[gmx-developers] Possible bug with particle decomposition

[Berk Hess]

On 12/21/2010 03:55 PM, Ondrej Marsalek wrote:
> On Tue, Dec 21, 2010 at 15:45, Berk Hess <hess at cbr.su.se> wrote:
>   
>> I fixed this bug.
>>     
> Thank you.
>
>   
>> I don't think it's really important though, domain decomposition should
>> be faster
>> than particle decomposition.
>>     
> It is, even for small systems, when I am able to run it all.
>
>   
>> Or do you need it for something special?
>>     
> Nothing special, but sometimes there is trouble for small systems
> (buffer region sizes), especially during equilibration. Certainly
> nothing critical. I just thought that it might be of interest for code
> quality reasons, it could have been just an indication of something
> more serious.
>
> Ondrej
>   
No, it was simply an issue that MPI does not allow NULL pointers, even
for arrays of size 0.

But it's indeed good to report such things because of the reason you
mentioned.

Berk