[petsc-dev] Moving from BG/L to BG/P

[Satish Balay]

On Mon, 17 Oct 2011, Kevin.Buckley at ecs.vuw.ac.nz wrote:

> > Just to clarify - petsc-bgl-tools were useful for the first generation
> > of BlueGene [bg/l] with the first generation of software stack - where
> > mpixlc/mpixlf77/mpixlf90 wrappers to use IBM compilers with MPI on
> > these machines were absent.
> >
> > Subsequent machines [bg/p] and perhaps software stack update to bg/l
> > do provide proper mpixlc/mpixlf77/mpixlf90 warppers. [so
> > petsc-bgl-tools isn't needed anymore]
> >
> > As Jed suggested - use config/examples/arch-bgp-ibm-opt.py - modify it
> > appropriately for the externalpackages requred - and run it.
> >
> > Satish
> >
> 
> Thanks to both of you, for those pointers and clarification.
> 
> Initially, I was thinking that the researcher would want to
> stick with the 3.0 series so had not looked at the 3.2 sources,
> and hence not finding the
> 
>  config/examples/arch-bgp-ibm-opt.py
> 
> mentioned, however, we are going to move with the times.
> 
> I thus altered the external package options to match those of the
> 3.0.0-p9 build in a copy of that file and ran it.
> 
> I get some odd output when trying to pull in the latest
> SuperLU, which I grabbed from ANL
> 
>   '--download-superlu=/path/to/superlu_4.0-March_7_2010.tar.gz',
> 
> It starts to compile but then fails with this output
> 
> ...
> mpixlc_r -O3 -qarch=450d -qtune=450 -qmaxmem=-1  -DNoChange
> -DUSE_VENDOR_BLAS -c ilu_heap_relax_snode.c
> mpixlc_r -O3 -qarch=450d -qtune=450 -qmaxmem=-1  -DNoChange
> -DUSE_VENDOR_BLAS -c mark_relax.c
> c mc64ad.f
> c mc21.f
> ...
> make[1]: c: Command not found
> 
> Looking at
> 
>  externalpackages/SuperLU_4.0/SRC/Makefile
> 
> suggests to me that the default rule for
> 
> .f.o:
>         $(FORTRAN) $(FFLAGS) -c $<
> 
> isn't seeing the $(FORTRAN) or $(FFLAGS) expanded.
> 
> Along the right lines ? Or missed something obvious ?

PETSc-3.2 uses http://crd.lbl.gov/~xiaoye/SuperLU/superlu_4.2.tar.gz


Satish