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

bug in BuildSystem check
by Barry Smith 22 Apr '11

22 Apr '11

Matt, In the test for isinf() the compile warns about lack of prototype. BuildSystem should stop at this point and NOT set HAVE_ISINF. Could you please fix this? Thanks Barry sh: gcc -c -o conftest.o -Wall -Wwrite-strings -Wno-strict-aliasing -g3 conftest.c Executing: gcc -c -o conftest.o -Wall -Wwrite-strings -Wno-strict-aliasing -g3 conftest.c sh: Possible ERROR while running compiler: error message = {conftest.c: In function 'main': conftest.c:6: warning: implicit declaration of function 'isinf' conftest.c:6: warning: unused variable 'a' } Source: #include "confdefs.h" #include "conffix.h" #include <math.h> int main() { double b = 2.0; int a = isinf(b); ; return 0; } Pushing language C Popping language C sh: gcc -o conftest -Wall -Wwrite-strings -Wno-strict-aliasing -g3 conftest.o -L/opt/local/lib/gcc/sparc-sun-solaris2.10/4.4.1 -L/opt/local/lib/gcc -L/usr/ccs/lib -L/opt/local/lib -ldl -lgcc_eh -ldl -lsocket -lnsl Executing: gcc -o conftest -Wall -Wwrite-strings -Wno-strict-aliasing -g3 conftest.o -L/opt/local/lib/gcc/sparc-sun-solaris2.10/4.4.1 -L/opt/local/lib/gcc -L/usr/ccs/lib -L/opt/local/lib -ldl -lgcc_eh -ldl -lsocket -lnsl sh: Defined "HAVE_ISINF" to "1"

1 0

Re: [petsc-dev] pushing MASSIVE change to PETSc-dev
by Barry Smith 21 Apr '11

21 Apr '11

On Apr 21, 2011, at 11:29 AM, Ethan Coon wrote: > On Thu, 2011-04-21 at 10:25 -0500, Barry Smith wrote: >> Also you need not do VecDestroy(x->y); x->y = 0; since the destroy >> handles that automatically. > > Is this true in Fortran as well? It should be but test it. Barry > > Ethan > > -- > ------------------------------------ > Ethan Coon > Post-Doctoral Researcher > Applied Mathematics - T-5 > Los Alamos National Laboratory > 505-665-8289 > > http://www.ldeo.columbia.edu/~ecoon/ > ------------------------------------ >

1 0

Re: [petsc-dev] pushing MASSIVE change to PETSc-dev
by Ethan Coon 21 Apr '11

21 Apr '11

On Thu, 2011-04-21 at 10:25 -0500, Barry Smith wrote: > Also you need not do VecDestroy(x->y); x->y = 0; since the destroy > handles that automatically. Is this true in Fortran as well? Ethan -- ------------------------------------ Ethan Coon Post-Doctoral Researcher Applied Mathematics - T-5 Los Alamos National Laboratory 505-665-8289 http://www.ldeo.columbia.edu/~ecoon/ ------------------------------------

1 0

Re: [petsc-dev] pushing MASSIVE change to PETSc-dev
by Barry Smith 21 Apr '11

21 Apr '11

On Apr 21, 2011, at 10:18 AM, Jed Brown wrote: > One consequence of this change is that, in general, it should no longer be necessary to use the construct > > if (x->y) {ierr = YDestroy(x->y);CHKERRQ(ierr)} > > Instead, you should be able to always use > > ierr = YDestroy(&x->y);CHKERRQ(ierr); > > There are still some of the former around, but it should not be used in new code. Correct, I removed all that I found but may have missed a few. Also you need not do VecDestroy(x->y); x->y = 0; since the destroy handles that automatically. I will make another sweep through the code to try to eliminate the unneeded checks. Barry

1 0

Re: [petsc-dev] pushing MASSIVE change to PETSc-dev
by Jed Brown 21 Apr '11

21 Apr '11

One consequence of this change is that, in general, it should no longer be necessary to use the construct if (x->y) {ierr = YDestroy(x->y);CHKERRQ(ierr)} Instead, you should be able to always use ierr = YDestroy(&x->y);CHKERRQ(ierr); There are still some of the former around, but it should not be used in new code.

1 0

Re: [petsc-dev] PetscViewerASCIISynchronizedAllow
by Barry Smith 21 Apr '11

21 Apr '11

On Apr 21, 2011, at 10:11 AM, Jed Brown wrote: > On Thu, Apr 21, 2011 at 17:08, Barry Smith <bsmith(a)mcs.anl.gov> wrote: > I changed the level to intermediate. There is a fortran binding in /src/sys/viewer/impls/ascii/ftn-custom/zfilev.c perhaps there is a bug in it so you cannot see it? > > Then this note could be removed from the man page: > > Fortran Note: > This routine is not supported in Fortran. Fixed. Barry >

1 0

Re: [petsc-dev] PetscViewerASCIISynchronizedAllow
by Barry Smith 21 Apr '11

21 Apr '11

On Apr 21, 2011, at 10:05 AM, Jed Brown wrote: > On Thu, Apr 21, 2011 at 17:03, Jed Brown <jed(a)59a2.org> wrote: > Currently this needs to be called, but is not available from Fortran and is marked as developer level. Is there a plan to make it somehow automatic (I thought the function was introduced because it was expensive to do it automatically), or did someone forget to make Fortran bindings and set the Level? > > And should this have a push and pop model. I realized it when I was half done implemented but was too lazy to do it correctly. Also PetscSynchronizedFlush() should be changed to use the PetscViewerFlowControlStart() paradigm allowing bigger chunks then one message at a time for flow control, see for example VecView_MPI_Binary() again I was too lazy to do that correctly. So someone can do these things or add them to http://www.mcs.anl.gov/petsc/petsc-as/developers/projects.html Barry > Otherwise I still need a getter to do a synchronized view within a library call and leave the viewer in the same state as the user gave it to me in.

1 0

Re: [petsc-dev] PetscViewerASCIISynchronizedAllow
by Jed Brown 21 Apr '11

21 Apr '11

On Thu, Apr 21, 2011 at 17:08, Barry Smith <bsmith(a)mcs.anl.gov> wrote: > I changed the level to intermediate. There is a fortran binding in > /src/sys/viewer/impls/ascii/ftn-custom/zfilev.c perhaps there is a bug in it > so you cannot see it? > Then this note could be removed from the man page: Fortran Note: This routine is not supported in Fortran.

1 0

Re: [petsc-dev] PetscViewerASCIISynchronizedAllow
by Barry Smith 21 Apr '11

21 Apr '11

On Apr 21, 2011, at 10:03 AM, Jed Brown wrote: > Currently this needs to be called, but is not available from Fortran and is marked as developer level. Is there a plan to make it somehow automatic (I thought the function was introduced because it was expensive to do it automatically), or did someone forget to make Fortran bindings and set the Level? I changed the level to intermediate. There is a fortran binding in /src/sys/viewer/impls/ascii/ftn-custom/zfilev.c perhaps there is a bug in it so you cannot see it? Barry

1 0

Re: [petsc-dev] PetscViewerASCIISynchronizedAllow
by Jed Brown 21 Apr '11

21 Apr '11

On Thu, Apr 21, 2011 at 17:03, Jed Brown <jed(a)59a2.org> wrote: > Currently this needs to be called, but is not available from Fortran and is > marked as developer level. Is there a plan to make it somehow automatic (I > thought the function was introduced because it was expensive to do it > automatically), or did someone forget to make Fortran bindings and set the > Level? And should this have a push and pop model. Otherwise I still need a getter to do a synchronized view within a library call and leave the viewer in the same state as the user gave it to me in.

1 0