AMBER Archive (2006)

Subject: Re: AMBER: Installation amber 9 on IBM SP4

From: Scott Brozell (sbrozell_at_scripps.edu)
Date: Fri Oct 06 2006 - 12:13:15 CDT

Hi,

I was able to compile build_amoeba on an IBM. So this will not become
a bug fix without further info on the failing platform.
Here's where build_amoeba compiled:

Oct 06 12:03:00pm cu12 256> ~/amber/amber9/src/build_amoeba uname -a
AIX cu12 3 5 0024BB0A4C00
Oct 06 12:07:22pm cu12 257> ~/amber/amber9/src/build_amoeba which make
/usr/bin/make
Oct 06 12:07:28pm cu12 258> ~/amber/amber9/src/build_amoeba make -v
make: Not a recognized flag: v
usage: make [-eiknqrst] [-k|-S] [-d[A|adg[1|2]msv]] [-D variable] [-f makefile ] [-j [jobs]] [variable=value ...] [target ...]
Oct 06 12:07:33pm cu12 259> ~/amber/amber9/src/build_amoeba xlf90 -qversion
 IBM XL Fortran Enterprise Edition V9.1
 Version: 09.01.0000.0003

Oct 06 12:07:36pm cu12 260> ~/amber/amber9/src/build_amoeba make -n
        /lib/cpp -P -DMASSLIB -DNMLEQ -DCLINK_PLAIN -DXLF90 amoeba_parm.f90 > amoeba_parm.f
        xlf90 -c -qfixed -O3 -qmaxmem=-1 -qarch=auto -qtune=auto -c -qfree=f90 -o amoeba_parm.o amoeba_parm.f
        xlf90 -c -qfixed -O3 -qmaxmem=-1 -qarch=auto -qtune=auto -c cspline.f
        xlf90 -o amoeba_parm amoeba_parm.o cspline.o -L/usr/lib -lmassv -L/usr/lib -lblas
        /lib/cpp -P -DMASSLIB -DNMLEQ -DCLINK_PLAIN -DXLF90 new_crd_to_dyn.f90 > new_crd_to_dyn.f
        xlf90 -c -qfixed -O3 -qmaxmem=-1 -qarch=auto -qtune=auto -c -qfree=f90 -o new_crd_to_dyn.o new_crd_to_dyn.f
        cd ../lib; make nxtsec.o
        xlf90 -o new_crd_to_dyn new_crd_to_dyn.o ../lib/nxtsec.o -L/usr/lib -lmassv -L/usr/lib -lblas
        /lib/cpp -P -DMASSLIB -DNMLEQ -DCLINK_PLAIN -DXLF90 new_to_old_crd.f90 > new_to_old_crd.f
        xlf90 -c -qfixed -O3 -qmaxmem=-1 -qarch=auto -qtune=auto -c -qfree=f90 -o new_to_old_crd.o new_to_old_crd.f
        xlf90 -o new_to_old_crd new_to_old_crd.o ../lib/nxtsec.o -L/usr/lib -lmassv -L/usr/lib -lblas
        /lib/cpp -P -DMASSLIB -DNMLEQ -DCLINK_PLAIN -DXLF90 fix_new_inpcrd_vel.f90 > fix_new_inpcrd_vel.f
        xlf90 -c -qfixed -O3 -qmaxmem=-1 -qarch=auto -qtune=auto -c -qfree=f90 -o fix_new_inpcrd_vel.o fix_new_inpcrd_vel.f
        xlf90 -o fix_new_inpcrd_vel fix_new_inpcrd_vel.o ../lib/nxtsec.o -L/usr/lib -lmassv -L/usr/lib -lblas
Target "all" is up to date.

Scott

On Thu, 24 Aug 2006, Scott Brozell wrote:

> Hi,
>
> On Thu, 24 Aug 2006, Scott Brozell wrote:
>
> > Hi,
> >
> > On Thu, 24 Aug 2006, Myunggi Yi wrote:
> >
> > > The compilation stopped.
> > > The following is the end of the screen output.
> > >
> > > =====================================
> > > cd build_amoeba; make install
> > > /lib/cpp -P -DMASSLIB -DNMLEQ -DCLINK_PLAIN -DXLF90 > amoeba_parm.f
> > >
> > > What's going on?
> > > I waited for tens of minuites, but nothing happened.
> >
> > You are waiting because cpp is expecting input that is missing, namely
> > amoeba_parm.f90. The command should be:
> >
> > /lib/cpp -P -DMASSLIB -DNMLEQ -DCLINK_PLAIN -DXLF90 amoeba_parm.f90 > amoeba_parm.f
> >
> > It looks like $< is not being evaluated correctly.
>
> It may be that $< is only valid in an inference rule for a picky
> make, such as ibm's. In which we may have to bugfix the Makefile.
> On a technical note for future Ambers using .F90 instead of .f90
> would also avoid this issue if ibm supports the popular conventions
> on uppercase extensions indicating preprocessing.
>
> > >From build_amoeba/Makefile
> > amoeba_parm.o: amoeba_parm.f90
> > $(FPP) $< > $*.f
> > $(FC) -c $(FOPTFLAGS) $(FREEFORMAT_FLAG) -o $@ $*.f
> >
> > Note that config.h has a suffixes for .f90
> > I'll sleep on it; someone with access to an ibm may wake before me :)
> >
> >
> > You can fix this and the others like it by editing the Makefile:
> >
> > amoeba_parm.o: amoeba_parm.f90
> > $(FPP) $< > $*.f
> > $(FC) -c $(FOPTFLAGS) $(FREEFORMAT_FLAG) -o $@ $*.f
> >
> > ->
> >
> > amoeba_parm.o: amoeba_parm.f90
> > $(FPP) $*.f90 > $*.f
> > $(FC) -c $(FOPTFLAGS) $(FREEFORMAT_FLAG) -o $@ $*.f
> >
> > etc.
>
> Note, do a make clean after you edit the Makefile.
>
> > Good luck,
> > Scott
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu