On Wed, May 26, 2010, Mirko Cestari wrote:
> we are experiencing some troubles in installing AmberTools 1.4
> on our IBM-AIX system. We are compiling using xlc and xlf compilers
> after having modified accordingly the configure file.
>
> The installation procedure has been working correctly so far (apart
> from some minor issues like commas in "write(6,*), ...." lines) but
> now we are stuck because of this problem:
>
> >xlf90_r -c -O0 -ffree-form -qsuffix=cpp=f -o qm2_pm6_hof_module.o
> _qm2_pm6_hof_module.f
> "_qm2_pm6_hof_module.f", line 75.23: 1514-163 (S) Procedures specified
> in the same generic
> interface must all be functions or all be subroutines. Procedure or
> binding hofcorrectiongradient
> will not be included in the pm6_correction generic interface.
>
> Indeed file $AMBERHOME/AmberTools/src/sqm/_qm2_pm6_hof_module.f
> contains an interface in which there are a function and a subroutine
>
> ====
> interface pm6_correction
> module procedure hofCorrection <- function
> module procedure hofCorrectionGradient <- subroutine
> end interface
> ====
>
> Is there a way we can have this problem fixed?
Andy: is there any good reason to go through an interface here, which just
seems to obfuscate what is going on anyway? That is, is there any reason why
qm2_energy.f and qm2_get_qm_forces.f should not call hofCorrection() and
hofCorrectionGradient() directly, rather than trying to pretend like this is
C++ when it isn't?
We don't really support AIX anymore (lack of machines in developers' hands)
but if simple things can be done to make it work, we are willing to try.
See if the attached patch (apply it in amber11/AmberTools/src/sqm) helps.
(But it may be a losing cause if this only leads to the next problem....)
...dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sat May 29 2010 - 20:00:03 PDT
