ADL: autodock 3.05 on linux cluster?

Demet Akten dakten at halic.edu.tr
Wed Nov 21 09:08:16 PST 2007


Yes, that's right. The type of parallel processing I am talking about  
is not a "real" one. As far as I know, there is no algorithm  
implemented inside Autodock source code that divides the task into  
smaller tasks (for example energy evaluations) and process them in  
parallel. Then it would be possible to give one job which would use  
more than one processor. However, unlike Molecular Dynamics, a  
docking job always consists of multiple "independent" runs and you  
can always divide them. The only difficulty is to combine the results  
afterwards.


On 21.Kas.2007, at 18:18, Daniel Jana wrote:

> On 21/11/2007, Demet Akten <dakten at halic.edu.tr> wrote:
>> I have seen the benefit of parallel processing for docking large  
>> ligands
>> such as cyclosporin with around 15 degrees of freedom. Note that my
>> version of parallel processing is to divide a total of 100 runs  
>> into 10
>> different jobs.
>
> But, you end up running each of the 10 jobs in one processor, right?
> So, in total you end up using 10 processors but each job uses only
> one. This is no better than using a 2000 processor cluster to do
> screening with one job per ligand at a time. This is not parallel
> processing per se as there is no job running in more than one
> processor. You have different jobs running in different processors.
>
> Did I understand correctly what you said or am I making any mistake?
>
> Daniel
> ________________________________________________
> --- ADL: AutoDock List  --- http://autodock.scripps.edu/ 
> mailing_list ---



More information about the autodock mailing list