March 2019 - discuss - mailman3.cels.anl.gov

Re: [mpich-discuss] Error check in MPI_Startall with count=0
by Wesley Bland 20 Mar '19

20 Mar '19

I think a generous interpretation of the MPI Standard would probably agree with you. It's the same text for startall as it is for waitall (it should act like for i = 0 ... count; start). On Wed, Mar 20, 2019 at 2:51 PM Zhang, Junchao via discuss < discuss(a)mpich.org> wrote: > When count= 0, mpich should not raise error if array_of_requests is > NULL. Otherwise, one has to write code like if (count) MPI_Startall(count, > reqs). Note reqs is often NULL if count=0. See also MPI_Waitall, which > has the correct logic. > > MPI_Startall(int count, MPI_Request array_of_requests[]) { > ... > MPIR_ERRTEST_COUNT(count, mpi_errno); > MPIR_ERRTEST_ARGNULL(array_of_requests, "array_of_requests", > mpi_errno); > } > --Junchao Zhang > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss >

1 0

Error check in MPI_Startall with count=0
by Zhang, Junchao 20 Mar '19

20 Mar '19

When count= 0, mpich should not raise error if array_of_requests is NULL. Otherwise, one has to write code like if (count) MPI_Startall(count, reqs). Note reqs is often NULL if count=0. See also MPI_Waitall, which has the correct logic. MPI_Startall(int count, MPI_Request array_of_requests[]) { ... MPIR_ERRTEST_COUNT(count, mpi_errno); MPIR_ERRTEST_ARGNULL(array_of_requests, "array_of_requests", mpi_errno); } --Junchao Zhang

1 0

Re: [mpich-discuss] remove restrictive GCC check
by Jeff Hammond 14 Mar '19

14 Mar '19

Hi Ana, Thanks! I created https://github.com/pmodels/mpich/pull/3667 to kick off the CI systems... Jeff On Thu, Mar 14, 2019 at 12:02 PM Ana Guerrero López via discuss < discuss(a)mpich.org> wrote: > Hi, > > The current check in MPICH forcing to use exactly the same GCC version > to build MPICH and the user programs is too restrictive. I got bitten > by this and seems to be a recurrent problem, for example, see > https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=807666 > > The check was added many years ago by a problem with GCC3, > so I think it could be safely removed. Please see attached patch. > > Cheers, > Ana > > -- > Ana Guerrero Lopez > HPC Team - SUSE Labs > > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss > -- Jeff Hammond jeff.science(a)gmail.com http://jeffhammond.github.io/

1 0

remove restrictive GCC check
by Ana Guerrero López 14 Mar '19

14 Mar '19

Hi, The current check in MPICH forcing to use exactly the same GCC version to build MPICH and the user programs is too restrictive. I got bitten by this and seems to be a recurrent problem, for example, see https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=807666 The check was added many years ago by a problem with GCC3, so I think it could be safely removed. Please see attached patch. Cheers, Ana -- Ana Guerrero Lopez HPC Team - SUSE Labs

1 0

Re: [mpich-discuss] How to run program with arguments with mpich3
by Si, Min 06 Mar '19

06 Mar '19

Hi Kin, MPICH3.2.1 mpirun should work fine with any application arguments. Can you please first run a simple helloworld program without arguments ? mpirun -np 2 ./helloworld Then try to add some arguments in the program ? mpirun -np 2 ./helloworld arg1=xx arg2=xx If you still see the issue in the second step, please send the reproducer program to us. Thanks. Min On 2019/03/06 9:37, Yip, Kin via discuss wrote: Hi, Nobody has replied/advised and let me try another email before I give up on mpich3. The problem seems to be related to the "command line arguments" in my application (mcnpx). It worked in MPICH1 and I found the following in your user guide: https://www.mpich.org/static/downloads/3.3/mpich-3.3-userguide.pdf, you guys said : ======================================================================== MPICH1 (more precisely MPICH1's mpirun) required access to command line arguments in all application programs, including Fortran ones, and configure devoted some e ort to nding the libraries that contained the right versions of iargc and getarg and including those libraries with which the mpifort script linked MPI programs. Since MPICH does not require access to command line arguments to applications, these functions are optional, and configure does nothing special with them. If you need them in your applications, you will have to ensure that they are available in the Fortran environment you are using ================================================================== Why do I need to do when compiling MPICH3 (or mcnpx) to take care of the "iargc"/"getarg" part as mentioned in this note ?? I just ran mpich3 like mpich1 : $ ~/mpich3/bin/mpirun -np 5 -f machines ~kinyip/v270/bin/mpich3/mcnpx i=ptest n=ptest1. That's, putting the arguments "i=ptest n=ptest1." behind the application "mcnpx". Is it the normal way of using mpich3 ?? In your user guide, you don't have an example of running your application with argument. Can you please comment here ??? Kin -------- Forwarded Message -------- Subject: How to run program with arguments with mpich3 Date: Tue, 5 Mar 2019 10:01:48 -0500 From: Yip, Kin <kinyip(a)bnl.gov><mailto:[email protected]> To: discuss(a)mpich.org<mailto:[email protected]> <discuss(a)mpich.org><mailto:[email protected]> Hi, I've got an old code which works in MPICH1 (or 2). Now, I try to use MPICH3 . I've compiled MPICH3.2.1 with: env F77_GETARGDECL=" " CC="gcc" CXX="g++" FC="gfortran -ff2c -fno-second-underscore" ./configure --with-pm=hydra:gforker:remshell --enable-f90 --enable-f77 --enable-sharedlib --prefix=$HOME/mpich3 and then used it to compile the old code (MCNPX). When I run, $ ~/mpich3/bin/mpirun -np 5 -f machines ~kinyip/v270/bin/mpich3/mcnpx i=ptest n=ptest1. I've always got the error : bad trouble in mcnpx in routine getexm invalid number of command line arguments - use mpirun. It seems that it doesn't understand the "argument" above, which is " i=ptest n=ptest1. ". The gfortran does understand getarg etc. (otherwise, it'd not work with MPICH1). Do you have any advice to solve the "argument" problem.?? Kin _______________________________________________ discuss mailing list discuss(a)mpich.org<mailto:[email protected]> To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss

1 0

Re: [mpich-discuss] How to run program with arguments with mpich3
by Yip, Kin 06 Mar '19

06 Mar '19

Hi, Nobody has replied/advised and let me try another email before I give up on mpich3. The problem seems to be related to the "command line arguments" in my application (mcnpx). It worked in MPICH1 and I found the following in your user guide: https://www.mpich.org/static/downloads/3.3/mpich-3.3-userguide.pdf, you guys said : ======================================================================== MPICH1 (more precisely MPICH1's mpirun) required access to command line arguments in all application programs, including Fortran ones, and configure devoted some e ort to nding the libraries that contained the right versions of iargc and getarg and including those libraries with which the mpifort script linked MPI programs. Since MPICH does not require access to command line arguments to applications, these functions are optional, and configure does nothing special with them. If you need them in your applications, you will have to ensure that they are available in the Fortran environment you are using ================================================================== Why do I need to do when compiling MPICH3 (or mcnpx) to take care of the "iargc"/"getarg" part as mentioned in this note ?? I just ran mpich3 like mpich1 : $ ~/mpich3/bin/mpirun -np 5 -f machines ~kinyip/v270/bin/mpich3/mcnpx i=ptest n=ptest1. That's, putting the arguments "i=ptest n=ptest1." behind the application "mcnpx". Is it the normal way of using mpich3 ?? In your user guide, you don't have an example of running your application with argument. Can you please comment here ??? Kin -------- Forwarded Message -------- Subject: How to run program with arguments with mpich3 Date: Tue, 5 Mar 2019 10:01:48 -0500 From: Yip, Kin <kinyip(a)bnl.gov><mailto:[email protected]> To: discuss(a)mpich.org<mailto:[email protected]> <discuss(a)mpich.org><mailto:[email protected]> Hi, I've got an old code which works in MPICH1 (or 2). Now, I try to use MPICH3 . I've compiled MPICH3.2.1 with: env F77_GETARGDECL=" " CC="gcc" CXX="g++" FC="gfortran -ff2c -fno-second-underscore" ./configure --with-pm=hydra:gforker:remshell --enable-f90 --enable-f77 --enable-sharedlib --prefix=$HOME/mpich3 and then used it to compile the old code (MCNPX). When I run, $ ~/mpich3/bin/mpirun -np 5 -f machines ~kinyip/v270/bin/mpich3/mcnpx i=ptest n=ptest1. I've always got the error : bad trouble in mcnpx in routine getexm invalid number of command line arguments - use mpirun. It seems that it doesn't understand the "argument" above, which is " i=ptest n=ptest1. ". The gfortran does understand getarg etc. (otherwise, it'd not work with MPICH1). Do you have any advice to solve the "argument" problem.?? Kin

1 0

How to run program with arguments with mpich3
by Yip, Kin 05 Mar '19

05 Mar '19

Hi, I've got an old code which works in MPICH1 (or 2). Now, I try to use MPICH3 . I've compiled MPICH3.2.1 with: env F77_GETARGDECL=" " CC="gcc" CXX="g++" FC="gfortran -ff2c -fno-second-underscore" ./configure --with-pm=hydra:gforker:remshell --enable-f90 --enable-f77 --enable-sharedlib --prefix=$HOME/mpich3 and then used it to compile the old code (MCNPX). When I run, $ ~/mpich3/bin/mpirun -np 5 -f machines ~kinyip/v270/bin/mpich3/mcnpx i=ptest n=ptest1. I've always got the error : bad trouble in mcnpx in routine getexm invalid number of command line arguments - use mpirun. It seems that it doesn't understand the "argument" above, which is " i=ptest n=ptest1. ". The gfortran does understand getarg etc. (otherwise, it'd not work with MPICH1). Do you have any advice to solve the "argument" problem.?? Kin

1 0