>> It is assumed that the rank 0 process has access to the
complete
>> database so that it can calculate the correct effective
lengths for the entire DB
>- How the rank 0 process knows about to complete database,
if it has been fragmented? Or rank 0 process pass just parameters of length
complete DB?
All of the database fragments are accessible to rank0. It is a requirement of
mpiblast that the full database accessible to rank0. It needs this information
to compute the 'effective search space'. This effective search space is what is
passed to the workers. While related, this is different than simply passing the
length of the complete db.
> - What the PATCH does basically on NCBI Toolbox? Conceptually…
You can see the patch here. It is quite small.
http://mpiblast.cvs.sourceforge.net/mpiblast/mpiblast/ncbi_May2006_evalue.patch?revision=1.1&view=markup
The most important part is some extra fields is a key data structure in
blastdef.h. These are used to allow mpiblast to decide when it only has to do
rank0 tasks vs worker tasks, and additional statistics for the complete
database vs an individual fragment. The rest of the changes are related to
honoring those fields, or collecting and using the extra statistics.
>> For some workloads, the search space calculation can be rather time-consuming
>- About calculate to
effective length. Is it doing to all query sequence? Because this is onerous?
I'm afraid I don't understand this portion. Please explain again? I suspect
this also relates to the 'effective search space'.
--
Mike Cariaso * Bioinformatics Software * http://www.cariaso.com
-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
_______________________________________________
Mpiblast-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/mpiblast-users