This page introduces some examples of practical job scripts when submitting a batch job request to SX-ACE.
Mastering perfect description is not necessary. Please reuse these sample scripts by modifying areas you want to change.
The scripts below are categorized based on the purposes of the users. Please reuse files of your interest. You will need to modify places with "( )".
Please read carefully the following page before this page:
How to describe a job script
Please check the following items again.
1.Does "#!/bin/csh" appear on the very first line of the file?
2. Does the last line of the file have line feed code?
The above item 1 and this item are mandatory.
3.The above item 1 and this item are mandatory.
4.Is there any extra space in your script ile?
The typical example is like the space right after "2," in
"#PBS -l cpunum_job=2, elapstim_req=0:10:00".
This kind of extra space causes an error.
5.Did you correctly specify the amount of computational resource？
Each computational resources has the upper limitation in resource use.
Please see job class table for each resource.
6. In the case that your problem is not solved in spite of the trials above..
Please contact us from system@cmc.osaka-u.ac.jp.
Ex. 1: Execution on a single node
perform one-hour computation using a single SX-ACE node
Page useful when writing this script:How to describe a job script
|
1 2 3 4 5 |
#!/bin/csh #PBS –q ACE # execution queue #PBS –l elapstim_req=1:00:00,memsz_job=60GB # amount of resources cd $PBS_O_WORKDIR # current directory ./a.out # executable module |
Ex. 2: Using intranode parallelism (auto-parallelism)
Perform a one-hour 4 thread processing using auto-parallelism
This page is useful when writing this script: How to use auto-parallelism(SX-ACE)
|
1 2 3 4 5 6 |
#!/bin/csh #PBS -q ACE # execution queue #PBS -l elapstim_req=1:00:00,memsz_job=60GB # amount of resources #PBS -v F_RSVTASK=4 # number of threads for intra-node parallelism cd $PBS_O_WORKDIR # current directory ./a.out # executable module |
Ex3: Using intra-node parallelism (OpenMP)
perform one-hour 4 thread processing using OpenMP
This page is useful when writing this script: How to use OpenMP(SX-ACE)
|
1 2 3 4 5 6 |
#!/bin/csh #PBS -q ACE # execution queue #PBS -l elapstim_req=1:00:00,memsz_job=60GB # amount of resources #PBS -v OMP_NUM_THREADS=4 # number of threads for intra-node parallelism cd $PBS_O_WORKDIR # current directory ./a.out # executable module |
Ex4: Using MPI/SX (8 node x 4 core parallel processing)
perform one-hour 32 parallel processing using MPI
This page is useful when writing this script: How to use MPI/SX(SX-ACE)
|
1 2 3 4 5 |
#!/bin/csh #PBS -q ACE # execution queue #PBS -l elapstim_req=1:00:00,memsz_job=60GB # amount of resources #PBS -T mpisx # mandatory for SX/MPI #PBS -b 8 # number of nodes which you want to use |
Ex5: Using HPF/SX ( 4 node x 4 core parallel processing)
Perform one-hour 16 parallel processing using HPF with 4 SX-ACE nodes.
The HPF program is translated into an MPI executable module after compilation and so the job script file is the same as in the case of MPI.
This page is useful when writing this script: How to use HPF/SX(SX-ACE)
|
1 2 3 4 5 6 7 8 9 |
#!/bin/csh #PBS -q ACE #PBS -l memsz_job=60GB,elapstim_req=5:00:00 #PBS -b 4 # number of nodes #PBS -T mpisx # mandatory for MPI/SX setenv MPIPROGINF DETAIL # requesting performance information of MPI/SX program cd $PBS_O_WORKDIR mpirun -nn 4 -np 16 ./a.out # number of node after -nn, number of processes after -np |
Ex 6: Using MPI/SX(HPF/SX) and intranode parallelism (auto-parallelism) (8 node x 4 core)
Perform one-hour 32 parallel processing using ace 8 nodes x 4 cores.
(4 intra-node parallelism + 8 inter-node parallelism )
This page is useful when writing this script: Inter-node parallelism and intra-node parallelism
|
1 2 3 4 5 6 7 8 9 |
#!/bin/csh #PBS -q ACE #PBS -l elapstim_req=1:00:00,memsz_job=60GB #PBS -T mpisx # mandatory for SX/MPI #PBS -b 8 # number of nodes #PBS -v F_RSVTASK=4 # number of threads for auto-parallelism setenv MPIPROGINF DETAIL cd $PBS_O_WORKDIR mpirun -nn 8 -np 8 ./a.out # -nn（number of nodes) -np (number of MPI processes) |
Ex7: Using MPI/SX(HPF/SX) and intra-node parallelism (OpenMP) (8 node x 4 core)
Perform one-hour 32 parallel processing using SX-ACE 8 nodes x 4 cores.
(4 intra-node parallelism(OpenMP) + 8 inter-node parallelism (MPI(HPF)) )
This page is useful when writing this script: Inter-node parallelism and intra-node parallelism
|
1 2 3 4 5 6 7 8 9 |
#!/bin/csh #PBS -q ACE #PBS -l elapstim_req=1:00:00,memsz_job=60GB #PBS -T mpisx # mandatory for SX/MPI #PBS -b 8 # number of nodes #PBS -v OMP_NUM_THREADS=4 # number of OpenMP Threads setenv MPIPROGINF DETAIL cd $PBS_O_WORKDIR mpirun -nn 8 -np 8 ./a.out # -nn（number of nodes） -np (number of MPI processes) |
