On 2012-01-24, at 2:35 PM, David A Case wrote:
> On Tue, Jan 24, 2012, Robin Betz wrote:
>>
>> Currently AMBER serial compilation builds yacc, which is then used in the
>> parallel compilation. I believe that this is the only reason that serial
>> needs to be build before parallel.
>> Should yacc be built by AMBER at all? Most distributions have the byacc
>> package that will provide it. It could be added as a dependency instead and
>> removed from the AMBER source.
>
> I don't understand why yacc can't just be a dependency of the parallel
> build; I thought(?) that had already been done. Yacc is trivial to build,
> and I'd rather use the Amber version than rely on a (unknown version of)
> byacc that might be available on "most distributions".
I had been working on this recently but got distracted with other issues.
byacc and ucpp both need to be added to the parallel build, which should not be a problem. I have done this but not pushed it yet.
The next problem is with PBSA. AFAIK, the situation currently is
-PBSA.MPI is the only program that uses MPI PBSA and there is no MPI PBSA library.
-the PBSA Makefile refuses to build in parallel if the serial libraries have not already been built.
-both sander.MPI and mpinab link the serial PBSA library.
If MPI PBSA is not going to be used in sander.MPI or mpinab, then I suggest that PBSA functionality be removed from these programs and the PBSA Makefile can proceed with an MPI build without a prior serial build. Users can always compile the serial versions to get PBSA functionality.
Tyler
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Tue Jan 24 2012 - 13:30:04 PST
