AMBER Archive (2006)

Subject: Re: AMBER: building AMBER9 with gfortran/cygwin

From: David A. Case (case_at_scripps.edu)
Date: Fri Aug 25 2006 - 11:39:40 CDT

On Fri, Aug 25, 2006, Andreas Svrcek-Seiler wrote:

>
> Here's a workaround:
>
> 4) (related to point 2): ifort doesn't compile lmod.f after
> applying this patch.

Oops, I missed this the first time around. So I really can't recommend this
patch in general. Here's what amber10 (development) does:

182c182,183
< character(len=len(MVPM_FORWARD)) :: matrix_vector_product_method=MVPM_FORWARD 

---
>    character(len=7) :: matrix_vector_product_method=MVPM_FORWARD ! gfortran hack
194,195c195,197
<    character(len=len(MC_METHOD_QUICK_QUENCH)) :: Monte_Carlo_method &
<                                                      = MC_METHOD_METROPOLIS
---
>    character(len=12) :: Monte_Carlo_method = MC_METHOD_METROPOLIS
257c259,260
<    character(len=len(XMIN_METHOD_LBFGS)) :: xmin_method = XMIN_METHOD_LBFGS
---
>    character(len=5) :: xmin_method = XMIN_METHOD_LBFGS

Pretty ugly, but should work on all compilers.  My guess is that this is a
limitation/bug in gfortran, since all other compilers accept it, but I can't
find anything official that states that len() of a parameter string needs to
be considered a constant.  Someone should volunteer to clean up the code.

...dac

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu