[gmx-developers] Implicit solvent + Still radii + sse2_double = segmentation fault

Mark Abraham Mark.Abraham at anu.edu.au
Wed Nov 2 06:03:31 CET 2011


On 2/11/2011 3:41 PM, Julius Su wrote:
> Hi everyone,
>
> I observed segmentation fault errors in Gromacs 4.5.5 when the 
> following options are enabled:
>
> implicit_solvent = GBSA
> gb_algorithm = Still
>
> and additionally the sse2_double optimized kernel 
> (src/mdlib/genborn_sse2_double.c) is used.
>
> After some investigation, I believe that line 1199 of 
> src/mdlib/genborn.c should be changed from
>      calc_gb_rad_still_sse2_double(cr,fr,md->nr,top, atype, x[0], nl, 
> born);
> to
>     calc_gb_rad_still_sse2_double(cr,fr,born->nr,top, atype, x[0], nl, 
> born);
> so that the correct number of atoms is passed to the Born radius 
> function.
>
> This eliminates the segmentation fault and enables the program to 
> function properly.

Thanks, the fix looks right and I have added it to the repository.

Mark



More information about the gromacs.org_gmx-developers mailing list