On Dec 8, 2012, at 12:11 PM, Jed Brown <jedbrown at mcs.anl.gov> wrote:
> This commit says "rm unused variables" but its primary purpose appears to be > to add non-compliant debugging code. > > https://bitbucket.org/petsc/petsc-dev/commits/01c3be9a4ea1 > > I still don't understand why we aren't using PETSc events to profile this, If PETSc events do not provide enough useful information for optimizing code then we should think about what can be added/changed to PETSc events and log summary to so that they can be used. Perhaps alternative -log_summary's that do not print as much stuff etc? Barry > but can we all please strive to check in only compliant code? You can always > make a personal fork on bitbucket, commit all the debugging fragments you > want, then clean up your patches with histedit, and push to petsc/petsc-dev. > > > /home/jed/petsc/src/mat/impls/aij/seq/matmatmult.c:37:24: warning: format > specifies type 'int' but the argument has type 'PetscInt' (aka 'long long') > [-Wformat] > printf(" Mat %d %d, 2MultSymbolic_SeqAIJ_Scalable time: > %g\n",A->rmap->N,A->cmap->N,t1-t0); > ~^ > ~~~~~~~~~~ > %lld > /home/jed/petsc/src/mat/impls/aij/seq/matmatmult.c:37:27: warning: format > specifies type 'int' but the argument has type 'PetscInt' (aka 'long long') > [-Wformat] > printf(" Mat %d %d, 2MultSymbolic_SeqAIJ_Scalable time: > %g\n",A->rmap->N,A->cmap->N,t1-t0); > ~^ > ~~~~~~~~~~ > %lld > /home/jed/petsc/src/mat/impls/aij/seq/matmatmult.c:46:24: warning: format > specifies type 'int' but the argument has type 'PetscInt' (aka 'long long') > [-Wformat] > printf(" Mat %d %d, 2MultSymbolic_SeqAIJ time: > %g\n",A->rmap->N,A->cmap->N,t1-t0); > ~^ ~~~~~~~~~~ > %lld > /home/jed/petsc/src/mat/impls/aij/seq/matmatmult.c:46:27: warning: format > specifies type 'int' but the argument has type 'PetscInt' (aka 'long long') > [-Wformat] > printf(" Mat %d %d, 2MultSymbolic_SeqAIJ time: > %g\n",A->rmap->N,A->cmap->N,t1-t0); > ~^ > ~~~~~~~~~~ > %lld > 4 warnings generated. > [ 11%] Building CXX object > CMakeFiles/petsc.dir/src/mat/impls/aij/mpi/mpiptap.c.o > clang: warning: treating 'c' input as 'c++' when in C++ mode, this behavior > is deprecated > /home/jed/petsc/src/mat/impls/aij/seq/matmatmatmult.c:61:20: warning: format > specifies type 'int' but the argument has type 'PetscInt' (aka 'long long') > [-Wformat] > printf(" Mat %d %d, 3MultSymbolic_SeqAIJ_Scalable time: %g + %g = > %g\n",A->rmap->N,A->cmap->N,t1-t0,t2-t1,t2-t0); > ~^ > ~~~~~~~~~~ > %lld > /home/jed/petsc/src/mat/impls/aij/seq/matmatmatmult.c:61:23: warning: format > specifies type 'int' but the argument has type 'PetscInt' (aka 'long long') > [-Wformat] > printf(" Mat %d %d, 3MultSymbolic_SeqAIJ_Scalable time: %g + %g = > %g\n",A->rmap->N,A->cmap->N,t1-t0,t2-t1,t2-t0); > ~^ > ~~~~~~~~~~ > %lld > /home/jed/petsc/src/mat/impls/aij/seq/matmatmatmult.c:68:20: warning: format > specifies type 'int' but the argument has type 'PetscInt' (aka 'long long') > [-Wformat] > printf(" Mat %d %d, 3MultSymbolic_SeqAIJ time: %g + %g = > %g\n",A->rmap->N,A->cmap->N,t1-t0,t2-t1,t2-t0); > ~^ ~~~~~~~~~~ > %lld > /home/jed/petsc/src/mat/impls/aij/seq/matmatmatmult.c:68:23: warning: format > specifies type 'int' but the argument has type 'PetscInt' (aka 'long long') > [-Wformat] > printf(" Mat %d %d, 3MultSymbolic_SeqAIJ time: %g + %g = > %g\n",A->rmap->N,A->cmap->N,t1-t0,t2-t1,t2-t0); > ~^ > ~~~~~~~~~~ > %lld > 4 warnings generated. >
