February 2013 - petsc-dev - mailman3.cels.anl.gov

(no subject)
by Barry Smith 15 Feb '13

15 Feb '13

We should probably have a petscconf.h and a petsc-private/petscconfimpl.h wc -l include/*.h include/petsc-private/*.h arch-gnu/include/petscconf.h | sort 102 include/petsccharacteristic.h 113 include/petscbt.h 114 include/petsc-private/petscaxpy.h 122 include/petscblaslapack.h 124 include/petsc-private/logimpl.h 131 include/petsc-private/petscfptimpl.h 140 include/petscctable.h 176 include/petsctime.h 181 include/petsc-private/threadcommimpl.h 182 include/petscblaslapack_c.h 182 include/petscblaslapack_caps.h 185 include/petscdmplex.h 193 include/petscsf.h 218 include/petsc-private/dmimpl.h 219 include/petsc-private/tsimpl.h 223 include/petsc-private/fortranimpl.h 242 include/petsc-private/snesimpl.h 250 include/petscoptions.h 256 include/petscdm.h 262 include/petscblaslapack_uscore.h 266 include/petscdmda.h 276 include/petsc-private/kspimpl.h 282 include/petscblaslapack_stdcall.h 304 include/petscdmmesh.h 314 include/petscdraw.h 334 include/petsc-private/vecimpl.h 451 include/petscmath.h 455 include/petscviewer.h 461 include/petscis.h 533 include/petsclog.h 564 include/petscvec.h 596 include/petscpc.h 649 include/petscksp.h 741 include/petscsnes.h 754 include/petscerror.h 845 include/petscts.h 858 include/petsc-private/petscimpl.h 1036 arch-gnu/include/petscconf.h 1504 include/petsc-private/matimpl.h 1930 include/petscmat.h 2399 include/petscsys.h 20713 total

1 0

Re: [petsc-dev] Forward declarations of types?
by Barry Smith 15 Feb '13

15 Feb '13

Damn Mecurial. I know this won't happen with git On Feb 15, 2013, at 6:10 PM, Jed Brown <jedbrown(a)mcs.anl.gov> wrote: > On Fri, Feb 15, 2013 at 4:31 PM, Barry Smith <bsmith(a)mcs.anl.gov> wrote: > > Pushed. Off to play Pandemic > > Oops. > > https://bitbucket.org/petsc/petsc-dev/commits/3410d5fe5456048e71b071a2162d6… >

1 0

Re: [petsc-dev] Forward declarations of types?
by Jed Brown 15 Feb '13

15 Feb '13

On Fri, Feb 15, 2013 at 4:31 PM, Barry Smith <bsmith(a)mcs.anl.gov> wrote: > > Pushed. Off to play Pandemic > Oops. https://bitbucket.org/petsc/petsc-dev/commits/3410d5fe5456048e71b071a2162d6…

1 0

Re: [petsc-dev] Forward declarations of types?
by Barry Smith 15 Feb '13

15 Feb '13

On Feb 15, 2013, at 4:33 PM, Jed Brown <jedbrown(a)mcs.anl.gov> wrote: > > On Fri, Feb 15, 2013 at 4:15 PM, Barry Smith <bsmith(a)mcs.anl.gov> wrote: > Why petsc-private/${class}types.h ? Why in the private directory. It is fine and harmless to include in the public part? plus it allows users to do the same trick and not include all the PETSc stuff in much of their code if we end up doing it in a systematic way? > > Okay, that's fine. It could also be > > include/petsc/${class}types.h > include/petsc/private/${class}impl.h > > if we wanted to reduce the number of files in plain include/. silly and unneccessary > This is mostly relevant when someone installs with --prefix=/usr. > > > My feeling was that "regular" users could work with a PETSc install where petsc-private wasn't even installed and we just install that so that people can write their own KSP etc etc. With this addition now petsc-private becomes required. > > Anyone who writes a plugin needs petsc-private already, but I get the point.

1 0

Re: [petsc-dev] Forward declarations of types?
by Chetan Jhurani 15 Feb '13

15 Feb '13

Interjecting. Talking of conventions, it is likely that Boost picked it up from the C++ iostream (heavy) and iosfwd (light) standard headers. Chetan From: petsc-dev-bounces(a)mcs.anl.gov [mailto:[email protected]] On Behalf Of Jed Brown Sent: Friday, February 15, 2013 2:51 PM To: For users of the development version of PETSc Subject: Re: [petsc-dev] Forward declarations of types? On Fri, Feb 15, 2013 at 4:39 PM, Jed Brown <jedbrown(a)mcs.anl.gov> wrote: petsc*fwd.h or petsc*types.h? Is there an established convention here? Here's one documented convention. http://lists.boost.org/Archives/boost/2001/07/14827.php

1 0

Re: [petsc-dev] Forward declarations of types?
by Barry Smith 15 Feb '13

15 Feb '13

On Feb 15, 2013, at 4:39 PM, Jed Brown <jedbrown(a)mcs.anl.gov> wrote: > > On Fri, Feb 15, 2013 at 4:27 PM, Matthew Knepley <knepley(a)gmail.com> wrote: > I think we should just double all include files: > > petscksp.h > petsckspfwd.h > > You don't have to know anything about petsc*fwd.h unless you > want lightweight includes, and there is no way to misinterpret what it is. > > petsc*fwd.h or petsc*types.h? Is there an established convention here? types has already been pushed and types it is. Barrys-MacBook-Pro:src barrysmith$ ls /usr/include/*type*.h /usr/include/_types.h /usr/include/ecpgtype.h /usr/include/lber_types.h /usr/include/sqltypes.h /usr/include/_wctype.h /usr/include/inttypes.h /usr/include/nl_types.h /usr/include/runetype.h /usr/include/wctype.h /usr/include/ctype.h /usr/include/isqltypes.h /usr/include/sql3types.h Barrys-MacBook-Pro:src barrysmith$ ls /usr/include/*fwd*.h ls: /usr/include/*fwd*.h: No such file or directory :-) > > I agree that this makes the code much more systematic. It will impact users significantly if we apply this to Mat and Vec, but then include-what-you-use would work precisely.

1 0

Re: [petsc-dev] Forward declarations of types?
by Jed Brown 15 Feb '13

15 Feb '13

On Fri, Feb 15, 2013 at 4:39 PM, Jed Brown <jedbrown(a)mcs.anl.gov> wrote: > petsc*fwd.h or petsc*types.h? Is there an established convention here? Here's one documented convention. http://lists.boost.org/Archives/boost/2001/07/14827.php

1 0

Re: [petsc-dev] Forward declarations of types?
by Jed Brown 15 Feb '13

15 Feb '13

On Fri, Feb 15, 2013 at 4:27 PM, Matthew Knepley <knepley(a)gmail.com> wrote: > I think we should just double all include files: > > petscksp.h > petsckspfwd.h > > You don't have to know anything about petsc*fwd.h unless you > want lightweight includes, and there is no way to misinterpret what it is. > petsc*fwd.h or petsc*types.h? Is there an established convention here? I agree that this makes the code much more systematic. It will impact users significantly if we apply this to Mat and Vec, but then include-what-you-use would work precisely.

1 0

Re: [petsc-dev] Forward declarations of types?
by Jed Brown 15 Feb '13

15 Feb '13

On Fri, Feb 15, 2013 at 4:15 PM, Barry Smith <bsmith(a)mcs.anl.gov> wrote: > Why petsc-private/${class}types.h ? Why in the private directory. It is > fine and harmless to include in the public part? plus it allows users to do > the same trick and not include all the PETSc stuff in much of their code if > we end up doing it in a systematic way? > Okay, that's fine. It could also be include/petsc/${class}types.h include/petsc/private/${class}impl.h if we wanted to reduce the number of files in plain include/. This is mostly relevant when someone installs with --prefix=/usr. > > My feeling was that "regular" users could work with a PETSc install > where petsc-private wasn't even installed and we just install that so that > people can write their own KSP etc etc. With this addition now > petsc-private becomes required. > Anyone who writes a plugin needs petsc-private already, but I get the point.

1 0

Re: [petsc-dev] Forward declarations of types?
by Barry Smith 15 Feb '13

15 Feb '13

Pushed. Off to play Pandemic On Feb 15, 2013, at 4:10 PM, Jed Brown <jedbrown(a)mcs.anl.gov> wrote: > On Fri, Feb 15, 2013 at 3:59 PM, Barry Smith <bsmith(a)mcs.anl.gov> wrote: > > Ok, I did an implementation with petscdrawtypes.h included by petscdraw.h, petscksp.h and petscsnes.h > > Should it go in petsc-private because it generally doesn't make sense for users to include directly? Or stay out of petsc-private because it's "safe" to include if that's what they want? > > not ideal (I hate having lots of little includes around) but it does keep the PetscDrawLG typedef out of petscys.h > > People ok with me pushing? > > Yup, should I try doing the same with petscdm.h? I see it as perhaps the biggest funny dependency exposed by include-graph.sh. > > > Barry > > Currently only PetscDrawLG is in there, since that is the only one needed but others or all could be put there if desired. > > On Feb 15, 2013, at 3:44 PM, Jed Brown <jedbrown(a)mcs.anl.gov> wrote: > > > On Fri, Feb 15, 2013 at 3:38 PM, Jed Brown <jedbrown(a)mcs.anl.gov> wrote: > > > > Heh, I thought I remembered that duplicate typedefs were invalid in C, but gcc happily compiled it without a warning. In fact, they _are_ allowed in C11, but not in our old, crusty version of C. > > > > From the looks of it, the C forum never explicitly intended for it to be invalid, but as a consequence of the classification of typedef in C (typedef is classified differently in C++), it would have needed to be explicitly allowed. C11 added that clause. > > > > http://stackoverflow.com/a/8595191/33208 > > > >

1 0