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

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

23 Apr '12

Sure, one sec. On Mon, Apr 23, 2012 at 4:29 PM, Matthew Knepley <knepley(a)gmail.com> wrote: > On Mon, Apr 23, 2012 at 5:27 PM, Daniel Lowell <redratio1(a)gmail.com>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: >> > > Can you send the whole make.log to petsc-maint? > > Matt > > >> #$ 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 >> >> > > > -- > 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 > -- Daniel Lowell

1 0

Re: [petsc-dev] 4.2 and PETSc
by Matthew Knepley 23 Apr '12

23 Apr '12

On Mon, Apr 23, 2012 at 5:27 PM, Daniel Lowell <redratio1(a)gmail.com> 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: > Can you send the whole make.log to petsc-maint? Matt > #$ 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 > > -- 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

MetSetBlockSize issue
by Mark F. Adams 23 Apr '12

23 Apr '12

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] Cross compile with .c and .cu files
by Matthew Knepley 23 Apr '12

23 Apr '12

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 \ > > > > > > > > > > > > > > -- 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

4.2 and PETSc
by Daniel Lowell 23 Apr '12

23 Apr '12

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] Cross compile with .c and .cu files
by Satish Balay 23 Apr '12

23 Apr '12

'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 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

Cross compile with .c and .cu files
by Daniel Lowell 23 Apr '12

23 Apr '12

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 \ -- Daniel Lowell

1 0

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

23 Apr '12

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

23 Apr '12

It's a mistake. I'll push a fix. - Peter On Mon, Apr 23, 2012 at 3:09 PM, Stephan Kramer <s.kramer(a)imperial.ac.uk>wrote: > 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

petsc-dev build error in petscsnes.h90
by Stephan Kramer 23 Apr '12

23 Apr '12

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