November 2011 - petsc-dev - mailman3.cels.anl.gov

snes/examples/tutorials/ex27.c
by Jed Brown 15 Nov '11

15 Nov '11

http://petsc.cs.iit.edu/petsc/petsc-dev/rev/c3192eb7cd93 I would rather not delete this example because it is the example code accompanying a paper that I like. I wrote essentially the same thing in ts/examples/tutorials/ex26.c (which uses TSPseudo), but having the simple, executable example that exactly reproduces the figure in the paper is still useful. Using TS is somewhat less clear because you have to look at library code to see the algorithm, but I could also live with simply documenting how to get the same results with ts/.../ex26.c. Preference? As a general rule, many elliptic problems arise as the steady state of some dynamical system. If we write the code using the TS interface, then we can always solve the steady problem by taking an infinite step size with nonlinearly implicit Euler. Should we recommend that all such problems be written using the TS interface?

1 0

Re: [petsc-dev] src/ksp/ksp/examples/tutorials/ex31.c do we need?
by Barry Smith 14 Nov '11

14 Nov '11

On Nov 14, 2011, at 10:27 PM, Jed Brown wrote: > On Mon, Nov 14, 2011 at 22:25, Barry Smith <bsmith(a)mcs.anl.gov> wrote: > Do we still want src/ksp/ksp/examples/tutorials/ex31.c do we need? Or with all the cool new TS stuff is this no longer needed? > > Some people just really like to write everything out. I think it's valuable to have at least one example like that. Ok, needs to conversion away from DMMG

1 0

Re: [petsc-dev] src/ksp/ksp/examples/tutorials/ex31.c do we need?
by Jed Brown 14 Nov '11

14 Nov '11

On Mon, Nov 14, 2011 at 22:25, Barry Smith <bsmith(a)mcs.anl.gov> wrote: > Do we still want src/ksp/ksp/examples/tutorials/ex31.c do we need? Or > with all the cool new TS stuff is this no longer needed? > Some people just really like to write everything out. I think it's valuable to have at least one example like that.

1 0

src/ksp/ksp/examples/tutorials/ex31.c do we need?
by Barry Smith 14 Nov '11

14 Nov '11

Do we still want src/ksp/ksp/examples/tutorials/ex31.c do we need? Or with all the cool new TS stuff is this no longer needed? Barry

1 0

Re: [petsc-dev] Generating Assembly code for PETSc routines
by Jed Brown 14 Nov '11

14 Nov '11

On Mon, Nov 14, 2011 at 16:44, Matthew Knepley <knepley(a)gmail.com> wrote: > That will display the build command. Then just cut&paste, inserting your > -S flag. Alternatively, you can disassemble the function body, either with "objdump -d --prefix-addresses" or with gdb's "disassemble".

1 0

Re: [petsc-dev] Generating Assembly code for PETSc routines
by Matthew Knepley 14 Nov '11

14 Nov '11

On Mon, Nov 14, 2011 at 10:43 PM, Chekuri S. Choudary < cchoudary(a)rnet-tech.com> wrote: > I was wondering how to make use of the compiler options within the PETSc > build system to generate the assembly code for the selected PETSc routines. > **** > > I tried CFLAGS=–S flag (gcc compiler) in the makefile in subfolders but > that doesn’t output anything.**** > > ** ** > > I tried –S flag in the configure options as following but that didn’t work. > **** > > ** ** > > ./configure --download-f-blas-lapack=1 --with-x=0 --with-debugging=1 > --CFLAGS="-O3 -g -ggdb –S -fopenmp" --FFLAGS="-O3 -g -fopenmp" > --download-hypre=1 --with-hdf5=1 --download-hdf5=1 --with-batch=0 > --known-mpi-shared-libraries=no --with-cuda=1 --with-cudac="nvcc -m64" > --with-cusp=1 --with-thrust=1 PETSC_ARCH=PFLOTRAN1 > Easiest way: cd <source dir> touch <source file> make That will display the build command. Then just cut&paste, inserting your -S flag. Matt > > > ** > > Thanks**** > > Shiva**** > > ** ** > -- 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

Generating Assembly code for PETSc routines
by Chekuri S. Choudary 14 Nov '11

14 Nov '11

I was wondering how to make use of the compiler options within the PETSc build system to generate the assembly code for the selected PETSc routines. I tried CFLAGS=-S flag (gcc compiler) in the makefile in subfolders but that doesn't output anything. I tried -S flag in the configure options as following but that didn't work. ./configure --download-f-blas-lapack=1 --with-x=0 --with-debugging=1 --CFLAGS="-O3 -g -ggdb -S -fopenmp" --FFLAGS="-O3 -g -fopenmp" --download-hypre=1 --with-hdf5=1 --download-hdf5=1 --with-batch=0 --known-mpi-shared-libraries=no --with-cuda=1 --with-cudac="nvcc -m64" --with-cusp=1 --with-thrust=1 PETSC_ARCH=PFLOTRAN1 Thanks Shiva

1 0

Re: [petsc-dev] no manual pages for DM subclasses
by Matthew Knepley 14 Nov '11

14 Nov '11

On Mon, Nov 14, 2011 at 7:25 PM, Barry Smith <bsmith(a)mcs.anl.gov> wrote: > > None of these beasties have manual pages. > > #define DMDA "da" > #define DMADDA "adda" > #define DMCOMPOSITE "composite" > #define DMSLICED "sliced" > #define DMMESH "mesh" > #define DMCARTESIAN "cartesian" > #define DMIGA "iga" > #define DMREDUNDANT "redundant" > I will do DMMESH, DMCARTESIAN, AND DMIGA. Matt -- 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] DMCartesian vs DMMesh?
by Jed Brown 14 Nov '11

14 Nov '11

On Mon, Nov 14, 2011 at 13:28, Barry Smith <bsmith(a)mcs.anl.gov> wrote: > DMSIEVE, makes perfect sense, like PCSPAI How about DMXQZ because it sounds faster? I see no problem with naming software components like motorcycles. I think DM names should reflect the interface that users see when using them to build the algebraic ingredients like residual and Jacobian. Sieve is certainly not the "standard" unstructured grid interface, so I think it can have a non-generic name, even if it is completely general.

1 0

Re: [petsc-dev] DMCartesian vs DMMesh?
by Barry Smith 14 Nov '11

14 Nov '11

On Nov 14, 2011, at 1:25 PM, Matthew Knepley wrote: > On Mon, Nov 14, 2011 at 7:21 PM, Barry Smith <bsmith(a)mcs.anl.gov> wrote: > > On Nov 14, 2011, at 1:08 PM, Matthew Knepley wrote: > > > On Mon, Nov 14, 2011 at 6:18 PM, Barry Smith <bsmith(a)mcs.anl.gov> wrote: > > > > > > > > Also since Mesh is a generic word but DMMesh is built on Sieve should DMMesh be renamed to DMSieve? > > > > I don't like that, because no one will understand what we are talking about. DMMesh makes sense. I might be > > alright with DMUnstructuredMesh, but that is overkill. > > > > Matt > > DMMesh is kind of like having a PCPRECONDITIONER > > I hear the criticism. I don't hear the alternative. DMSIEVE, makes perfect sense, like PCSPAI Clearly we need documentation on DM and a flow chart as to possible choices for different types of problems to lead people to the right DM. Barry > What other word do people use for an unstructured mesh? > > Matt > > > Barry > > > > > > > Barry > > > > -- > > 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