February 2014 - discuss - mailman3.cels.anl.gov

Re: [mpich-discuss] Error "hydra requires VA_ARGS support", RHEL 5.10, icc
by Lu, Huiwei 27 Feb '14

27 Feb '14

Done reviewing. Tested with CentOS 6.5, icc version 13.1.3. $ export CFLAGS='-O3 -static-intel' $ export CXXFLAGS='-O3 -static-intel' $ export FFLAGS='-O3 -static-intel' $ ./configure && make — Huiwei On Feb 27, 2014, at 7:36 AM, Balaji, Pavan <balaji(a)anl.gov> wrote: > > I’ve pushed a review branch in for Huiwei. Once he signs off on it, I’ll commit it in an create a “nightly" tarball. I’ll send out an email once that’s done. It should be in today. > > — Pavan > > On Feb 27, 2014, at 4:36 AM, Бакиров, Медер <bakirov(a)almetico.com> wrote: > >> On Thu, Feb 27, 2014 at 12:59 PM, Balaji, Pavan <balaji(a)anl.gov> wrote: >>> Alternatively, you can wait till this is reviewed internally and committed to mpich/master. Then it'll show up in the nightly "release-like" tarballs. >> >> >> Thank you very much! I think, I will wait till this is reviewed and >> committed. Where can I follow a progress on this? >> >> With the best regards, >> Meder >> _______________________________________________ >> discuss mailing list discuss(a)mpich.org >> To manage subscription options or unsubscribe: >> https://lists.mpich.org/mailman/listinfo/discuss > > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss

1 0

Re: [mpich-discuss] Error "hydra requires VA_ARGS support", RHEL 5.10, icc
by Balaji, Pavan 27 Feb '14

27 Feb '14

I’ve pushed a review branch in for Huiwei. Once he signs off on it, I’ll commit it in an create a “nightly" tarball. I’ll send out an email once that’s done. It should be in today. — Pavan On Feb 27, 2014, at 4:36 AM, Бакиров, Медер <bakirov(a)almetico.com> wrote: > On Thu, Feb 27, 2014 at 12:59 PM, Balaji, Pavan <balaji(a)anl.gov> wrote: >> Alternatively, you can wait till this is reviewed internally and committed to mpich/master. Then it'll show up in the nightly "release-like" tarballs. > > > Thank you very much! I think, I will wait till this is reviewed and > committed. Where can I follow a progress on this? > > With the best regards, > Meder > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss

1 0

Re: [mpich-discuss] Error "hydra requires VA_ARGS support", RHEL 5.10, icc
by Бакиров, Медер 27 Feb '14

27 Feb '14

On Thu, Feb 27, 2014 at 12:59 PM, Balaji, Pavan <balaji(a)anl.gov> wrote: > Alternatively, you can wait till this is reviewed internally and committed to mpich/master. Then it'll show up in the nightly "release-like" tarballs. Thank you very much! I think, I will wait till this is reviewed and committed. Where can I follow a progress on this? With the best regards, Meder

1 0

Re: [mpich-discuss] Error "hydra requires VA_ARGS support", RHEL 5.10, icc
by Balaji, Pavan 27 Feb '14

27 Feb '14

On Feb 27, 2014, at 12:35 AM, Бакиров, Медер <bakirov(a)almetico.com> wrote: > On Wed, Feb 26, 2014 at 9:18 PM, Lu, Huiwei <huiweilu(a)mcs.anl.gov> wrote: >> Could you please try to compile MPICH-3.1 with less flags? For example, >> export CFLAGS='-O3' >> export CXXFLAGS='-O3' >> export FFLAGS='-O3' > > With these flags MPICH-3.1 compiles without any errors. > > I also played with flags, and found a problem flag ('-static-intel'), > which leads to compilation errors. Is there a possible workaround, I > can apply, to successfully compile MPICH-3.1 with full flags, > including '-static-intel’? Looks like -static-intel throws warnings sometimes that mpich does not like: % icc -static-intel test.c icc: warning #10237: -lcilkrts linked in dynamically, static library not available Maybe we are being too cautious here with disabling VA_ARGS if the compiler throws warnings. A better model might be to disable VA_ARGS only if the compiler is throwing additional warnings compared to the case when VA_ARGS is not used. The attached patch should improve this, but to try it out you’ll need to download the latest autoconf, automake and libtool, and run ./autogen.sh in the top-level directory before running configure. Alternatively, you can wait till this is reviewed internally and committed to mpich/master. Then it’ll show up in the nightly “release-like” tarballs. — Pavan

1 0

Re: [mpich-discuss] Error "hydra requires VA_ARGS support", RHEL 5.10, icc
by Бакиров, Медер 27 Feb '14

27 Feb '14

On Wed, Feb 26, 2014 at 9:18 PM, Lu, Huiwei <huiweilu(a)mcs.anl.gov> wrote: > Could you please try to compile MPICH-3.1 with less flags? For example, > export CFLAGS='-O3' > export CXXFLAGS='-O3' > export FFLAGS='-O3' With these flags MPICH-3.1 compiles without any errors. I also played with flags, and found a problem flag ('-static-intel'), which leads to compilation errors. Is there a possible workaround, I can apply, to successfully compile MPICH-3.1 with full flags, including '-static-intel'? Thanks! With the best regards, Meder Bakirov

1 0

Re: [mpich-discuss] Error "hydra requires VA_ARGS support", RHEL 5.10, icc
by Latham, Robert J. 26 Feb '14

26 Feb '14

On Wed, 2014-02-26 at 19:36 +0000, Latham, Robert J. wrote: > On Wed, 2014-02-26 at 17:47 +0000, Balaji, Pavan wrote: > > Note for MPICH developers: It’ll be good to figure out what exactly changed between 3.0.4 and 3.1 that is causing this. My understanding is that this check for warnings has been there in MPICH for a long time. > > git-bisect is awesome for this sort of thing. Pavan! % git bisect start v3.1 v3.0.4 Bisecting: 467 revisions left to test after this (roughly 9 steps) ... ccf7f70c3252d4e38933de00dd2f4bb49a389a61 is the first bad commit commit ccf7f70c3252d4e38933de00dd2f4bb49a389a61 Author: Pavan Balaji <balaji(a)mcs.anl.gov> Date: Sat May 18 22:38:40 2013 -0500 Warning squash for clang. For the VA_ARGS test, clang was accepting it in C89 mode, but throwing warnings. This causes configure to enable it, but the build to throw many warnings. This patch will disable VA_ARGS support unless the compiler can accept it without warnings. No reviewer. :040000 040000 cb59b5abfef07e8076e405141f701798ecde55f9 5727410f767d93a42a31cd461c37750e588223e0 M confdb bisect run success

1 0

Re: [mpich-discuss] Bug??
by Lu, Huiwei 26 Feb '14

26 Feb '14

Could you give us a small program using 2D arrays that causes the error? — Huiwei Lu Postdoc Appointee Mathematics and Computer Science Division, Argonne National Laboratory http://www.mcs.anl.gov/~huiweilu/ On Feb 26, 2014, at 3:26 PM, Soheil Hooshdaran <shooshdaran577(a)gmail.com> wrote: > > Hello. > It seems that MPICH3.0.4 doesn't work as expected with 2D arrays but it doesn't have that problem with 1-dimesional arrays. Can it be fixed? > > Thanks > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss

1 0

Bug??
by Soheil Hooshdaran 26 Feb '14

26 Feb '14

Hello. It seems that MPICH3.0.4 doesn't work as expected with 2D arrays but it doesn't have that problem with 1-dimesional arrays. Can it be fixed? Thanks

1 0

Re: [mpich-discuss] Compiling with -std=f2003 and mpif.h
by Junchao Zhang 26 Feb '14

26 Feb '14

Hi, Jed, Could you suggest a solution so that I can patch MPICH? Thanks. --Junchao Zhang On Wed, Feb 26, 2014 at 12:34 PM, Jed Brown <jed(a)jedbrown.org> wrote: > I have a user that wants me to check their F2003 code which uses mpif.h. > Attempting to compile with gfortran-4.8.2 -std=f2003 results in errors > coming from mpif.h. I tried configuring MPICH (v3.1-16-g38a5e61) using > FCFLAGS=-std=f2003, but I get build errors. GFortran does not have a > -std=gnu2003, so it appears that GNU extensions and F2003+ features are > mutually exclusive. What is MPICH's intent here? > > > mpifnoext.h:16.18: > Included at src/binding/f90/mpi_constants.f90:6: > > CHARACTER*1 MPI_ARGVS_NULL(1,1) > 1 > Warning: Obsolescent feature: Old-style character length at (1) > mpifnoext.h:17.18: > Included at src/binding/f90/mpi_constants.f90:6: > > CHARACTER*1 MPI_ARGV_NULL(1) > 1 > Warning: Obsolescent feature: Old-style character length at (1) > mpifnoext.h:528.16: > Included at src/binding/f90/mpi_constants.f90:6: > > integer*8 MPI_DISPLACEMENT_CURRENT > 1 > Error: GNU Extension: Nonstandard type declaration INTEGER*8 at (1) > mpifnoext.h:529.42: > Included at src/binding/f90/mpi_constants.f90:6: > > PARAMETER (MPI_DISPLACEMENT_CURRENT=-54278278) > 1 > Error: Symbol 'mpi_displacement_current' at (1) has no IMPLICIT type > > _______________________________________________ > discuss mailing list discuss(a)mpich.org > To manage subscription options or unsubscribe: > https://lists.mpich.org/mailman/listinfo/discuss >

1 0

Re: [mpich-discuss] Error "hydra requires VA_ARGS support", RHEL 5.10, icc
by Latham, Robert J. 26 Feb '14

26 Feb '14

On Wed, 2014-02-26 at 17:47 +0000, Balaji, Pavan wrote: > Note for MPICH developers: It’ll be good to figure out what exactly changed between 3.0.4 and 3.1 that is causing this. My understanding is that this check for warnings has been there in MPICH for a long time. git-bisect is awesome for this sort of thing. ==rob

1 0