dolfin team mailing list archive

[Barry Smith <petsc-maint@xxxxxxxxxxx>]

> I don't understand. Isn't there always only one single PETSC_ARCH on
> each system you compile PETSc?

   WOW! Matter and anit-matter collide :-)

   The only reason for PETSC_ARCH is to support several different sets
of libraries simultaneously; especially debug and optimized but also
with different compilers etc. It is a rare day in hack that any machine
at ANL has only a single PETSC_ARCH built.

  If we didn't think multiple PETSC_ARCH's were important, we would just
use the "standard open-source" install model.

   Barry


On Tue, 5 Jul 2005, Anders Logg wrote:

> On Tue, Jul 05, 2005 at 02:37:55PM -0500, Matthew Knepley wrote:
> 
> > >>   make <option>   <-->    petsc-config <option>
> > >> 
> > >> and you would will need all the bmake directories.
> > >
> > > Yes, but the bmake directories would not be needed after PETSc has
> > > been installed and the script been generated. Or am I missing something?
> > 
> >   This is important. Unless you want to allow only a single PETSC_ARCH at install
> > time, you would need a separate petsc-config for each arch, or a single complicated
> > one. That is why we use make which handles this both seamlessly, and also dynamically
> > in that new archs are incorporated automatically as they appear, whereas the petsc-config
> > model would require considerable fixup, or a separate script each time.
> > 
> >       Matt
> 
> 
> We also use make to dynamically (as determined by configure at
> compile-time) get all the flags for compilation of DOLFIN, but during
> compilation make spits out these flags to a script.
> 
> /Anders
> 
>