[petsc-dev] external package ML and 64 bit

[Matthew Knepley]

On Thu, Mar 5, 2015 at 11:35 AM, Barry Smith <bsmith at mcs.anl.gov> wrote:

>
>    Matt,
>
>     Please don't put such fixes into random branches you happen to be
> working on (like knepley/feature-parallel-configure). It is better to go
> directly into their only little branch.
>

I thought it was fine because it was a configure branch.


>     Your fix doesn't work. I have pushed a fix in the branch
> barry/error-on-64bitindices-with-ml and merged into next. Also explained
> why 64 bit indices with ML are not going to happen in ml.py
>

I accepted the email on faith.


>    self.requires32bitint  = 1;  # ml uses a combination of "global"
> indices that can be 64 bit and local indices that are always int therefore
> it is
>                                  # essentially impossible to use ML's 64
> bit integer mode with PETSc's --with-64-bit-indices
>
>    Note: also this is the "Trilinos style" of having these two kinds of
> indices. I debated doing this with PETSc but felt it was too complex with
> little payoff.
>

Do they mix the two kinds of indices in the interface?

   Matt


>   Barry
>
>
>
> > On Mar 5, 2015, at 11:02 AM, Matthew Knepley <knepley at gmail.com> wrote:
> >
> > On Thu, Mar 5, 2015 at 11:01 AM, Barry Smith <bsmith at mcs.anl.gov> wrote:
> >
> >   Matt,
> >
> >    Please ALWAYS list the branch(s)/commit that you did this in when you
> say you "pushed a fix"
> >
> >
> https://bitbucket.org/petsc/petsc/commits/ba4f728f6c00f645cbd73e74de34ed390600afde
> >
> >   Matt
> >
> >   Barry
> >
> > > On Mar 5, 2015, at 9:12 AM, Matthew Knepley <knepley at gmail.com> wrote:
> > >
> > > I pushed a fix.
> > >
> > >    Matt
> > >
> > > On Thu, Mar 5, 2015 at 2:18 AM, Stefano Zampini <
> stefano.zampini at gmail.com> wrote:
> > > PETSc is broken when configured with --download-ml
> --with-64-bit-indices.
> > > ML has support for 64 bit integers (using ML's configure flag
> --with-ml-64bit-integer) but one needs to specify the type of 64 bit
> integer (long long or long or whatever...)
> > >
> > > Better to turn off 64 bit builds of ML before adding support for it
> during configure.
> > >
> > >
> > > Stefano
> > >
> > >
> > >
> > > --
> > > 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
>
>


-- 
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
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20150305/43ba6db8/attachment.html>