Ok, I fixed this. Here is what happened. I have compiled Ray with OpenMPI. But run it with mpirun from MPICH (because there is no command to switch implementation in my submission script). This resulted in no processors used. At the same time I used pre-compiled (with MPICH) version of my test program with the same submission script. It was run with mpirun from MPICH and used a lot of processors.
After setting submission script to use mpirun from OpenMPI, everything is fine. Thank you and sorry for taking your time. 2011/9/15 Sébastien Boisvert <[email protected]>: > Hello, > > Your command looks fine to me. > > > > Can you provide the PBS submission script for the simple program you > described ? > > > > p.s. sourceforge blocked your original message as it was too large -- it was > larger than 40000 bytes. > > p.s. 2: you should paste your logs on pastebin.com and link it in your mail > instead of attaching a large file. > > > Sébastien > > http://github.com/sebhtml/ray > >> ________________________________________ >> De : Fedor Gusev [[email protected]] >> Date d'envoi : 15 septembre 2011 09:54 >> À : [email protected] >> Objet : Ray uses only one processor? >> >> Hello every one. >> >> It seems like Ray does not use all the processors I want it to use. >> >> I am running Ray with script, which I submit to Torque >> >> ----------------------------- >> #PBS -m ea >> #PBS -M [email protected] >> #PBS -lnodes=3:ppn=8 >> >> . /data/software/exports >> >> cd $PBS_O_WORKDIR >> >> mpirun -np 24 Ray -o assembly_59 -k 59 -s short.fastq >> ----------------------------- >> >> >> The output is from Ray contains this: >> ----------------------------- >> Number of MPI ranks: 1 >> Ray master MPI rank: 0 >> Ray slave MPI ranks: 0-0 >> ----------------------------- >> >> It seems Ray does not pick up all 24 processors. >> >> A simple program called the exact same way (with mpirun and Torque) >> uses 24 processors: >> ----------------------------- >> int main(int argc, char *argv[]) { >> int numprocs, rank, namelen; >> char processor_name[MPI_MAX_PROCESSOR_NAME]; >> >> MPI_Init(&argc, &argv); >> MPI_Comm_size(MPI_COMM_WORLD, &numprocs); >> MPI_Comm_rank(MPI_COMM_WORLD, &rank); >> MPI_Get_processor_name(processor_name, &namelen); >> >> printf("Process %d on %s out of %d\n", rank, processor_name, numprocs); >> >> MPI_Finalize(); >> } >> ----------------------------- >> >> What can be the problem? >> >> -- >> Kind regards, >> Fedor Gusev. >> -- Kind regards, Fedor Gusev. ------------------------------------------------------------------------------ Doing More with Less: The Next Generation Virtual Desktop What are the key obstacles that have prevented many mid-market businesses from deploying virtual desktops? How do next-generation virtual desktops provide companies an easier-to-deploy, easier-to-manage and more affordable virtual desktop model.http://www.accelacomm.com/jaw/sfnl/114/51426474/ _______________________________________________ Denovoassembler-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/denovoassembler-users
