Hi Mengjuei,
What you describe might work but I would like to avoid copying module files at all. There are no conflicts with module names for sander. I just tested this and it works with gfortran. Getting rid of copying the .mod files just requires replacement of
LOCALFLAGS = $(FREEFORMAT_FLAG)
with
LOCALFLAGS = $(FREEFORMAT_FLAG) -I../../AmberTools/src/pbsa -I../../AmberTools/src/sqm \
-I../../AmberTools/src/rism -I../../include
in sander's Makefile. We really should be using this which is why I was asking yesterday for people to test whether this works on platforms I do not have access to. If you have access to something else we are supporting apart from GNU, Intel or PGI, please try to compile the test I sent and let me know whether it works, it just takes a minute.
We could eventually choose one common place where module files are put when the source is compiled, for example $AMBERHOME/include. But this requires a little more thinking and testing. I believe that also in this case we woldn't have conflicts with module names. If there would be conflicts it probably means that module names are not chosen wisely and should changed or that we have code duplications or helper routines that should end up in one module.
All the best,
Andy
On May 19, 2011, at 11:33 AM, Mengjuei Hsieh wrote:
> Hi,
>
> It seems to me that Amber used to support PathScale as a compiler suite,
> so what was the suggestion by then to deal with module filename with
> different cases? For example, a module file in gfortran/intel fortran
> like qmmm_module.mod would become QMMM_MODULE.mod in Open64 and PathScale.
>
> Reading from the archive of the developers' list, the solution was simply
> cp orig/*.mod dest/ , but that doesn't look very robust. The other way
> to do it is using "-I" flag, but I can imagine some cases of conflicted
> module names.
>
> Can we do "find" and then "xargs" in Makefile? They are both POSIX
> commands, but I am not sure how portable the commands are. Basically it
> would look like:
>
> find ../../AmberTools/src/sqm -iname qmmm_module.mod \
> -o -iname qmmm_vsolv_module.mod -o -iname qmmm_vsolv_module.mod \
> -o -iname qmmm_struct_module.mod -o -iname qmmm_nml_module.mod \
> -o -iname qmmm_qmtheorymodule.mod -o -iname qmmm_struct_module.mod \
> -o -iname elementorbitalindex.mod | xargs -I xxx cp xxx .
>
> Ideas? :p
>
> Regards,
> --
> Mengjuei Hsieh, Molecular Bio & Biochem, University of California Irvine
>
> _______________________________________________
> AMBER-Developers mailing list
> AMBER-Developers.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber-developers
--
Dr. Andreas W. Goetz
San Diego Supercomputer Center
Tel : +1-858-822-4771
Email: agoetz.sdsc.edu
Web : www.awgoetz.de
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Thu May 19 2011 - 15:00:03 PDT
