[Users] error message MPI
Roland Haas
rhaas at illinois.edu
Mon Oct 15 09:04:32 CDT 2018
Hello Maria,
the commit that you cherry picked is now in master and this should be
fine in the future.
Yours,
Roland
> That actually worked!
> The only complaint I had was at PAPI, but after I disabled it, the compilation went all the way to the end.
>
> Thank you!
> Maria
>
> _______________________
> Dr. Maria C. Babiuc Hamilton
> Department of Physics
> Marshall University
> S257 Science Building
> Huntington, WV, 25755
> Phone: (304)696-2754
>
>
> ________________________________________
> From: Roland Haas <rhaas at illinois.edu>
> Sent: Tuesday, October 9, 2018 2:28 PM
> To: Hamilton, Maria
> Cc: Steven R. Brandt; users at einsteintoolkit.org
> Subject: Re: [Users] error message MPI
>
> Hello Maria,
>
> would you mind trying the following:
>
> cd repos/carpet
> git cherry-pick 5073af2c
>
> then try and compile?
>
> This commit is part of the proposed updated to Carpet that Steve is currently reviewing and that we will commit in short order.
>
> Yours,
> Roland
>
> > Hello Steve,
> >
> > I tried again to compile cactus on osx, in fact I erase the whole cactus, did a fresh download of the current development version, brew reinstalled gcc-8 --without-multilib and brew reinstall openmpi --build-from-source --cc=gcc-8, then did a ./simfactory/bin/sim setup-silent --optionlist=osx-homebrew.cfg --runscript generic-mpi.run and ./simfactory/bin/sim build.
> > The good news is that I don't have any more complains about LORENE and PAPI.
> > The bad news is that I have the same errors in dh.cc
> >
> > COMPILING arrangements/Carpet/CarpetLib/src/dh.cc
> > /Users/babiuc/Cactus/arrangements/Carpet/CarpetLib/src/dh.cc: In function 'ompi_datatype_t* mpi_datatype(const dh::light_dboxes&)':
> > /Users/babiuc/Cactus/arrangements/Carpet/CarpetLib/src/dh.cc:2278:23 error: 'MPI_UB' was not declared in this scope
> > {1, sizeof s, MPI_UB, "MPI_UB", "MPI_UB"}};
> > ^~~~~~
> > /Users/babiuc/Cactus/arrangements/Carpet/CarpetLib/src/dh.cc:2278:23 note: suggested alternative: 'MPI_IO'
> > {1, sizeof s, MPI_UB, "MPI_UB", "MPI_UB"}};
> > ^~~~~~
> > MPI_IO
> > /Users/babiuc/Cactus/arrangements/Carpet/CarpetLib/src/dh.cc: In function 'ompi_datatype_t* mpi_datatype(const dh::fast_dboxes&)':
> > /Users/babiuc/Cactus/arrangements/Carpet/CarpetLib/src/dh.cc:2330:23 error: 'MPI_UB' was not declared in this scope
> > {1, sizeof s, MPI_UB, "MPI_UB", "MPI_UB"}};
> > ^~~~~~
> > /Users/babiuc/Cactus/arrangements/Carpet/CarpetLib/src/dh.cc:2330:23 note: suggested alternative: 'MPI_IO'
> > {1, sizeof s, MPI_UB, "MPI_UB", "MPI_UB"}};
> > ^~~~~~
> > MPI_IO
> > make[3]: *** [dh.cc.o] Error 1
> >
> > Maria
> >
> > _______________________
> > Dr. Maria C. Babiuc Hamilton
> > Department of Physics
> > Marshall University
> > S257 Science Building
> > Huntington, WV, 25755
> > Phone: (304)696-2754
> >
> >
> > ________________________________________
> > From: users-bounces at einsteintoolkit.org <users-bounces at einsteintoolkit.org> on behalf of Steven R. Brandt <sbrandt at cct.lsu.edu>
> > Sent: Thursday, October 4, 2018 1:21 PM
> > To: users at einsteintoolkit.org
> > Subject: Re: [Users] error message MPI
> >
> > I recently pushed a change for osx. If you redo your setup (i.e. remove
> > etc/defs.local.ini and mdb/machines/your-machine.ini) and try again, it
> > may help.
> >
> > Is mpicxx in your command path? I.e. if you type 'which mpicxx' do you
> > see anything?
> >
> > Thanks.
> >
> > --Steve
> >
> >
> > On 10/04/2018 08:43 AM, Hamilton, Maria wrote:
> > > Roland,
> > >
> > > A quick update. My hacking managed to get the code compiled, but it hangs on when running. I'm sure I did something wrong.
> > > I'll wait for your pointers after you have time to install and run Cactus on Mojave.
> > >
> > > Maria
> > >
> > > _______________________
> > > Dr. Maria C. Babiuc Hamilton
> > > Department of Physics
> > > Marshall University
> > > S257 Science Building
> > > Huntington, WV, 25755
> > > Phone: (304)696-2754
> > >
> > >
> > > ________________________________________
> > > From: users-bounces at einsteintoolkit.org <users-bounces at einsteintoolkit.org> on behalf of Hamilton, Maria <babiuc at marshall.edu>
> > > Sent: Wednesday, October 3, 2018 10:52 PM
> > > To: rhaas at illinois.edu
> > > Cc: users at einsteintoolkit.org
> > > Subject: Re: [Users] error message MPI
> > >
> > > Roland,
> > >
> > > OK, keep me updated. Honestly, I am afraid to update to Mojave!
> > >
> > > I do wander why Cactus consistently breaks down after any new updates I do on my mac laptop. Is it because of home-brew?
> > > Or shall I just stick with older versions of the software libraries?
> > > It finally compiled, after I commented out LORENE, PAPI, and the Meudon thorns, and did some hacking in the CarpetLib thorn.
> > > Clearly, my new brew update made MPU_UB deprecated.
> > > I ended up defining MPI-UB everywhere it was needed as:
> > > #define MPI_UB ((MPI_Datatype)16)
> > > This should have been in an mpi.h file, but apparently is not.
> > > It also complained about MPI_Type_struct, which I commented out.
> > >
> > > I have an older Ubuntu workstation, and the new Cactus compiled there without any problems.
> > >
> > > I reinstalled also the new Cactus on BigGreen, the cluster at Marshall University, and it went well.
> > >
> > > One last thing: I could not find any ubuntu.cfg.
> > > Also, I don't see a cents.cfg either.
> > > Any suggestion about what to use instead?
> > > I would like to have some flexibility on what version of compilers to use.
> > >
> > > cheers,
> > > Maria
> > >
> > > _______________________
> > > Dr. Maria C. Babiuc Hamilton
> > > Department of Physics
> > > Marshall University
> > > S257 Science Building
> > > Huntington, WV, 25755
> > > Phone: (304)696-2754
> > >
> > >
> > > ________________________________________
> > > From: Roland Haas <rhaas at illinois.edu>
> > > Sent: Wednesday, October 3, 2018 10:20 AM
> > > To: Hamilton, Maria
> > > Cc: users at einsteintoolkit.org
> > > Subject: Re: [Users] error message MPI
> > >
> > > Hello Maria,
> > >
> > > this will take a while. Installing Mojave is complicated for me and I
> > > likely will not get to it until at least after Oct 15th, sorry.
> > >
> > > Yours,
> > > Roland
> > >
> > >> Hello Roland,
> > >>
> > >> OK, thanks! It would be great if you let me know how the compilation went on your new mac.
> > >> The errors I had were popping out at Carpet/CarpetLib/src/dh.cc were for the compilation from scratch.
> > >>
> > >> I did try to use sim as well, as you suggested, and I had a cascade of errors.
> > >> The first complaint was about LORENE:
> > >> _______________
> > >> LORENE: Installing...
> > >> ~/Cactus/configs/sim/scratch
> > >> LORENE: Cleaning up...
> > >> LORENE: Done.
> > >> Creating /Users/babiuc/Cactus/configs/sim/lib/libthorn_LORENE.a
> > >> make: *** [sim] Error 2
> > >> __________
> > >>
> > >>
> > >> I commented LORENE out and I had to comment out 3 more thorns:
> > >> CST error 1:
> > >> -> Thorn 'Meudon_Bin_BH' requires the capability 'LORENE'.
> > >> Please add a thorn that provides 'LORENE' to your ThornList or remove 'Meudon_Bin_BH' from it !
> > >>
> > >>
> > >> CST error 2:
> > >> -> Thorn 'Meudon_Bin_NS' requires the capability 'LORENE'.
> > >> Please add a thorn that provides 'LORENE' to your ThornList or remove 'Meudon_Bin_NS' from it !
> > >>
> > >>
> > >> CST error 3:
> > >> -> Thorn 'Meudon_Mag_NS' requires the capability 'LORENE'.
> > >> Please add a thorn that provides 'LORENE' to your ThornList or remove 'Meudon_Mag_NS' from it !
> > >>
> > >>
> > >> Then I had an error at thorn PAPI:
> > >> _______
> > >> Checking status of thorn PAPI
> > >> PAPI: Preparing directory structure...
> > >> PAPI: Unpacking archive...
> > >> ~/Cactus/configs/sim/scratch/build/PAPI ~/Cactus/configs/sim/scratch
> > >> PAPI: Applying patches...
> > >> ~/Cactus/configs/sim/scratch/build/PAPI/papi-5.3.0 ~/Cactus/configs/sim/scratch/build/PAPI ~/Cactus/configs/sim/scratch
> > >> Cannot complete in-place edit of ./src/libpfm-3.y/examples_v2.x/ia64/mont_opcode.c: failed to rename work file './src/libpfm-3.y/examples_v2.x/ia64/XXxgZzZf' to './src/libpfm-3.y/examples_v2.x/ia64/mont_opcode.c': Bad file descriptor, <> line 32848.
> > >> make[2]: *** [make.checked] Error 2
> > >> make[1]: *** [/Users/babiuc/Cactus/configs/sim/lib/libthorn_PAPI.a] Error 2
> > >> make[1]: *** Waiting for unfinished jobs....
> > >> Creating /Users/babiuc/Cactus/configs/sim/lib/libthorn_LocalReduce.a
> > >> make: *** [sim] Error 2
> > >> ___________________
> > >>
> > >> In the end I arrived at the same error:
> > >>
> > >> ____
> > >> COMPILING arrangements/Carpet/CarpetLib/src/dh.cc
> > >> /Users/babiuc/Cactus/arrangements/Carpet/CarpetLib/src/dh.cc: In function 'ompi_datatype_t* mpi_datatype(const dh::light_dboxes&)':
> > >> /Users/babiuc/Cactus/arrangements/Carpet/CarpetLib/src/dh.cc:2278:23: error: 'MPI_UB' was not declared in this scope
> > >> {1, sizeof s, MPI_UB, "MPI_UB", "MPI_UB"}};
> > >> ^~~~~~
> > >> /Users/babiuc/Cactus/arrangements/Carpet/CarpetLib/src/dh.cc:2278:23: note: suggested alternative: 'MPI_IO'
> > >> {1, sizeof s, MPI_UB, "MPI_UB", "MPI_UB"}};
> > >> ^~~~~~
> > >> MPI_IO
> > >> /Users/babiuc/Cactus/arrangements/Carpet/CarpetLib/src/dh.cc: In function 'ompi_datatype_t* mpi_datatype(const dh::fast_dboxes&)':
> > >> /Users/babiuc/Cactus/arrangements/Carpet/CarpetLib/src/dh.cc:2330:23: error: 'MPI_UB' was not declared in this scope
> > >> {1, sizeof s, MPI_UB, "MPI_UB", "MPI_UB"}};
> > >> ^~~~~~
> > >> /Users/babiuc/Cactus/arrangements/Carpet/CarpetLib/src/dh.cc:2330:23: note: suggested alternative: 'MPI_IO'
> > >> {1, sizeof s, MPI_UB, "MPI_UB", "MPI_UB"}};
> > >> ^~~~~~
> > >> MPI_IO
> > >> make[3]: *** [dh.cc.o] Error 1
> > >> _____________
> > >>
> > >>
> > >> Also, a thorn seems to be empty:
> > >> Missing thorn EinsteinInitialData/Hydro_RNSID
> > >>
> > >>
> > >>
> > >> Maria
> > >>
> > >> _______________________
> > >> Dr. Maria C. Babiuc Hamilton
> > >> Department of Physics
> > >> Marshall University
> > >> S257 Science Building
> > >> Huntington, WV, 25755
> > >> Phone: (304)696-2754
> > >>
> > >>
> > >> ________________________________________
> > >> From: Roland Haas <rhaas at illinois.edu>
> > >> Sent: Tuesday, October 2, 2018 9:01 AM
> > >> To: Hamilton, Maria
> > >> Cc: users at einsteintoolkit.org
> > >> Subject: Re: [Users] error message MPI
> > >>
> > >> Hello Maria,
> > >>
> > >> please try compiling from scratch. Ie do
> > >>
> > >> make sim-realclean (or even rm -rf configs/sim)
> > >>
> > >> and try compiling again. It is possible that a previous abortive
> > >> attempt to compile has created a file "mpi.h" in your Cactus tree and
> > >> that this file is messing up things now.
> > >>
> > >> If you already did start from scratch then I do not quite know what may
> > >> be going on (I do not normally use a Mac).
> > >>
> > >> I am updating my own mac today and will see if I get similar results.
> > >>
> > >> Yours,
> > >> Roland
> > >>
> > >>> Hello Roland,
> > >>>
> > >>> It turned out that I still have problems with MPI.
> > >>> I did not update to mojave, just did a brew update/brew upgrade.
> > >>> A smaller executable compiled fine, but compiling with most thorns activated, I have the following error:
> > >>>
> > >>> _____________________
> > >>> COMPILING arrangements/Carpet/CarpetLib/src/dh.cc
> > >>> /Users/babiuc/Cactus/arrangements/Carpet/CarpetLib/src/dh.cc: In function 'ompi_datatype_t* mpi_datatype(const dh::light_dboxes&)':
> > >>> /Users/babiuc/Cactus/arrangements/Carpet/CarpetLib/src/dh.cc:2278:23: error: 'MPI_UB' was not declared in this scope
> > >>> {1, sizeof s, MPI_UB, "MPI_UB", "MPI_UB"}};
> > >>> ^~~~~~
> > >>> /Users/babiuc/Cactus/arrangements/Carpet/CarpetLib/src/dh.cc:2278:23: note: suggested alternative: 'MPI_IO'
> > >>> {1, sizeof s, MPI_UB, "MPI_UB", "MPI_UB"}};
> > >>> ^~~~~~
> > >>> MPI_IO
> > >>> /Users/babiuc/Cactus/arrangements/Carpet/CarpetLib/src/dh.cc: In function 'ompi_datatype_t* mpi_datatype(const dh::fast_dboxes&)':
> > >>> /Users/babiuc/Cactus/arrangements/Carpet/CarpetLib/src/dh.cc:2330:23: error: 'MPI_UB' was not declared in this scope
> > >>> {1, sizeof s, MPI_UB, "MPI_UB", "MPI_UB"}};
> > >>> ^~~~~~
> > >>> /Users/babiuc/Cactus/arrangements/Carpet/CarpetLib/src/dh.cc:2330:23: note: suggested alternative: 'MPI_IO'
> > >>> {1, sizeof s, MPI_UB, "MPI_UB", "MPI_UB"}};
> > >>> ^~~~~~
> > >>> MPI_IO
> > >>> ___________________
> > >>>
> > >>> My mpic++ does not take -compile_info, but a show gives me this:
> > >>>
> > >>> $ mpic++ --showme
> > >>> clang++ -I/usr/local/Cellar/open-mpi/HEAD-73075b8/include -L/usr/local/opt/libevent/lib -L/usr/local/Cellar/open-mpi/HEAD-73075b8/lib -lmpi
> > >>>
> > >>> What do you suggest?
> > >>>
> > >>> thanks,
> > >>> Maria
> > >>>
> > >>> _______________________
> > >>> Dr. Maria C. Babiuc Hamilton
> > >>> Department of Physics
> > >>> Marshall University
> > >>> S257 Science Building
> > >>> Huntington, WV, 25755
> > >>> Phone: (304)696-2754
> > >>>
> > >>>
> > >>> ________________________________________
> > >>> From: Roland Haas <rhaas at illinois.edu>
> > >>> Sent: Monday, October 1, 2018 4:59 PM
> > >>> To: Hamilton, Maria
> > >>> Cc: users at einsteintoolkit.org
> > >>> Subject: Re: [Users] error message MPI
> > >>>
> > >>> Hello Maria,
> > >>>
> > >>> if by update you mean an update of the OS (to Mojave) or the package
> > >>> manager (Homebrew, Macports) then most likely the update made you MPI
> > >>> stack unusable.
> > >>>
> > >>> I would try the following:
> > >>>
> > >>> 1. if you updated the OS to Mojave then quite likely you will have to
> > >>> reinstall the package manager and all packages
> > >>> (https://apple.stackexchange.com/questions/85480/how-should-i-fix-my-homebrew-installation-after-upgrading-os-x)
> > >>> though if you are lucky a "brew upgrade" may do the trick
> > >>>
> > >>> 2. check the output of the commands the Cactus tried ie:
> > >>> usr/local/bin/mpic++ -compile_info and /usr/local/bin/mpic++ --showme
> > >>> to see what kind of error is produced. Likely you have to re-install
> > >>> the MPI package from your package manager (same as point 1. really).
> > >>>
> > >>> Yours,
> > >>> Roland
> > >>>
> > >>>> Hello,
> > >>>>
> > >>>>
> > >>>> I'm trying to compile the development Cactus version on my Mac laptop
> > >>>> after an update:
> > >>>>
> > >>>> $repos/CRL/GetComponents --update.
> > >>>>
> > >>>> Unfortunately, I ran into the following error:
> > >>>>
> > >>>> ********************************************************************************
> > >>>>
> > >>>> Running configuration script for thorn MPI:
> > >>>>
> > >>>> Found MPI compiler wrapper at /usr/local/bin/mpic++!
> > >>>>
> > >>>> sh: line 1: 30205 Abort trap: 6 /usr/local/bin/mpic++
> > >>>> -compile_info 2> /dev/null
> > >>>>
> > >>>> sh: line 1: 30207 Abort trap: 6 /usr/local/bin/mpic++
> > >>>> --showme 2> /dev/null
> > >>>>
> > >>>> ERROR: MPI could not be configured: neither automatic nor manual
> > >>>> configuration succeeded
> > >>>>
> > >>>>
> > >>>> CST error 1:
> > >>>>
> > >>>> -> Configuration script for thorn MPI returned exit code 5
> > >>>>
> > >>>> Error message: 'MPI could not be configured: neither automatic
> > >>>> nor manual configuration succeeded'
> > >>>>
> > >>>>
> > >>>> Finished running configuration script for thorn MPI.
> > >>>>
> > >>>>
> > >>>> ********************************************************************************
> > >>>>
> > >>>> This persists after real clean, and even after I erase the
> > >>>> configuration and reconfigure it.
> > >>>>
> > >>>> Please help,
> > >>>> Maria
> > >>>> ?
> > >>>>
> > >>>>
> > >>>> _______________________
> > >>>> Dr. Maria C. Babiuc Hamilton
> > >>>> Department of Physics
> > >>>> Marshall University
> > >>>> S257 Science Building
> > >>>> Huntington, WV, 25755
> > >>>> Phone: (304)696-2754
> > >>>>
> > >>>
> > >>>
> > >>> --
> > >>> My email is as private as my paper mail. I therefore support encrypting
> > >>> and signing email messages. Get my PGP key from http://pgp.mit.edu .
> > >>
> > >> --
> > >> My email is as private as my paper mail. I therefore support encrypting
> > >> and signing email messages. Get my PGP key from http://pgp.mit.edu .
> > >
> > > --
> > > My email is as private as my paper mail. I therefore support encrypting
> > > and signing email messages. Get my PGP key from http://pgp.mit.edu .
> > > _______________________________________________
> > > Users mailing list
> > > Users at einsteintoolkit.org
> > > http://lists.einsteintoolkit.org/mailman/listinfo/users
> > > _______________________________________________
> > > Users mailing list
> > > Users at einsteintoolkit.org
> > > http://lists.einsteintoolkit.org/mailman/listinfo/users
> >
> > _______________________________________________
> > Users mailing list
> > Users at einsteintoolkit.org
> > http://lists.einsteintoolkit.org/mailman/listinfo/users
> > _______________________________________________
> > Users mailing list
> > Users at einsteintoolkit.org
> > http://lists.einsteintoolkit.org/mailman/listinfo/users
>
>
> --
> My email is as private as my paper mail. I therefore support encrypting
> and signing email messages. Get my PGP key from http://pgp.mit.edu .
--
My email is as private as my paper mail. I therefore support encrypting
and signing email messages. Get my PGP key from http://pgp.mit.edu .
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 833 bytes
Desc: OpenPGP digital signature
Url : http://lists.einsteintoolkit.org/pipermail/users/attachments/20181015/1248ecd9/attachment.bin
More information about the Users
mailing list