>> s
> > >>>>>>>>>
> > >>>>>>>>> _______________________________________________
> > >>>>>>>>> llvm-bgq-discuss mailing list
> > >>>>>>>>> llvm-bgq-discuss at lists.alcf.anl.gov
> > >>>>>>>>> https://lists.alcf.anl.gov/mailman/listinfo/llvm-bgq-discuss
> > >>>>>>>>
> > >>>>>>>>
> > >>>>>>>
> > >>>>>>> --
> > >>>>>>> Hal Finkel
> > >>>>>>> Assistant Computational Scientist Leadership Computing
> > >>>>>>> Facility Argonne National Laboratory
> > >>>>>>> _______________________________________________
> > >>>>>>> llvm-bgq-discuss mailing list
> > >>>>>>> llvm-bgq-discuss at lists.alcf.anl.gov
> > >>>>>>> https://lists.alcf.anl.gov/mailman/listinfo/llvm-bgq-discuss
> > >>>>>>>
> > >>>>>>
> > >>>>>> --
> > >>>>>> Hal Finkel
> > >>>>>> Assistant Computational Scientist Leadership Computing Facility
> > >>>>>> Argonne National Laboratory
> > >>>>>
> > >>>>>
> > >>>>
> > >>>> --
> > >>>> Hal Finkel
> > >>>> Assistant Computational Scientist
> > >>>> Leadership Computing Facility
> > >>>> Argonne National Laboratory
> > >>> _______________________________________________
> > >>> llvm-bgq-discuss mailing list
> > >>> llvm-bgq-discuss at lists.alcf.anl.gov
> > >>> https://lists.alcf.anl.gov/mailman/listinfo/llvm-bgq-discuss
> > >>>
> > >>
> > >>
> > >> --
> > >> Thomas Heller
> > >> Friedrich-Alexander-Universität Erlangen-Nürnberg Department
> > >> Informatik -
> > >> Lehrstuhl Rechnerarchitektur Martensstr. 3
> > >> 91058 Erlangen
> > >> Tel.: 09131/85-27018
> > >> Fax: 09131/85-27912
> > >> Email: thomas.heller at cs.fau.de
> > >> _______________________________________________
> > >> llvm-bgq-discuss mailing list
> > >> llvm-bgq-discuss at lists.alcf.anl.gov
> > >> https://lists.alcf.anl.gov/mailman/listinfo/llvm-bgq-discuss
> >
> >
> > --
> > Thomas Heller
> > Friedrich-Alexander-Universität Erlangen-Nürnberg
> > Department Informatik - Lehrstuhl Rechnerarchitektur
> > Martensstr. 3
> > 91058 Erlangen
> > Tel.: 09131/85-27018
> > Fax: 09131/85-27912
> > Email: thomas.heller at cs.fau.de
> > _______________________________________________
> > llvm-bgq-discuss mailing list
> > llvm-bgq-discuss at lists.alcf.anl.gov
> > https://lists.alcf.anl.gov/mailman/listinfo/llvm-bgq-discuss
> >
>
> --
> Hal Finkel
> Assistant Computational Scientist
> Leadership Computing Facility
> Argonne National Laboratory
More information about the llvm-bgq-discuss
mailing list