[petsc-dev] MPI I/O tests failing in master after merging my branch

[Matthew Knepley]

On Thu, Mar 22, 2018 at 8:28 AM, Jed Brown <jed at jedbrown.org> wrote:

> Matthew Knepley <knepley at gmail.com> writes:
>
> > On Thu, Mar 22, 2018 at 7:09 AM, Jed Brown <jed at jedbrown.org> wrote:
> >
> >> Matthew Knepley <knepley at gmail.com> writes:
> >>
> >> > On Thu, Mar 22, 2018 at 6:15 AM, Lisandro Dalcin <dalcinl at gmail.com>
> >> wrote:
> >> >
> >> >> On 22 March 2018 at 12:51, Matthew Knepley <knepley at gmail.com>
> wrote:
> >> >> > On Thu, Mar 22, 2018 at 3:44 AM, Lisandro Dalcin <
> dalcinl at gmail.com>
> >> >> wrote:
> >> >> >>
> >> >> >>
> >> >> >> ftp://ftp.mcs.anl.gov/pub/petsc/nightlylogs/archive/
> >> >> 2018/03/21/examples_master_arch-linux-cxx-cmplx-pkgs-64idx_churn.log
> >> >> >>
> >> >> >> The thing is, PETSc does not currently have a way of mapping a
> plain
> >> C
> >> >> >> `int` type to a PETSc datatype. I'm wiling to take the required
> work
> >> >> >> to fix this issue.
> >> >> >
> >> >> >
> >> >> > Is this not what PetscMPIInt is for?
> >> >> >
> >> >>
> >> >> No, not at all, PetscMPIInt it not even mentioned in the
> PetscDatatype
> >> >> enumeration. The whole PetscDatatype this is kind of messy, and
> fixing
> >> >> it should take some time, and this for sure will no make my h-index
> >> >> explode, right?. For now, mapping MPI_INT -> PETSC_ENUM seems
> >> >> workaround the issue just fine.
> >> >>
> >> >
> >> > What I meant was, we already have a typedef 'PetscMPIInt' which is
> >> supposed
> >> > to reproduce MPI_INT, so why not add PETSC_MPI_INT to the
> PetscDataTypes
> >> > rather than map MPI_INT to PetscEnum?
> >>
> >> https://bitbucket.org/petsc/petsc/issues/188/remove-use-
> of-petscdatatype
> >>
> >
> > I agree, but this was about short term fixing.
>
> I guess I'd rather see a quick workaround if someone needs the
> capability now instead of expanding an interface that we're trying to
> remove.
>

Okay. As far as removing it. Why is it complicated? I have not looked
closely.

What I imagined was just mapping each Petsc typedef (do we list them all
anywhere),
to an MPI type, creating those that are missing (as we do for MPIU_INT). Is
there more
to it than that?

  Thanks,

     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

https://www.cse.buffalo.edu/~knepley/ <http://www.caam.rice.edu/~mk51/>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20180322/0affdde2/attachment-0001.html>