[Debichem-devel] OpenMM package for Debian/Ubuntu
Andreas Tille
andreas at an3as.eu
Thu Jan 23 18:59:23 UTC 2014
Hi Peter,
On Thu, Jan 23, 2014 at 10:43:54AM -0800, Peter Eastman wrote:
> We already have build options to omit particular plugins. But I don't see any value in distributing a pre-built version that omits the CUDA plugins. It would just mislead users into thinking they had a "real" OpenMM install, when in fact they had a crippled version, and then they would wonder why their performance was so bad.
I guess this was a missunderstanding: My suggestion was to provide
both: The free code as base and the non-free plugins in a separate
package. Otherwise everything needs to go to contrib / non-free (which
might make things a bit simpler technically but might not be in the
interest of users who want to link against the library).
In any case it would be interested how to build the source at all -
since as I wrote some hours ago the current build breaks.
Kind regards
Andreas.
> On Jan 23, 2014, at 10:26 AM, Andreas Tille <andreas at an3as.eu> wrote:
>
> > Hi Peter,
> >
> > On Thu, Jan 23, 2014 at 09:59:49AM -0800, Peter Eastman wrote:
> >> Hi Andreas,
> >>
> >>> The first and probably major problem is that the build depends on packages in non-free: nvidia-cuda-toolkit.
> >>
> >> All of the code that relies on CUDA is in plugins that get loaded dynamically at runtime. So CUDA is needed to build it, but not to run it. If CUDA isn't available at runtime, those plugins just don't get loaded.
> >
> > Thanks for the clarification. However, we also need to be buildability
> > inside main to comply with Debian policy. One possible solution would
> > be:
> >
> > 1. provide a build option which excludes non-free code
> > 2. provide a separate code tree (which might build-depend
> > from libopenmm libraries) and creates the needed plugins.
> >
> > Kind regards
> >
> > Andreas.
> >
> >
> > --
> > http://fam-tille.de
>
>
--
http://fam-tille.de
More information about the Debichem-devel
mailing list