November 2018 - discuss - mailman3.cels.anl.gov

Re: [mpich-discuss] Request AREQ0263985 assigned to your group: Nek5000 Queue: JLSE
by Congiu, Giuseppe 30 Nov '18

30 Nov '18

Hello, I am compiling on the login node. I need to run it on KNL nodes eventually. Thank you, Giuseppe On Nov 30, 2018, at 2:00 PM, Kumaran, Kalyan <kumaran(a)anl.gov<mailto:[email protected]>> wrote: Please send to discuss mailing list. What machine is this? On Nov 30, 2018, at 1:45 PM, Argonne Service Desk <argonne(a)service-now.com<mailto:[email protected]>> wrote: Hello, This request has been assigned to your group. Please assign it to an individual for resolution. Number: AREQ0263985<https://argonne.service-now.com/nav_to.do?uri=u_ad_hoc_request.do%3Fsys_id=…> Assignment Group: JLSE (CELS) Customer: Giuseppe Congiu Short Description: Nek5000 Queue: JLSE Details: received from: gcongiu(a)anl.gov<mailto:[email protected]> Hello, I am trying to compile a Nek5000 case on JLSE. When I run makenek I get the following linker error: mpif77 -O2 -fdefault-real-8 -fdefault-double-8 -cpp -DMPI -DUNDERSCORE -DTIMER -I/home/gcongiu/Nek5000/run/hillp -I/home/gcongiu/Nek5000/core -I./ -o nek5000 obj/hillp.o obj/drive.o obj/drive1.o obj/drive2.o obj/plan5.o obj/plan4.o obj/bdry.o obj/coef.o obj/conduct.o obj/connect1.o obj/connect2.o obj/dssum.o obj/edgec.o obj/eigsolv.o obj/gauss.o obj/genxyz.o obj/navier1.o obj/makeq.o obj/navier0.o obj/navier2.o obj/navier3.o obj/navier4.o obj/prepost.o obj/speclib.o obj/map2.o obj/mvmesh.o obj/ic.o obj/gfldr.o obj/ssolv.o obj/planx.o obj/math.o obj/mxm_wrapper.o obj/hmholtz.o obj/gfdm_par.o obj/gfdm_op.o obj/gfdm_solve.o obj/subs1.o obj/subs2.o obj/genbox.o obj/gmres.o obj/hsmg.o obj/convect.o obj/induct.o obj/perturb.o obj/navier5.o obj/navier6.o obj/navier7.o obj/navier8.o obj/fast3d.o obj/fasts.o obj/calcz.o obj/byte.o obj/chelpers.o obj/byte_mpi.o obj/postpro.o obj/intp_usr.o obj/cvode_driver.o obj/nek_comm.o obj/vprops.o obj/makeq_aux.o obj/papi.o obj/nek_in_situ.o obj/reader_rea.o obj/reader_par.o obj/reader_re2.o obj/finiparser.o obj/iniparser.o obj/dictionary.o obj/hpf.o obj/mxm_std.o obj/comm_mpi.o obj/singlmesh.o obj/blas.o obj/dsygv.o -L/home/gcongiu/Nek5000/3rd_party/gslib/lib -lgs /usr/bin/ld: obj/finiparser.o: undefined reference to symbol ‘__intel_sse2_strtok' Is there a module I can load that sets the right environment for the corresponding library? Thank you, Giuseppe Please let us know if this request has been incorrectly assigned to your group. Ref:MSG3494386

1 0

Re: [mpich-discuss] how to dynamically schedule process in mpich
by Zhou, Hui 29 Nov '18

29 Nov '18

What is your system? PBS? You have to prescribe a pool of nodes with qsub and run mpiexec -n 1 (for example). Then when you launch additional mpi process dynamically, hydra will spawn new processes within the prescribed pool. If your original pool is a single host, then all new dynamic process will be on that single host. If you describe multiple hosts, I believe hydra will do round-robin style allocation, i.e. the next processes will go to host 2, host 3, …., host N, host 1 (circle back). This is my current understanding. Let me know whether you can confirm. — Hui Zhou On Nov 29, 2018, at 4:41 PM, Shuwei Zhao <shuweizhao1991(a)gmail.com<mailto:[email protected]>> wrote: Hi Hui I cannot run mpiexec directly on my hosts because both rsh and ssh connections are disabled. I have to use qsub -P bnormal -pe dp 4 mpiexec ./my_exe to run my binary on remote hosts. Thanks Shuwei

1 0

Re: [mpich-discuss] how to dynamically schedule process in mpich
by Shuwei Zhao 29 Nov '18

29 Nov '18

Hi Hui I cannot run mpiexec directly on my hosts because both rsh and ssh connections are disabled. I have to use qsub -P bnormal -pe dp 4 mpiexec ./my_exe to run my binary on remote hosts. Thanks Shuwei On Wed, Nov 28, 2018 at 16:59 Zhou, Hui <zhouh(a)anl.gov> wrote: > Hi Shuwei, > > Were you able to run over multiple hosts directly, e.g.: > > ``` > $ mpiexec -n 2 -hosts “hostname_1,hostname_2” ./cpi > ``` > ? > > — > Hui Zhou > > > > > > > > > On Nov 28, 2018, at 3:33 PM, Shuwei Zhao <shuweizhao1991(a)gmail.com> wrote: > > Hi hui > > I tried your suggestion but looks like the newly spawned jobs are using > ssh to distribute on the new hosts. Howver In my machine farm, ssh & rsh > connections are disabled. have to submit through qsub network interface. Is > using qsub possible in mpi_comm_spawn? > > Thanks a lot > Shuwei > > On Mon, Nov 19, 2018 at 09:21 Zhou, Hui <zhouh(a)anl.gov> wrote: > >> On Sun, Nov 18, 2018 at 10:33:48PM -0600, Shuwei Zhao wrote: >> >2. About the MPI_Info object, I tried to use mpi_comm_spawn on the same >> >machine, it can work. But I don't know how to spawn processes cross >> >machines. For example, if I have 4 processes on machine 1, and I want to >> >spawn 4 more processes on other machines (resource managed by qsub), how >> do >> >I need to declare in the MPI_Info object? >> >> Did you try setting the host parameter in the info object? For >> example: >> >> MPI_Info info; >> MPI_Info_create(&info, "host", "A,B,C"); >> >> MPI_Comm_spawn("worker", MPI_ARGV_NULL, n_hosts, info, 0, >> MPI_COMM_SELF, &comm_workers, MPI_ERRCODES_IGNORE)? >> >> MPI_Info_free(&info); >> >> Replace "A,B,C" with host name list that you want to spawn process on. >> >> -- >> Hui Zhou >> > >

1 0

Re: [mpich-discuss] package issue on fedora
by bruno.guerraz＠orange.com 29 Nov '18

29 Nov '18

The answer is in the link provided by Zbyszek : " MPI implementation specific files MUST be installed in the directories used by the used MPI compiler ($MPI_BIN, $MPI_LIB and so on)." With my binaries installed under /usr/lib64/mpich/bin, I have the good requires -----Message d'origine----- De : Zbigniew Jędrzejewski-Szmek [mailto:[email protected]] Envoyé : mardi 27 novembre 2018 15:01 À : bruno via discuss Cc : GUERRAZ Bruno TGI/OLS Objet : Re: [mpich-discuss] package issue on fedora On Tue, Nov 27, 2018 at 01:40:32PM +0000, bruno via discuss wrote: > Hi all, > I am packaging a software that use MPICH on fedora. When I try to install the package, dnf tells me : > conflicting requests > - nothing provides libmpi.so.12()(64bit) needed by khiops-9.0.x86_64 > - nothing provides libmpicxx.so.12()(64bit) needed by > khiops-9.0.x86_64 > > And when I run "rpm -q --provides mpich" I have : > libmpi.so.12()(64bit)(mpich-x86_64) > libmpicxx.so.12()(64bit)(mpich-x86_64) The mpi libraries are not installed in the normal library search path in Fedora, hence the provides/requires tags use a suffix. This suffix also allows multiple mpi implementations to provide libmpi. See https://docs.fedoraproject.org/en-US/packaging-guidelines/MPI/. If you post your .spec file somewhere, I can take a look. Zbyszek _________________________________________________________________________________________________________________________ Ce message et ses pieces jointes peuvent contenir des informations confidentielles ou privilegiees et ne doivent donc pas etre diffuses, exploites ou copies sans autorisation. Si vous avez recu ce message par erreur, veuillez le signaler a l'expediteur et le detruire ainsi que les pieces jointes. Les messages electroniques etant susceptibles d'alteration, Orange decline toute responsabilite si ce message a ete altere, deforme ou falsifie. Merci. This message and its attachments may contain confidential or privileged information that may be protected by law; they should not be distributed, used or copied without authorisation. If you have received this email in error, please notify the sender and delete this message and its attachments. As emails may be altered, Orange is not liable for messages that have been modified, changed or falsified. Thank you.

1 0

Re: [mpich-discuss] how to dynamically schedule process in mpich
by Zhou, Hui 28 Nov '18

28 Nov '18

Hi Shuwei, Were you able to run over multiple hosts directly, e.g.: ``` $ mpiexec -n 2 -hosts “hostname_1,hostname_2” ./cpi ``` ? — Hui Zhou On Nov 28, 2018, at 3:33 PM, Shuwei Zhao <shuweizhao1991(a)gmail.com<mailto:[email protected]>> wrote: Hi hui I tried your suggestion but looks like the newly spawned jobs are using ssh to distribute on the new hosts. Howver In my machine farm, ssh & rsh connections are disabled. have to submit through qsub network interface. Is using qsub possible in mpi_comm_spawn? Thanks a lot Shuwei On Mon, Nov 19, 2018 at 09:21 Zhou, Hui <zhouh(a)anl.gov<mailto:[email protected]>> wrote: On Sun, Nov 18, 2018 at 10:33:48PM -0600, Shuwei Zhao wrote: >2. About the MPI_Info object, I tried to use mpi_comm_spawn on the same >machine, it can work. But I don't know how to spawn processes cross >machines. For example, if I have 4 processes on machine 1, and I want to >spawn 4 more processes on other machines (resource managed by qsub), how do >I need to declare in the MPI_Info object? Did you try setting the host parameter in the info object? For example: MPI_Info info; MPI_Info_create(&info, "host", "A,B,C"); MPI_Comm_spawn("worker", MPI_ARGV_NULL, n_hosts, info, 0, MPI_COMM_SELF, &comm_workers, MPI_ERRCODES_IGNORE)? MPI_Info_free(&info); Replace "A,B,C" with host name list that you want to spawn process on. -- Hui Zhou

1 0

Re: [mpich-discuss] how to dynamically schedule process in mpich
by Zhou, Hui 28 Nov '18

28 Nov '18

Hi Shuwei, Were you able to run over multiple hosts directly, e.g.: ``` $ mpiexec -n 2 -hosts “hostname_1,hostname_2” ./cpi ``` ? — Hui Zhou On Nov 28, 2018, at 3:33 PM, Shuwei Zhao <shuweizhao1991(a)gmail.com<mailto:[email protected]>> wrote: Hi hui I tried your suggestion but looks like the newly spawned jobs are using ssh to distribute on the new hosts. Howver In my machine farm, ssh & rsh connections are disabled. have to submit through qsub network interface. Is using qsub possible in mpi_comm_spawn? Thanks a lot Shuwei On Mon, Nov 19, 2018 at 09:21 Zhou, Hui <zhouh(a)anl.gov<mailto:[email protected]>> wrote: On Sun, Nov 18, 2018 at 10:33:48PM -0600, Shuwei Zhao wrote: >2. About the MPI_Info object, I tried to use mpi_comm_spawn on the same >machine, it can work. But I don't know how to spawn processes cross >machines. For example, if I have 4 processes on machine 1, and I want to >spawn 4 more processes on other machines (resource managed by qsub), how do >I need to declare in the MPI_Info object? Did you try setting the host parameter in the info object? For example: MPI_Info info; MPI_Info_create(&info, "host", "A,B,C"); MPI_Comm_spawn("worker", MPI_ARGV_NULL, n_hosts, info, 0, MPI_COMM_SELF, &comm_workers, MPI_ERRCODES_IGNORE)? MPI_Info_free(&info); Replace "A,B,C" with host name list that you want to spawn process on. -- Hui Zhou

1 0

Re: [mpich-discuss] how to dynamically schedule process in mpich
by Shuwei Zhao 28 Nov '18

28 Nov '18

Hi hui I tried your suggestion but looks like the newly spawned jobs are using ssh to distribute on the new hosts. Howver In my machine farm, ssh & rsh connections are disabled. have to submit through qsub network interface. Is using qsub possible in mpi_comm_spawn? Thanks a lot Shuwei On Mon, Nov 19, 2018 at 09:21 Zhou, Hui <zhouh(a)anl.gov> wrote: > On Sun, Nov 18, 2018 at 10:33:48PM -0600, Shuwei Zhao wrote: > >2. About the MPI_Info object, I tried to use mpi_comm_spawn on the same > >machine, it can work. But I don't know how to spawn processes cross > >machines. For example, if I have 4 processes on machine 1, and I want to > >spawn 4 more processes on other machines (resource managed by qsub), how > do > >I need to declare in the MPI_Info object? > > Did you try setting the host parameter in the info object? For > example: > > MPI_Info info; > MPI_Info_create(&info, "host", "A,B,C"); > > MPI_Comm_spawn("worker", MPI_ARGV_NULL, n_hosts, info, 0, > MPI_COMM_SELF, &comm_workers, MPI_ERRCODES_IGNORE)? > > MPI_Info_free(&info); > > Replace "A,B,C" with host name list that you want to spawn process on. > > -- > Hui Zhou >

1 0

Re: [mpich-discuss] package issue on fedora
by Zbigniew Jędrzejewski-Szmek 27 Nov '18

27 Nov '18

On Tue, Nov 27, 2018 at 01:40:32PM +0000, bruno via discuss wrote: > Hi all, > I am packaging a software that use MPICH on fedora. When I try to install the package, dnf tells me : > conflicting requests > - nothing provides libmpi.so.12()(64bit) needed by khiops-9.0.x86_64 > - nothing provides libmpicxx.so.12()(64bit) needed by khiops-9.0.x86_64 > > And when I run "rpm -q --provides mpich" I have : > libmpi.so.12()(64bit)(mpich-x86_64) > libmpicxx.so.12()(64bit)(mpich-x86_64) The mpi libraries are not installed in the normal library search path in Fedora, hence the provides/requires tags use a suffix. This suffix also allows multiple mpi implementations to provide libmpi. See https://docs.fedoraproject.org/en-US/packaging-guidelines/MPI/. If you post your .spec file somewhere, I can take a look. Zbyszek

1 0

package issue on fedora
by bruno.guerraz＠orange.com 27 Nov '18

27 Nov '18

Hi all, I am packaging a software that use MPICH on fedora. When I try to install the package, dnf tells me : conflicting requests - nothing provides libmpi.so.12()(64bit) needed by khiops-9.0.x86_64 - nothing provides libmpicxx.so.12()(64bit) needed by khiops-9.0.x86_64 And when I run "rpm -q --provides mpich" I have : libmpi.so.12()(64bit)(mpich-x86_64) libmpicxx.so.12()(64bit)(mpich-x86_64) there is not an exact match between Requires and Provides... how can I specify that libmpi.so.12()(64bit)(mpich-x86_64) is libmpi.so.12()(64bit) ? For the record, with the same spec file I have built packages for redhat and centOS with no issue. Thank you for your help Bruno _________________________________________________________________________________________________________________________ Ce message et ses pieces jointes peuvent contenir des informations confidentielles ou privilegiees et ne doivent donc pas etre diffuses, exploites ou copies sans autorisation. Si vous avez recu ce message par erreur, veuillez le signaler a l'expediteur et le detruire ainsi que les pieces jointes. Les messages electroniques etant susceptibles d'alteration, Orange decline toute responsabilite si ce message a ete altere, deforme ou falsifie. Merci. This message and its attachments may contain confidential or privileged information that may be protected by law; they should not be distributed, used or copied without authorisation. If you have received this email in error, please notify the sender and delete this message and its attachments. As emails may be altered, Orange is not liable for messages that have been modified, changed or falsified. Thank you.

1 0

Re: [mpich-discuss] _get_addr error while running application using MPICH
by Balaji, Pavan 26 Nov '18

26 Nov '18

> On Nov 26, 2018, at 10:12 AM, Zhou, Hui <zhouh(a)anl.gov> wrote: > Is there difference between mpirun and mpiexec. I understand the official standards only specifies mpiexec, and mpirun often (is it always?) is a link to mpiexec. Is there `unofficial` convention for mpirun, such as always present, always a link, no difference to mpiexec at all? It appears to me that mpirun is more wide spread than mpiexec (I always have been using mpirun). They are the same. The history is that MPI-1 didn't specify anything, so implementations started using "mpirun". When MPI-2 specified a launcher, they wanted to use a different name and came up with "mpiexec". In current implementations there is no difference. In mpich, they are both symlinks to mpiexec.hydra. -- Pavan

1 0