This is a test version of Biostars. For the public version, visit https://www.biostars.org.
DIAMOND: Does DIAMOND support openmp?

I try to run DIAMOND on Portable Batch System(PBS), and find that when I try to use 50 cpus, the command are executed 50 times. So I wonder if DIAMOND support openmp? PS: it doesn't effect the outcome.

diamond openmp

DIAMOND does not support OpenMP AFAIK. You can use the --threads option to use more than one process in parallel. Can you post the exact command line you are using in your PBS script? Not sure what DB you are searching against but DIAMOND requires significant amount of RAM.

Ok, here are my pbsscript and command:

pbs script:

  #!/bin/bash

  #PBS -N meg

  #PBS -l nodes=1:ppn=50

  #PBS -j oe

  #PBS -q   smp

cd  $PBS_O_WORKDIR

echo my job id is $PBS_JOBID | tee  $PBS_JOBID.log

echo run nodes is following: | tee -a $PBS_JOBID.log

cat $PBS_NODEFILE | tee  -a  $PBS_JOBID.log

source  /public/home/software/mpi/openmpi.sh

echo begin time is `date` | tee -a  $PBS_JOBID.log

id=`echo $PBS_JOBID|awk -F. '{print $1}' `

mpirun /public/home/blast | tee -a openmpi.log 

echo end time is `date` | tee -a  $PBS_JOBID.log


**command line:**

diamond blastx -p 16 --query test.fasta --db diamond_database/nr.dmnd --taxonlist 2 --daa test2.daa

Is there anything I can do to make this quicker?

Thanks

First of all submitting to a smp runtime is not doing anything useful. You can just submit to a regular non-smp runtime.

PBS -l nodes=1:ppn=50

Looks like you are requesting 50 cores but then in your DIAMOND command line you are only using 16 threads. Make sure those two numbers match.

I don't see a specific memory request. Assign at least 80G (that is what I remember using) for the DIAMOND job using nr database.

PBS -l mem=80G

Depending on number of your input sequences it could take hours or day+ for the entire job to complete. There is not much you can do to speed this up (besides allocating plenty of cores and memory). Make sure there is an output file being written that keeps growing. If you wish to parse the data later then choose an appropriate (tabular is useful) format.

I am not sure why you have this in your script.

mpirun /public/home/blast | tee -a openmpi.log

That does not have anything to do with DIAMOND.

0 answers

No answers yet.

Log in to answer this question.