[petsc-dev] Fwd: [mumps-users] MUMPS MPI (5.2.1) does not compile

Pierre Jolivet pierre.jolivet at enseeiht.fr
Mon Jul 22 14:26:55 CDT 2019


Ah, shoot, I misunderstood the mail from the MUMPS ML, which states that there is a FTBS, but fixes something else…
I just tried myself and it seems OK as well.
The only difference I see with my user's ./configure and the MUMPS ML vs. Satish and me is that we are using MPICH (IMPI in my case) and they are using OpenMPI.

I’ll ask for the user to send its configure.log to petsc-maint.

Thanks and sorry for the trouble,
Pierre

> On 22 Jul 2019, at 9:06 PM, Balay, Satish <balay at mcs.anl.gov> wrote:
> 
> And the error below is with cmumps_save_restore.F - but I don't see fixes to this file in the patches.
> 
> So I tried a build with intel-17 - and that worked fine for me.
> 
> balay at es^/sandbox/balay/petsc(master=) $ ifort --version
> ifort (IFORT) 17.0.2 20170213
> Copyright (C) 1985-2017 Intel Corporation.  All rights reserved.
> balay at es^/sandbox/balay/petsc(master=) $ ./configure CC=icc FC=ifort CXX=icpc --download-mpich --download-scalapack --download-mumps --with-blaslapack-dir=$MKL_HOME
> 
> Attaching the corresponding configure.log
> 
> Satish
> 
> On Mon, 22 Jul 2019, Smith, Barry F. via petsc-dev wrote:
> 
>> 
>>  Pierre,
>> 
>>   I see four patches here. 
>> 
>> 1) for examples
>> 2) for shared library support (but only for Linux????) (two files for MUMPS and PORD) We can't use this unless also supports Mac etc
>> 3) MUMPS-Makefile.par.inc  What is this for?
>> 
>>  How does any of them resolve the problem with ifort failing to compile?
>> 
>>  Barry
>> 
>> 
>>> On Jul 22, 2019, at 1:28 PM, Pierre Jolivet via petsc-dev <petsc-dev at mcs.anl.gov> wrote:
>>> 
>>> We are having trouble compiling git.mumps with similar errors (using ifort 2017, yikes).
>>> Could the forwarded patches be applied to petsc/pkg-mumps, please?
>>> 
>>> Thanks in advance (and sorry if this should be handled @petsc-maint),
>>> Pierre
>>> 
>>>> Begin forwarded message:
>>>> 
>>>> From: "Antonio Trande" (anto.trande at gmail.com via mumps-users Mailing List) <mumps-users at listes.ens-lyon.fr>
>>>> Subject: [mumps-users] MUMPS MPI (5.2.1) does not compile
>>>> Date: 29 June 2019 at 5:56:28 PM CEST
>>>> To: "mumps-users at listes.ens-lyon.fr" <mumps-users at listes.ens-lyon.fr>
>>>> Reply-To: mumps-users at listes.ens-lyon.fr
>>>> 
>>>> 
>>>> Hi all.
>>>> 
>>>> MUMPS-5.2.1 is not compiling on Fedora 30 because of this error:
>>>> 
>>>> /usr/lib64/openmpi/bin/mpif77 -O2 -g -pipe -Wall -Werror=format-security
>>>> -Wp,-D_FORTIFY_SOURCE=2 -Wp,-D_GLIBCXX_ASSERTIONS -fexceptions
>>>> -fstack-protector-strong -grecord-gcc-switches
>>>> -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1
>>>> -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -mtune=generic
>>>> -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection
>>>> -Dscotch -Dmetis -Dptscotch -pthread -I/usr/lib64/gfortran/modules
>>>> -Dintel_ -Wno-unused-dummy-argument -Wno-maybe-uninitialized
>>>> -I/usr/include/openmpi-x86_64 -I/usr/include/openblas
>>>> -I/usr/include/openmpi-x86_64 -Dpord -I. -I../include -fPIC -c
>>>> cmumps_save_restore.F -o cmumps_save_restore.o
>>>> cmumps_save_restore.F:8105:50:
>>>> 
>>>> 8105 |                IF(associated(id%MPITOOMP_PROCS_MAP)) THEN
>>>>     |                                                  1
>>>> Error: 'mpitoomp_procs_map' at (1) is not a member of the 'cmumps_struc'
>>>> structure
>>>> cmumps_save_restore.F:8108:49:
>>>> 
>>>> 8108 |      &                 size(id%MPITOOMP_PROCS_MAP,1)*SIZE_INT
>>>>     |                                                 1
>>>> Error: 'mpitoomp_procs_map' at (1) is not a member of the 'cmumps_struc'
>>>> structure
>>>> cmumps_save_restore.F:8113:72:
>>>> 
>>>> 8113 |             elseif(trim(mode).EQ."save") then
>>>>     |
>>>>      1
>>>> Error: Unexpected ELSE IF statement at (1)
>>>> cmumps_save_restore.F:8114:50:
>>>> 
>>>> 8114 |                IF(associated(id%MPITOOMP_PROCS_MAP)) THEN
>>>>     |                                                  1
>>>> Error: 'mpitoomp_procs_map' at (1) is not a member of the 'cmumps_struc'
>>>> structure
>>>> cmumps_save_restore.F:8115:67:
>>>> 
>>>> 8115 |                   write(unit,iostat=err)
>>>> size(id%MPITOOMP_PROCS_MAP,1)
>>>>     |                                                                   1
>>>> Error: 'mpitoomp_procs_map' at (1) is not a member of the 'cmumps_struc'
>>>> structure
>>>> cmumps_save_restore.F:8124:62:
>>>> 
>>>> 8124 |                   write(unit,iostat=err) id%MPITOOMP_PROCS_MAP
>>>>     |                                                              1
>>>> Error: 'mpitoomp_procs_map' at (1) is not a member of the 'cmumps_struc'
>>>> structure
>>>> cmumps_save_restore.F:8125:72:
>>>> 
>>>> 8125 |                ELSE
>>>>     |
>>>>      1
>>>> Error: Unexpected ELSE statement at (1)
>>>> cmumps_save_restore.F:8136:18:
>>>> 
>>>> 8136 |                ENDIF
>>>>     |                  1
>>>> Error: Expecting END SELECT statement at (1)
>>>> cmumps_save_restore.F:8145:72:
>>>> 
>>>> 8145 |             elseif(trim(mode).EQ."restore") then
>>>>     |
>>>>      1
>>>> Error: Unexpected ELSE IF statement at (1)
>>>> cmumps_save_restore.F:8146:44:
>>>> 
>>>> 8146 |                nullify(id%MPITOOMP_PROCS_MAP)
>>>>     |                                            1
>>>> Error: 'mpitoomp_procs_map' at (1) is not a member of the 'cmumps_struc'
>>>> structure
>>>> cmumps_save_restore.F:8163:48:
>>>> 
>>>> 8163 |                   allocate(id%MPITOOMP_PROCS_MAP(size_array1),
>>>>     |                                                1
>>>> Error: 'mpitoomp_procs_map' at (1) is not a member of the 'cmumps_struc'
>>>> structure
>>>> cmumps_save_restore.F:8171:61:
>>>> 
>>>> 8171 |                   read(unit,iostat=err) id%MPITOOMP_PROCS_MAP
>>>>     |                                                             1
>>>> Error: 'mpitoomp_procs_map' at (1) is not a member of the 'cmumps_struc'
>>>> structure
>>>> cmumps_save_restore.F:8184:15:
>>>> 
>>>> 8184 |             endif
>>>>     |               1
>>>> Error: Expecting END SELECT statement at (1)
>>>> cmumps_save_restore.F:8188:64:
>>>> 
>>>> 8188 |                SIZE_VARIABLES(i1)=SIZE_INT*size(id%METIS_OPTIONS,1)
>>>>     |                                                                1
>>>> Error: 'metis_options' at (1) is not a member of the 'cmumps_struc'
>>>> structure
>>>> cmumps_save_restore.F:8190:54:
>>>> 
>>>> 8190 |                write(unit,iostat=err) id%METIS_OPTIONS
>>>>     |                                                      1
>>>> Error: 'metis_options' at (1) is not a member of the 'cmumps_struc'
>>>> structure
>>>> cmumps_save_restore.F:8200:64:
>>>> 
>>>> 8200 |                SIZE_VARIABLES(i1)=SIZE_INT*size(id%METIS_OPTIONS,1)
>>>>     |                                                                1
>>>> Error: 'metis_options' at (1) is not a member of the 'cmumps_struc'
>>>> structure
>>>> cmumps_save_restore.F:8201:53:
>>>> 
>>>> 8201 |                read(unit,iostat=err) id%METIS_OPTIONS
>>>>     |                                                     1
>>>> Error: 'metis_options' at (1) is not a member of the 'cmumps_struc'
>>>> structure
>>>> make[3]: *** [Makefile:396: cmumps_save_restore.o] Error 1
>>>> 
>>>> I'm patching MUMPS for compiling shared libraries for Fedora. I'm
>>>> attaching the patch used.
>>>> 
>>>> -- 
>>>> ---
>>>> Antonio Trande
>>>> Fedora Project
>>>> mailto 'sagitter at fedoraproject dot org'
>>>> GPG key: 0x6e0331dd1699e4d7
>>>> GPG key server: https://keys.fedoraproject.org/
>>>> Subscription/unsubscription: see https://listes.ens-lyon.fr/sympa/subscribe/mumps-users
>>> 
>>> <MUMPS-examples-mpilibs.patch><MUMPS-Makefile.par.inc><MUMPS-shared.patch><MUMPS-shared-pord.patch>
>> 
> <configure.log.gz>



More information about the petsc-dev mailing list