April 2012 - petsc-dev - mailman3.cels.anl.gov

Re: [petsc-dev] MetSetBlockSize issue
by Barry Smith 23 Apr '12

23 Apr '12

Yes, look, for example how ex32 is handled at the bottom of the makefile. Thanks Barry On Apr 23, 2012, at 5:50 PM, Mark F. Adams wrote: > > On Apr 23, 2012, at 5:59 PM, Barry Smith wrote: > >> >> It was changed a while ago that MatSetBlockSize() couldn't be set after the matrix was full instantiated. I guess that example did not get fixed because it is not listed in the makefile to run in the makeall. > > May I add it? > >> >> I have fixed the example to call MatSetBlockSize() at a safe point. >> >> Barry >> >> On Apr 23, 2012, at 4:29 PM, Mark F. Adams wrote: >> >>> ex55.c in ksp is failing with: >>> >>> [0]PETSC ERROR: --------------------- Error Message ------------------------------------ >>> [0]PETSC ERROR: Arguments are incompatible! >>> [0]PETSC ERROR: Cannot change block size 1 to 2! >>> [0]PETSC ERROR: ------------------------------------------------------------------------ >>> >>> on this line 57 of ex55.c: >>> >>> ierr = MatSetBlockSize(Amat,2); CHKERRQ(ierr); >>> >>> Any idea what happened here? >>> >>> Mark >> >> >

1 0

Detecting MPI_C_DOUBLE_COMPLEX when clanguage!='C'
by Jed Brown 23 Apr '12

23 Apr '12

if self.getDefaultLanguage() == 'C': mpitypes.extend([('MPI_C_DOUBLE_COMPLEX', 'c-double-complex')]) Why don't we check for this regardless of C language? The C and C++ types are binary compatible and there is an advantage to using the predefined data types: it lets us also use the predefined MPI_Ops, which would let us use MPIU_SUM with PetscSF (and possibly use more optimized reduction implementations in other cases).

1 0

Re: [petsc-dev] MetSetBlockSize issue
by Mark F. Adams 23 Apr '12

23 Apr '12

On Apr 23, 2012, at 5:59 PM, Barry Smith wrote: > > It was changed a while ago that MatSetBlockSize() couldn't be set after the matrix was full instantiated. I guess that example did not get fixed because it is not listed in the makefile to run in the makeall. May I add it? > > I have fixed the example to call MatSetBlockSize() at a safe point. > > Barry > > On Apr 23, 2012, at 4:29 PM, Mark F. Adams wrote: > >> ex55.c in ksp is failing with: >> >> [0]PETSC ERROR: --------------------- Error Message ------------------------------------ >> [0]PETSC ERROR: Arguments are incompatible! >> [0]PETSC ERROR: Cannot change block size 1 to 2! >> [0]PETSC ERROR: ------------------------------------------------------------------------ >> >> on this line 57 of ex55.c: >> >> ierr = MatSetBlockSize(Amat,2); CHKERRQ(ierr); >> >> Any idea what happened here? >> >> Mark > >

1 0

Re: [petsc-dev] petsc-dev build error in petscsnes.h90
by Peter Brune 23 Apr '12

23 Apr '12

The fix is pushed. - Peter On Mon, Apr 23, 2012 at 3:14 PM, Alexander Grayver <agrayver(a)gfz-potsdam.de>wrote: > I also have this problem and solve it by editing this file manually. > > Regards, > Alexander > > > ----- Reply message ----- > From: "Stephan Kramer" <s.kramer(a)imperial.ac.uk> > To: "For users of the development version of PETSc" <petsc-dev(a)mcs.anl.gov > > > Subject: [petsc-dev] petsc-dev build error in petscsnes.h90 > Date: Mon, Apr 23, 2012 22:09 > > > Hi all, > > When building petsc-dev, configured --with-fortran-interfaces=1, I had to > add the following lines to include/finclude/petscsnesdef.h: > > diff -r fc934c1d84bc include/finclude/petscsnesdef.h > --- a/include/finclude/petscsnesdef.h Mon Apr 23 11:39:32 2012 -0500 > +++ b/include/finclude/petscsnesdef.h Mon Apr 23 21:02:14 2012 +0100 > @@ -12,6 +12,13 @@ > #define SNESType character*(80) > #define SNESMSType character*(80) > #define SNESConvergedReason PetscEnum > +#define SNESQNRestartType PetscEnum > +#define SNESQNCompositionType PetscEnum > +#define SNESQNScaleType PetscEnum > +#define SNESNCGType PetscEnum > +#define SNESNGMRESRestartType PetscEnum > +#define SNESNGMRESSelectType PetscEnum > +#define SNESNormType PetscEnum > #define SNESLineSearchType PetscEnum > #define MatMFFD PetscFortranAddr > #define MatMFFDType PetscFortranAddr > > to avoid compile errors in include/finclude/ftn-auto/petscsnes.h90 > (included from include/finclude/petscsnes.h90). Is this a bug, or did > something else go wrong? > > Cheers > Stephan >

1 0

Re: [petsc-dev] Cross compile with .c and .cu files
by Satish Balay 23 Apr '12

23 Apr '12

Ah - libfast is written to work with mixed builds. But there is abug with sharing a variable with .c and .cu builds rules. Pushed a fix now. The makefile below should now be: SOURCECU = diakerns.cu SOURCEC = dia.c OBJSCU = diakerns.o OBJSC = dia.o Satish On Mon, 23 Apr 2012, Matthew Knepley wrote: > On Mon, Apr 23, 2012 at 5:26 PM, Satish Balay <balay(a)mcs.anl.gov> wrote: > > > 'make libfast' can't handle different file types [but 'make lib' can]. > > > > For libfast to work - you can split up the sources into multiple dirs. > > > > i.e > > > > src/mat/impls/dia/seq/dia.c > > src/mat/impls/dia/seq/seqcusp/diakerns.cu > > > The Python build can handle this. > > Matt > > > > > > Satish > > > > On Mon, 23 Apr 2012, Daniel Lowell wrote: > > > > > Hi all, > > > > > > I'm building a matrix type with two separate files dai.c and diakerns.cu > > . > > > If I "make all" in the PETSc root dir, it will be unable to find the > > object > > > file from the .cu file; i.e., I get this: > > > > > > /usr/bin/ar: diakerns.o: No such file or directory > > > > > > However if I merely do a "make lib" inside the matrix type directory I > > have > > > no problem. This has happened to other types I've create, I end up just > > > shoving everything into one large .cu file to quickly deal with it. > > > Here what the folder level makefile looks like: > > > > > > > > > > > > SOURCECU = diakerns.cu > > > SOURCEC = dia.c > > > SOURCEF = > > > SOURCEH = dia.h > > > OBJSC = dia.o diakerns.o > > > OBJSF = > > > LIBBASE = libpetscmat > > > DIRS = > > > MANSEC = Mat > > > LOCDIR = src/mat/impls/dia/seq/ > > > > > > > > > > > > > > > This is what my configuration script for petsc root looks like: > > > > > > > > > MPI_DIR="--with-mpi-dir=$MPICH2_HOME" > > > LAD="--download-f-blas-lapack=yes" > > > CUD="--with-cuda-dir=/soft/cuda-4.2/cuda" > > > ./config/configure.py $MPI_DIR $LAD $CUD \ > > > --with-debugging=1 \ > > > --with-cudac="nvcc -m64 -g -G -v" \ > > > --with-precision=double \ > > > --with-clanguage=c \ > > > --with-cuda-arch=sm_20 \ > > > > > > > > > > > > > > > > > > > > > > > > >

1 0

Re: [petsc-dev] MetSetBlockSize issue
by Barry Smith 23 Apr '12

23 Apr '12

It was changed a while ago that MatSetBlockSize() couldn't be set after the matrix was full instantiated. I guess that example did not get fixed because it is not listed in the makefile to run in the makeall. I have fixed the example to call MatSetBlockSize() at a safe point. Barry On Apr 23, 2012, at 4:29 PM, Mark F. Adams wrote: > ex55.c in ksp is failing with: > > [0]PETSC ERROR: --------------------- Error Message ------------------------------------ > [0]PETSC ERROR: Arguments are incompatible! > [0]PETSC ERROR: Cannot change block size 1 to 2! > [0]PETSC ERROR: ------------------------------------------------------------------------ > > on this line 57 of ex55.c: > > ierr = MatSetBlockSize(Amat,2); CHKERRQ(ierr); > > Any idea what happened here? > > Mark

1 0

Re: [petsc-dev] 4.2 and PETSc
by Daniel Lowell 23 Apr '12

23 Apr '12

Sending the make.log to petsc-maint now. On Mon, Apr 23, 2012 at 4:38 PM, Satish Balay <balay(a)mcs.anl.gov> wrote: > but you missed copy/pasting the actual compile command. So incomplete info > here.. > > satish > > On Mon, 23 Apr 2012, Daniel Lowell wrote: > > > I added -v in the local folder's makefile to produce verbose output. > > > > On Mon, Apr 23, 2012 at 4:33 PM, Satish Balay <balay(a)mcs.anl.gov> wrote: > > > > > Hm - we've seen these errors with -g3 but not with -g. Are you sure > > > there is no '-g3' somewhere on the compile command? > > > > > > With latest petsc-dev/buildsystem one should not see -g3 [set by > configure] > > > > > > Also - why is there 'cicc' in a petsc build? > > > > > > Satish > > > > > > On Mon, 23 Apr 2012, Daniel Lowell wrote: > > > > > > > Hi all, > > > > > > > > > > > > Has anyone run into any problems with NVCC 4.1 or .42 on PETSc with > > > > arch=sm_20? > > > > I have having real issues with it. I keep getting this: > > > > > > > > #$ cicc -arch compute_20 -m64 -ftz=0 -prec_div=1 -prec_sqrt=1 > -fmad=1 -g > > > > -O0 "/tmp/tmpxft_00000b8b_00000000-10_vecgpu" > > > > "/tmp/tmpxft_00000b8b_00000000-7_vecgpu.cpp3.i" -o > > > > "/tmp/tmpxft_00000b8b_00000000-2_vecgpu.ptx" > > > > <built-in>(2): error: "__STDC_HOSTED__" is predefined; attempted > > > > redefinition ignored > > > > > > > > <built-in>(8): error: "__WCHAR_TYPE__" is predefined; attempted > > > > redefinition ignored > > > > > > > > <built-in>(115): error: "__x86_64" is predefined; attempted > redefinition > > > > ignored > > > > > > > > <built-in>(116): error: "__x86_64__" is predefined; attempted > > > redefinition > > > > ignored > > > > > > > > <built-in>(126): error: "__linux__" is predefined; attempted > redefinition > > > > ignored > > > > > > > > <built-in>(128): error: "__unix__" is predefined; attempted > redefinition > > > > ignored > > > > > > > > 6 errors detected in the compilation of > > > > "/tmp/tmpxft_00000b8b_00000000-7_vecgpu.cpp3.i". > > > > # --error 0x1 -- > > > > > > > > > > > > 4.0 works fine, no issues. > > > > 4.0 version uses the older nvopencc, in 4.1 and 4.2 they use > something > > > > called cicc, for target architecture sm_2x. Of course it does no > mention > > > > cicc in the updated nvcc documentation at all. > > > > > > > > Tried 4.2 with GCC 4.3 and 4.4 without any difference in behavior. > > > > > > > > > > > > PETSc configuration script: > > > > > > > > MPI_DIR="--with-mpi-dir=$MPICH2_HOME" > > > > LAD="--download-f-blas-lapack=yes" > > > > CUD="--with-cuda-dir=/soft/cuda-4.2/cuda" > > > > ./config/configure.py $MPI_DIR $LAD $CUD \ > > > > --with-debugging=1 \ > > > > --with-cudac="nvcc -m64 -g -G" \ > > > > --with-precision=double \ > > > > --with-clanguage=c \ > > > > --with-cuda-arch=sm_20 > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > -- Daniel Lowell

1 0

Re: [petsc-dev] 4.2 and PETSc
by Satish Balay 23 Apr '12

23 Apr '12

but you missed copy/pasting the actual compile command. So incomplete info here.. satish On Mon, 23 Apr 2012, Daniel Lowell wrote: > I added -v in the local folder's makefile to produce verbose output. > > On Mon, Apr 23, 2012 at 4:33 PM, Satish Balay <balay(a)mcs.anl.gov> wrote: > > > Hm - we've seen these errors with -g3 but not with -g. Are you sure > > there is no '-g3' somewhere on the compile command? > > > > With latest petsc-dev/buildsystem one should not see -g3 [set by configure] > > > > Also - why is there 'cicc' in a petsc build? > > > > Satish > > > > On Mon, 23 Apr 2012, Daniel Lowell wrote: > > > > > Hi all, > > > > > > > > > Has anyone run into any problems with NVCC 4.1 or .42 on PETSc with > > > arch=sm_20? > > > I have having real issues with it. I keep getting this: > > > > > > #$ cicc -arch compute_20 -m64 -ftz=0 -prec_div=1 -prec_sqrt=1 -fmad=1 -g > > > -O0 "/tmp/tmpxft_00000b8b_00000000-10_vecgpu" > > > "/tmp/tmpxft_00000b8b_00000000-7_vecgpu.cpp3.i" -o > > > "/tmp/tmpxft_00000b8b_00000000-2_vecgpu.ptx" > > > <built-in>(2): error: "__STDC_HOSTED__" is predefined; attempted > > > redefinition ignored > > > > > > <built-in>(8): error: "__WCHAR_TYPE__" is predefined; attempted > > > redefinition ignored > > > > > > <built-in>(115): error: "__x86_64" is predefined; attempted redefinition > > > ignored > > > > > > <built-in>(116): error: "__x86_64__" is predefined; attempted > > redefinition > > > ignored > > > > > > <built-in>(126): error: "__linux__" is predefined; attempted redefinition > > > ignored > > > > > > <built-in>(128): error: "__unix__" is predefined; attempted redefinition > > > ignored > > > > > > 6 errors detected in the compilation of > > > "/tmp/tmpxft_00000b8b_00000000-7_vecgpu.cpp3.i". > > > # --error 0x1 -- > > > > > > > > > 4.0 works fine, no issues. > > > 4.0 version uses the older nvopencc, in 4.1 and 4.2 they use something > > > called cicc, for target architecture sm_2x. Of course it does no mention > > > cicc in the updated nvcc documentation at all. > > > > > > Tried 4.2 with GCC 4.3 and 4.4 without any difference in behavior. > > > > > > > > > PETSc configuration script: > > > > > > MPI_DIR="--with-mpi-dir=$MPICH2_HOME" > > > LAD="--download-f-blas-lapack=yes" > > > CUD="--with-cuda-dir=/soft/cuda-4.2/cuda" > > > ./config/configure.py $MPI_DIR $LAD $CUD \ > > > --with-debugging=1 \ > > > --with-cudac="nvcc -m64 -g -G" \ > > > --with-precision=double \ > > > --with-clanguage=c \ > > > --with-cuda-arch=sm_20 > > > > > > > > > > > > > > > > > > > > > >

1 0

Re: [petsc-dev] 4.2 and PETSc
by Daniel Lowell 23 Apr '12

23 Apr '12

I added -v in the local folder's makefile to produce verbose output. On Mon, Apr 23, 2012 at 4:33 PM, Satish Balay <balay(a)mcs.anl.gov> wrote: > Hm - we've seen these errors with -g3 but not with -g. Are you sure > there is no '-g3' somewhere on the compile command? > > With latest petsc-dev/buildsystem one should not see -g3 [set by configure] > > Also - why is there 'cicc' in a petsc build? > > Satish > > On Mon, 23 Apr 2012, Daniel Lowell wrote: > > > Hi all, > > > > > > Has anyone run into any problems with NVCC 4.1 or .42 on PETSc with > > arch=sm_20? > > I have having real issues with it. I keep getting this: > > > > #$ cicc -arch compute_20 -m64 -ftz=0 -prec_div=1 -prec_sqrt=1 -fmad=1 -g > > -O0 "/tmp/tmpxft_00000b8b_00000000-10_vecgpu" > > "/tmp/tmpxft_00000b8b_00000000-7_vecgpu.cpp3.i" -o > > "/tmp/tmpxft_00000b8b_00000000-2_vecgpu.ptx" > > <built-in>(2): error: "__STDC_HOSTED__" is predefined; attempted > > redefinition ignored > > > > <built-in>(8): error: "__WCHAR_TYPE__" is predefined; attempted > > redefinition ignored > > > > <built-in>(115): error: "__x86_64" is predefined; attempted redefinition > > ignored > > > > <built-in>(116): error: "__x86_64__" is predefined; attempted > redefinition > > ignored > > > > <built-in>(126): error: "__linux__" is predefined; attempted redefinition > > ignored > > > > <built-in>(128): error: "__unix__" is predefined; attempted redefinition > > ignored > > > > 6 errors detected in the compilation of > > "/tmp/tmpxft_00000b8b_00000000-7_vecgpu.cpp3.i". > > # --error 0x1 -- > > > > > > 4.0 works fine, no issues. > > 4.0 version uses the older nvopencc, in 4.1 and 4.2 they use something > > called cicc, for target architecture sm_2x. Of course it does no mention > > cicc in the updated nvcc documentation at all. > > > > Tried 4.2 with GCC 4.3 and 4.4 without any difference in behavior. > > > > > > PETSc configuration script: > > > > MPI_DIR="--with-mpi-dir=$MPICH2_HOME" > > LAD="--download-f-blas-lapack=yes" > > CUD="--with-cuda-dir=/soft/cuda-4.2/cuda" > > ./config/configure.py $MPI_DIR $LAD $CUD \ > > --with-debugging=1 \ > > --with-cudac="nvcc -m64 -g -G" \ > > --with-precision=double \ > > --with-clanguage=c \ > > --with-cuda-arch=sm_20 > > > > > > > > > > > > -- Daniel Lowell

1 0

Re: [petsc-dev] 4.2 and PETSc
by Satish Balay 23 Apr '12

23 Apr '12

Hm - we've seen these errors with -g3 but not with -g. Are you sure there is no '-g3' somewhere on the compile command? With latest petsc-dev/buildsystem one should not see -g3 [set by configure] Also - why is there 'cicc' in a petsc build? Satish On Mon, 23 Apr 2012, Daniel Lowell wrote: > Hi all, > > > Has anyone run into any problems with NVCC 4.1 or .42 on PETSc with > arch=sm_20? > I have having real issues with it. I keep getting this: > > #$ cicc -arch compute_20 -m64 -ftz=0 -prec_div=1 -prec_sqrt=1 -fmad=1 -g > -O0 "/tmp/tmpxft_00000b8b_00000000-10_vecgpu" > "/tmp/tmpxft_00000b8b_00000000-7_vecgpu.cpp3.i" -o > "/tmp/tmpxft_00000b8b_00000000-2_vecgpu.ptx" > <built-in>(2): error: "__STDC_HOSTED__" is predefined; attempted > redefinition ignored > > <built-in>(8): error: "__WCHAR_TYPE__" is predefined; attempted > redefinition ignored > > <built-in>(115): error: "__x86_64" is predefined; attempted redefinition > ignored > > <built-in>(116): error: "__x86_64__" is predefined; attempted redefinition > ignored > > <built-in>(126): error: "__linux__" is predefined; attempted redefinition > ignored > > <built-in>(128): error: "__unix__" is predefined; attempted redefinition > ignored > > 6 errors detected in the compilation of > "/tmp/tmpxft_00000b8b_00000000-7_vecgpu.cpp3.i". > # --error 0x1 -- > > > 4.0 works fine, no issues. > 4.0 version uses the older nvopencc, in 4.1 and 4.2 they use something > called cicc, for target architecture sm_2x. Of course it does no mention > cicc in the updated nvcc documentation at all. > > Tried 4.2 with GCC 4.3 and 4.4 without any difference in behavior. > > > PETSc configuration script: > > MPI_DIR="--with-mpi-dir=$MPICH2_HOME" > LAD="--download-f-blas-lapack=yes" > CUD="--with-cuda-dir=/soft/cuda-4.2/cuda" > ./config/configure.py $MPI_DIR $LAD $CUD \ > --with-debugging=1 \ > --with-cudac="nvcc -m64 -g -G" \ > --with-precision=double \ > --with-clanguage=c \ > --with-cuda-arch=sm_20 > > > > >

1 0