March 2015 - petsc-dev - mailman3.cels.anl.gov

Re: [petsc-dev] did I mess up master?
by Jed Brown 31 Mar '15

31 Mar '15

Mark Adams <mfadams(a)lbl.gov> writes: > I see, I forgot to pull. > > These were trivial changes: changed a variable name (that is a key word) > and fixed a test that did not compile anymore so I did it directly on > master. The commit messages says "demonstrates a bug". Is there a bug? What is it? As for non-fast-forward merges, Junio just published a blog post explaining the rationale for maintaining first-parent history and the way to gracefully retry in case of racy integration. (His advice is the same as ours.) He also mentions discussion of a "git rebase --first-parent" feature that would make this process easier. http://git-blame.blogspot.com/2015/03/fun-with-non-fast-forward.html

1 0

Re: [petsc-dev] PETSC_COMM_SELF == 0
by Jed Brown 31 Mar '15

31 Mar '15

Mark Adams <mfadams(a)lbl.gov> writes: > BTW, I spent most of Friday finding that if I did not include the .h90 file > ISGetIndicesF90 would segv at the call site. That's Fortran productivity for you!

1 0

Re: [petsc-dev] PETSC_COMM_SELF == 0
by Mark Adams 31 Mar '15

31 Mar '15

On Tue, Mar 31, 2015 at 4:13 PM, Matthew Knepley <knepley(a)gmail.com> wrote: > On Tue, Mar 31, 2015 at 3:05 PM, Mark Adams <mfadams(a)lbl.gov> wrote: > >> Humm, that is what I see (appended). I put a print in PETSC and see this >> but I still have PETSC_COMM_SELF==0 in my fortran code. >> > > Is it possible that you do not have the right #include? Can you put > exactly the same includes as ex61f > Yea, I tried that. and changed it to .h90 BTW, I spent most of Friday finding that if I did not include the .h90 file ISGetIndicesF90 would segv at the call site. > > Matt > > >> Strange, the compile lines look close enough and the example ex61f sees a >> good PETSC_COMM_SELF ... >> >> Mark >> >> (gdb) where >> #0 PetscInitializeFortran () at >> /Users/markadams/Codes/petsc/src/sys/objects/ftn-custom/zstartf.c:48 >> #1 0x000000010027d985 in petscinitialize_ (filename=0x101eaab40 "", >> ierr=0x7fff5fbff18c, len=80) at >> /Users/markadams/Codes/petsc/src/sys/objects/ftn-custom/zstart.c:398 >> #2 0x00000001000015b2 in tpetsc () at ex61f.F90:111 >> #3 0x0000000100002a3d in main (argc=1, argv=0x7fff5fbff968 >> "/Users/markadams/Codes/petsc/src/ksp/ksp/examples/tutorials/ex61f") at >> ex61f.F90:60 >> Current language: auto; currently minimal >> (gdb) n >> 50 if (PETSC_COMM_WORLD) c1 = MPI_Comm_c2f(PETSC_COMM_WORLD); >> (gdb) p c1 >> $1 = 0 >> (gdb) n >> 51 c2 = MPI_Comm_c2f(PETSC_COMM_SELF); >> (gdb) p c1 >> $2 = 1140850688 >> (gdb) >> >> On Tue, Mar 31, 2015 at 1:32 PM, Satish Balay <balay(a)mcs.anl.gov> wrote: >> >>> 1. Check if this is reproduceable with a petsc example for your build. >>> >>> 2. if you run in the debugger - and place a breakpoint in >>> PetscInitializeFortran() - you should see the following trace. >>> >>> (gdb) where >>> #0 PetscInitializeFortran () at >>> /home/balay/petsc/src/sys/objects/ftn-custom/zstartf.c:48 >>> #1 0x00007ffff66c396d in petscinitialize_ (filename=0x603640 >>> <petscfortran1_> "", >>> ierr=0x7fffffffe0bc, len=80) at >>> /home/balay/petsc/src/sys/objects/ftn-custom/zstart.c:398 >>> #2 0x0000000000401a42 in MAIN__ () at ex2f.F:102 >>> (gdb) n >>> 50 if (PETSC_COMM_WORLD) c1 = MPI_Comm_c2f(PETSC_COMM_WORLD); >>> (gdb) p c1 >>> $1 = 0 >>> (gdb) n >>> 51 c2 = MPI_Comm_c2f(PETSC_COMM_SELF); >>> (gdb) p c1 >>> $2 = 1140850688 >>> >>> >>> Satish >>> >>> On Tue, 31 Mar 2015, Mark Adams wrote: >>> >>> > I'm not sure where we are. My code works on Edison but fails on OSX. >>> > Should I call PetscInitializeFortran? I do not have a C main. >>> > >>> > >>> > On Tue, Mar 31, 2015 at 12:14 PM, Barry Smith <bsmith(a)mcs.anl.gov> >>> wrote: >>> > >>> > > >>> > > PetscErrorCode PetscInitializeFortran(void) >>> > > { >>> > > MPI_Fint c1=0,c2=0; >>> > > >>> > > if (PETSC_COMM_WORLD) c1 = MPI_Comm_c2f(PETSC_COMM_WORLD); >>> > > c2 = MPI_Comm_c2f(PETSC_COMM_SELF); >>> > > petscsetcommonblock_(&c1,&c2); >>> > > >>> > > Note that this routine is called automatically by the Fortran >>> > > petscinitialize function. >>> > > >>> > > Barry >>> > > >>> > > > On Mar 31, 2015, at 10:19 AM, Matthew Knepley <knepley(a)gmail.com> >>> wrote: >>> > > > >>> > > > On Tue, Mar 31, 2015 at 10:13 AM, Mark Adams <mfadams(a)lbl.gov> >>> wrote: >>> > > > No, I don't have a C main program (like the doc says). >>> > > > >>> > > > I can see COMM_WORLD set in the Fortran initialize wrapper, but >>> where is >>> > > COMM_SELF set? >>> > > > >>> > > > MAtt >>> > > > >>> > > > On Tue, Mar 31, 2015 at 11:05 AM, Matthew Knepley < >>> knepley(a)gmail.com> >>> > > wrote: >>> > > > On Tue, Mar 31, 2015 at 5:44 AM, Mark Adams <mfadams(a)lbl.gov> >>> wrote: >>> > > > I'm finding that my fortran code on OSX has PETSC_COMM_SELF == 0. >>> > > > >>> > > > I can not find where PETSC_COMM_SELF is created. >>> > > > >>> > > > Its set in PetscInitializeFortran(). Did you call that? >>> > > > >>> > > > Matt >>> > > > >>> > > > >>> > > > Mark >>> > > > >>> > > > >>> > > > >>> > > > -- >>> > > > What most experimenters take for granted before they begin their >>> > > experiments is infinitely more interesting than any results to which >>> their >>> > > experiments lead. >>> > > > -- Norbert Wiener >>> > > > >>> > > > >>> > > > >>> > > > >>> > > > -- >>> > > > What most experimenters take for granted before they begin their >>> > > experiments is infinitely more interesting than any results to which >>> their >>> > > experiments lead. >>> > > > -- Norbert Wiener >>> > > >>> > > >>> > >>> >>> >> > > > -- > What most experimenters take for granted before they begin their > experiments is infinitely more interesting than any results to which their > experiments lead. > -- Norbert Wiener >

1 0

Re: [petsc-dev] PETSC_COMM_SELF == 0
by Matthew Knepley 31 Mar '15

31 Mar '15

On Tue, Mar 31, 2015 at 3:05 PM, Mark Adams <mfadams(a)lbl.gov> wrote: > Humm, that is what I see (appended). I put a print in PETSC and see this > but I still have PETSC_COMM_SELF==0 in my fortran code. > Is it possible that you do not have the right #include? Can you put exactly the same includes as ex61f Matt > Strange, the compile lines look close enough and the example ex61f sees a > good PETSC_COMM_SELF ... > > Mark > > (gdb) where > #0 PetscInitializeFortran () at > /Users/markadams/Codes/petsc/src/sys/objects/ftn-custom/zstartf.c:48 > #1 0x000000010027d985 in petscinitialize_ (filename=0x101eaab40 "", > ierr=0x7fff5fbff18c, len=80) at > /Users/markadams/Codes/petsc/src/sys/objects/ftn-custom/zstart.c:398 > #2 0x00000001000015b2 in tpetsc () at ex61f.F90:111 > #3 0x0000000100002a3d in main (argc=1, argv=0x7fff5fbff968 > "/Users/markadams/Codes/petsc/src/ksp/ksp/examples/tutorials/ex61f") at > ex61f.F90:60 > Current language: auto; currently minimal > (gdb) n > 50 if (PETSC_COMM_WORLD) c1 = MPI_Comm_c2f(PETSC_COMM_WORLD); > (gdb) p c1 > $1 = 0 > (gdb) n > 51 c2 = MPI_Comm_c2f(PETSC_COMM_SELF); > (gdb) p c1 > $2 = 1140850688 > (gdb) > > On Tue, Mar 31, 2015 at 1:32 PM, Satish Balay <balay(a)mcs.anl.gov> wrote: > >> 1. Check if this is reproduceable with a petsc example for your build. >> >> 2. if you run in the debugger - and place a breakpoint in >> PetscInitializeFortran() - you should see the following trace. >> >> (gdb) where >> #0 PetscInitializeFortran () at >> /home/balay/petsc/src/sys/objects/ftn-custom/zstartf.c:48 >> #1 0x00007ffff66c396d in petscinitialize_ (filename=0x603640 >> <petscfortran1_> "", >> ierr=0x7fffffffe0bc, len=80) at >> /home/balay/petsc/src/sys/objects/ftn-custom/zstart.c:398 >> #2 0x0000000000401a42 in MAIN__ () at ex2f.F:102 >> (gdb) n >> 50 if (PETSC_COMM_WORLD) c1 = MPI_Comm_c2f(PETSC_COMM_WORLD); >> (gdb) p c1 >> $1 = 0 >> (gdb) n >> 51 c2 = MPI_Comm_c2f(PETSC_COMM_SELF); >> (gdb) p c1 >> $2 = 1140850688 >> >> >> Satish >> >> On Tue, 31 Mar 2015, Mark Adams wrote: >> >> > I'm not sure where we are. My code works on Edison but fails on OSX. >> > Should I call PetscInitializeFortran? I do not have a C main. >> > >> > >> > On Tue, Mar 31, 2015 at 12:14 PM, Barry Smith <bsmith(a)mcs.anl.gov> >> wrote: >> > >> > > >> > > PetscErrorCode PetscInitializeFortran(void) >> > > { >> > > MPI_Fint c1=0,c2=0; >> > > >> > > if (PETSC_COMM_WORLD) c1 = MPI_Comm_c2f(PETSC_COMM_WORLD); >> > > c2 = MPI_Comm_c2f(PETSC_COMM_SELF); >> > > petscsetcommonblock_(&c1,&c2); >> > > >> > > Note that this routine is called automatically by the Fortran >> > > petscinitialize function. >> > > >> > > Barry >> > > >> > > > On Mar 31, 2015, at 10:19 AM, Matthew Knepley <knepley(a)gmail.com> >> wrote: >> > > > >> > > > On Tue, Mar 31, 2015 at 10:13 AM, Mark Adams <mfadams(a)lbl.gov> >> wrote: >> > > > No, I don't have a C main program (like the doc says). >> > > > >> > > > I can see COMM_WORLD set in the Fortran initialize wrapper, but >> where is >> > > COMM_SELF set? >> > > > >> > > > MAtt >> > > > >> > > > On Tue, Mar 31, 2015 at 11:05 AM, Matthew Knepley < >> knepley(a)gmail.com> >> > > wrote: >> > > > On Tue, Mar 31, 2015 at 5:44 AM, Mark Adams <mfadams(a)lbl.gov> >> wrote: >> > > > I'm finding that my fortran code on OSX has PETSC_COMM_SELF == 0. >> > > > >> > > > I can not find where PETSC_COMM_SELF is created. >> > > > >> > > > Its set in PetscInitializeFortran(). Did you call that? >> > > > >> > > > Matt >> > > > >> > > > >> > > > Mark >> > > > >> > > > >> > > > >> > > > -- >> > > > What most experimenters take for granted before they begin their >> > > experiments is infinitely more interesting than any results to which >> their >> > > experiments lead. >> > > > -- Norbert Wiener >> > > > >> > > > >> > > > >> > > > >> > > > -- >> > > > What most experimenters take for granted before they begin their >> > > experiments is infinitely more interesting than any results to which >> their >> > > experiments lead. >> > > > -- Norbert Wiener >> > > >> > > >> > >> >> > -- What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead. -- Norbert Wiener

1 0

Re: [petsc-dev] PETSC_COMM_SELF == 0
by Mark Adams 31 Mar '15

31 Mar '15

Humm, that is what I see (appended). I put a print in PETSC and see this but I still have PETSC_COMM_SELF==0 in my fortran code. Strange, the compile lines look close enough and the example ex61f sees a good PETSC_COMM_SELF ... Mark (gdb) where #0 PetscInitializeFortran () at /Users/markadams/Codes/petsc/src/sys/objects/ftn-custom/zstartf.c:48 #1 0x000000010027d985 in petscinitialize_ (filename=0x101eaab40 "", ierr=0x7fff5fbff18c, len=80) at /Users/markadams/Codes/petsc/src/sys/objects/ftn-custom/zstart.c:398 #2 0x00000001000015b2 in tpetsc () at ex61f.F90:111 #3 0x0000000100002a3d in main (argc=1, argv=0x7fff5fbff968 "/Users/markadams/Codes/petsc/src/ksp/ksp/examples/tutorials/ex61f") at ex61f.F90:60 Current language: auto; currently minimal (gdb) n 50 if (PETSC_COMM_WORLD) c1 = MPI_Comm_c2f(PETSC_COMM_WORLD); (gdb) p c1 $1 = 0 (gdb) n 51 c2 = MPI_Comm_c2f(PETSC_COMM_SELF); (gdb) p c1 $2 = 1140850688 (gdb) On Tue, Mar 31, 2015 at 1:32 PM, Satish Balay <balay(a)mcs.anl.gov> wrote: > 1. Check if this is reproduceable with a petsc example for your build. > > 2. if you run in the debugger - and place a breakpoint in > PetscInitializeFortran() - you should see the following trace. > > (gdb) where > #0 PetscInitializeFortran () at > /home/balay/petsc/src/sys/objects/ftn-custom/zstartf.c:48 > #1 0x00007ffff66c396d in petscinitialize_ (filename=0x603640 > <petscfortran1_> "", > ierr=0x7fffffffe0bc, len=80) at > /home/balay/petsc/src/sys/objects/ftn-custom/zstart.c:398 > #2 0x0000000000401a42 in MAIN__ () at ex2f.F:102 > (gdb) n > 50 if (PETSC_COMM_WORLD) c1 = MPI_Comm_c2f(PETSC_COMM_WORLD); > (gdb) p c1 > $1 = 0 > (gdb) n > 51 c2 = MPI_Comm_c2f(PETSC_COMM_SELF); > (gdb) p c1 > $2 = 1140850688 > > > Satish > > On Tue, 31 Mar 2015, Mark Adams wrote: > > > I'm not sure where we are. My code works on Edison but fails on OSX. > > Should I call PetscInitializeFortran? I do not have a C main. > > > > > > On Tue, Mar 31, 2015 at 12:14 PM, Barry Smith <bsmith(a)mcs.anl.gov> > wrote: > > > > > > > > PetscErrorCode PetscInitializeFortran(void) > > > { > > > MPI_Fint c1=0,c2=0; > > > > > > if (PETSC_COMM_WORLD) c1 = MPI_Comm_c2f(PETSC_COMM_WORLD); > > > c2 = MPI_Comm_c2f(PETSC_COMM_SELF); > > > petscsetcommonblock_(&c1,&c2); > > > > > > Note that this routine is called automatically by the Fortran > > > petscinitialize function. > > > > > > Barry > > > > > > > On Mar 31, 2015, at 10:19 AM, Matthew Knepley <knepley(a)gmail.com> > wrote: > > > > > > > > On Tue, Mar 31, 2015 at 10:13 AM, Mark Adams <mfadams(a)lbl.gov> > wrote: > > > > No, I don't have a C main program (like the doc says). > > > > > > > > I can see COMM_WORLD set in the Fortran initialize wrapper, but > where is > > > COMM_SELF set? > > > > > > > > MAtt > > > > > > > > On Tue, Mar 31, 2015 at 11:05 AM, Matthew Knepley <knepley(a)gmail.com > > > > > wrote: > > > > On Tue, Mar 31, 2015 at 5:44 AM, Mark Adams <mfadams(a)lbl.gov> wrote: > > > > I'm finding that my fortran code on OSX has PETSC_COMM_SELF == 0. > > > > > > > > I can not find where PETSC_COMM_SELF is created. > > > > > > > > Its set in PetscInitializeFortran(). Did you call that? > > > > > > > > Matt > > > > > > > > > > > > Mark > > > > > > > > > > > > > > > > -- > > > > What most experimenters take for granted before they begin their > > > experiments is infinitely more interesting than any results to which > their > > > experiments lead. > > > > -- Norbert Wiener > > > > > > > > > > > > > > > > > > > > -- > > > > What most experimenters take for granted before they begin their > > > experiments is infinitely more interesting than any results to which > their > > > experiments lead. > > > > -- Norbert Wiener > > > > > > > > > >

1 0

Re: [petsc-dev] PETSC_COMM_SELF == 0
by Satish Balay 31 Mar '15

31 Mar '15

1. Check if this is reproduceable with a petsc example for your build. 2. if you run in the debugger - and place a breakpoint in PetscInitializeFortran() - you should see the following trace. (gdb) where #0 PetscInitializeFortran () at /home/balay/petsc/src/sys/objects/ftn-custom/zstartf.c:48 #1 0x00007ffff66c396d in petscinitialize_ (filename=0x603640 <petscfortran1_> "", ierr=0x7fffffffe0bc, len=80) at /home/balay/petsc/src/sys/objects/ftn-custom/zstart.c:398 #2 0x0000000000401a42 in MAIN__ () at ex2f.F:102 (gdb) n 50 if (PETSC_COMM_WORLD) c1 = MPI_Comm_c2f(PETSC_COMM_WORLD); (gdb) p c1 $1 = 0 (gdb) n 51 c2 = MPI_Comm_c2f(PETSC_COMM_SELF); (gdb) p c1 $2 = 1140850688 Satish On Tue, 31 Mar 2015, Mark Adams wrote: > I'm not sure where we are. My code works on Edison but fails on OSX. > Should I call PetscInitializeFortran? I do not have a C main. > > > On Tue, Mar 31, 2015 at 12:14 PM, Barry Smith <bsmith(a)mcs.anl.gov> wrote: > > > > > PetscErrorCode PetscInitializeFortran(void) > > { > > MPI_Fint c1=0,c2=0; > > > > if (PETSC_COMM_WORLD) c1 = MPI_Comm_c2f(PETSC_COMM_WORLD); > > c2 = MPI_Comm_c2f(PETSC_COMM_SELF); > > petscsetcommonblock_(&c1,&c2); > > > > Note that this routine is called automatically by the Fortran > > petscinitialize function. > > > > Barry > > > > > On Mar 31, 2015, at 10:19 AM, Matthew Knepley <knepley(a)gmail.com> wrote: > > > > > > On Tue, Mar 31, 2015 at 10:13 AM, Mark Adams <mfadams(a)lbl.gov> wrote: > > > No, I don't have a C main program (like the doc says). > > > > > > I can see COMM_WORLD set in the Fortran initialize wrapper, but where is > > COMM_SELF set? > > > > > > MAtt > > > > > > On Tue, Mar 31, 2015 at 11:05 AM, Matthew Knepley <knepley(a)gmail.com> > > wrote: > > > On Tue, Mar 31, 2015 at 5:44 AM, Mark Adams <mfadams(a)lbl.gov> wrote: > > > I'm finding that my fortran code on OSX has PETSC_COMM_SELF == 0. > > > > > > I can not find where PETSC_COMM_SELF is created. > > > > > > Its set in PetscInitializeFortran(). Did you call that? > > > > > > Matt > > > > > > > > > Mark > > > > > > > > > > > > -- > > > What most experimenters take for granted before they begin their > > experiments is infinitely more interesting than any results to which their > > experiments lead. > > > -- Norbert Wiener > > > > > > > > > > > > > > > -- > > > What most experimenters take for granted before they begin their > > experiments is infinitely more interesting than any results to which their > > experiments lead. > > > -- Norbert Wiener > > > > >

1 0

Re: [petsc-dev] did I mess up master?
by Mark Adams 31 Mar '15

31 Mar '15

I see, I forgot to pull. These were trivial changes: changed a variable name (that is a key word) and fixed a test that did not compile anymore so I did it directly on master. Mark On Mon, Mar 30, 2015 at 12:02 PM, Jed Brown <jed(a)jedbrown.org> wrote: > Satish Balay <balay(a)mcs.anl.gov> writes: > > I interpreted this merge as 'commits were made directly to master' - > > not a separate feature branch. > > It either means developing on 'master' (which has always been > discouraged) or doing a fast-forward merge of a topic branch (also > discouraged, but easier to do accidentally). >

1 0

Re: [petsc-dev] PETSC_COMM_SELF == 0
by Mark Adams 31 Mar '15

31 Mar '15

I'm not sure where we are. My code works on Edison but fails on OSX. Should I call PetscInitializeFortran? I do not have a C main. On Tue, Mar 31, 2015 at 12:14 PM, Barry Smith <bsmith(a)mcs.anl.gov> wrote: > > PetscErrorCode PetscInitializeFortran(void) > { > MPI_Fint c1=0,c2=0; > > if (PETSC_COMM_WORLD) c1 = MPI_Comm_c2f(PETSC_COMM_WORLD); > c2 = MPI_Comm_c2f(PETSC_COMM_SELF); > petscsetcommonblock_(&c1,&c2); > > Note that this routine is called automatically by the Fortran > petscinitialize function. > > Barry > > > On Mar 31, 2015, at 10:19 AM, Matthew Knepley <knepley(a)gmail.com> wrote: > > > > On Tue, Mar 31, 2015 at 10:13 AM, Mark Adams <mfadams(a)lbl.gov> wrote: > > No, I don't have a C main program (like the doc says). > > > > I can see COMM_WORLD set in the Fortran initialize wrapper, but where is > COMM_SELF set? > > > > MAtt > > > > On Tue, Mar 31, 2015 at 11:05 AM, Matthew Knepley <knepley(a)gmail.com> > wrote: > > On Tue, Mar 31, 2015 at 5:44 AM, Mark Adams <mfadams(a)lbl.gov> wrote: > > I'm finding that my fortran code on OSX has PETSC_COMM_SELF == 0. > > > > I can not find where PETSC_COMM_SELF is created. > > > > Its set in PetscInitializeFortran(). Did you call that? > > > > Matt > > > > > > Mark > > > > > > > > -- > > What most experimenters take for granted before they begin their > experiments is infinitely more interesting than any results to which their > experiments lead. > > -- Norbert Wiener > > > > > > > > > > -- > > What most experimenters take for granted before they begin their > experiments is infinitely more interesting than any results to which their > experiments lead. > > -- Norbert Wiener > >

1 0

Re: [petsc-dev] restricting info
by Mark Adams 31 Mar '15

31 Mar '15

Better, but still pretty loud. But this is not bad: 13:18 merge_em ~/Codes/epsi/XGC-1_examples/em_tearing$ mpirun -n 16 ../../XGC1_3/xgc-em |grep GAMG [0] PCGAMGProlongator_AGG(): New grid 4400 nodes [0] PCGAMGOptProlongator_AGG(): Smooth P0: max eigen=1.932173e+00 min=2.303153e-03 PC=jacobi [0] PCGAMGProlongator_AGG(): New grid 736 nodes [0] PCGAMGOptProlongator_AGG(): Smooth P0: max eigen=1.656603e+00 min=7.944341e-03 PC=jacobi [0] PCGAMGCreateLevel_GAMG(): Aggregate processors noop: new_size=16, neq(loc)=45 [0] PCGAMGProlongator_AGG(): New grid 80 nodes [0] PCGAMGOptProlongator_AGG(): Smooth P0: max eigen=1.683069e+00 min=6.834695e-02 PC=jacobi [0] PCGAMGCreateLevel_GAMG(): Number of equations (loc) 4 with simple aggregation [0] PCGAMGProlongator_AGG(): New grid 8 nodes [0] PCGAMGOptProlongator_AGG(): Smooth P0: max eigen=1.435867e+00 min=5.953441e-01 PC=jacobi [0] PCGAMGCreateLevel_GAMG(): Number of equations (loc) 4 with simple aggregation [0] PCSetUp_GAMG(): 5 levels, grid complexity = nan I am missing output that I used to have with "verbose". Will see if I can add some of it back in. Thanks, Mark On Tue, Mar 31, 2015 at 11:39 AM, Tobin Isaac <tisaac(a)ices.utexas.edu> wrote: > > Is `-info -info_exclude null,vec,mat` still too noisy? > > On Tue, Mar 31, 2015 at 11:10:35AM -0400, Mark Adams wrote: > > Now that 'verbose' is not in GAMG anymore, how can I get something > similar? > > that is not super noisy. > > Mark >

1 0

Re: [petsc-dev] Patch for ksp tutorial ex56
by Patrick Sanan 31 Mar '15

31 Mar '15

I’m using a branch off of master on a fork , but I’ll grab those changes from petsc/petsc/master - thanks! Il giorno 31/mar/2015, alle ore 06:03 PM, Hong <hzhang(a)mcs.anl.gov> ha scritto: > Patrick: > We have replaced this segment of code (reading a data file) with initialized C arrays in petsc master branch. Are you using petsc-release version? > > Hong > > I tested this out and the strange behavior is very likely because the C string “sp" is not null-terminated, so there is undefined behavior when using snprintf . On different compilers this adds some extra characters (sometimes newlines as below, but arbitrary characters in general). Replacing sp with a null-terminated string fixed the problem for me: > > diff --git a/src/ksp/ksp/examples/tutorials/ex56.c b/src/ksp/ksp/examples/tutorials/ex56.c > index 16278bb..edccd03 100644 > --- a/src/ksp/ksp/examples/tutorials/ex56.c > +++ b/src/ksp/ksp/examples/tutorials/ex56.c > @@ -136,9 +136,10 @@ int main(int argc,char **args) > { > FILE *file; > char fname[256]; > - char sp[1]; > + char sp[2]; > const char* petsc_dir = getenv("PETSC_DIR"); > sp[0] = PETSC_DIR_SEPARATOR; > + sp[1]='\0'; > if (!petsc_dir) { > SETERRQ(PETSC_COMM_SELF,PETSC_ERR_USER,"'PETSC_DIR' not set."); > } > > Il giorno 11/mar/2015, alle ore 02:12 AM, Hong <hzhang(a)mcs.anl.gov> ha scritto: > >> Sascha: >> I'll replace *.txt file with a C-array to simply ex56.c. >> Please ignore my previous request. >> >> Hong >> >> Hong, >> I do not have an immediate explanation for this. In particular, I wonder why >> there is no line break after 'sp=/' in this line: >> > file = 0x0, sp=/,fname=/Users/Hong/soft/petsc/ >> but line breaks in the file path. >> >> On my machine I tested with make runex56, too. I will look at this again tomorrow. >> Can anyone else confirm this issue? >> >> Sascha >> >> >> >> On 10 Mar 2015, at 22:22, Hong <hzhang(a)mcs.anl.gov> wrote: >> >> > Sascha: >> > I prepared a patch regarding the reading of the data file in ksp tutorial ex56. >> > It sets the path to the data file required for running the example relative to PETSC_DIR >> > instead of relative to the tutorials folder to allow running the example from any folder. >> > >> > Also, I used PETSC_DIR_SEPARATOR instead of the currently hard coded ‘/‘ >> > so that the code would also run on a windows machine … if anyone cared. >> > This fails on my MacPro by running 'make runex56': >> > >> > file = 0x0, sp=/,fname=/Users/Hong/soft/petsc/ >> > src/ >> > ksp/ >> > ksp/ >> > examples/ >> > tutorials/ >> > data/ >> > elem_3d_elast_v_25.txt >> > main failed to open input file '/Users/Hong/soft/petsc/ >> > >> > I added >> > file = fopen(fname, "r"); >> > printf("file = %p, sp=%s,fname=%s\n",file,sp,fname); >> > if (file == 0) { >> > >> > as you see, 'sp=/' is followed by a 'new line'. >> > I do not see this behavior on other machines, e.g., linux or an old iMac. >> > Any suggestion? >> > >> > Hong >> > >> > >> > Please discard the patch if you do not want to change the current behavior. >> > >> > Best, >> > Sascha >> > >> > >> >> > >

1 0