Your First Program on Cori
To run a program on Cori you must either compile your own code or use a pre-built application. In the following example, we'll guide you through compiling and running the "Hello World" code shown below. Cori has two kinds of compute nodes: "Intel Haswell" and "Intel Xeon Phi". This pages refers to the Haswell nodes only, for simplicity.
"Hello World" Example
Using your favorite text editor (e.g. vi, emacs) open a new file called "helloWorld.f90". You can copy the example below by using "view source" button that will appear in the upper right hand corner of the example when you move your cursor there and paste the contents into your file.
integer :: myPE, numProcs, ierr
call MPI_COMM_RANK(MPI_COMM_WORLD, myPE, ierr)
call MPI_COMM_SIZE(MPI_COMM_WORLD, numProcs, ierr)
print *, "Hello from Processor ", myPE
end program helloWorld
Compile the Program
Use the compiler "wrappers" to compile codes on Cori: use ftn for Fortran, cc for C, and CC for C++. The following example will compile the text file "helloWorld.f90" into the binary executable "helloWorld".
% ftn -o helloWorld helloWorld.f90
Run the Program
To run a code on Cori you (1) submit a request to the batch system (SLURM at NERSC); and (2) launch your job on to the compute nodes using the 'srun' command. There are two ways to submit a request to the batch system: (1) You can request an interactive session using the salloc command or your can submit a batch script (see below). (Note that there is no 'mpirun' command – which is used by many MPI implementations – on Cori.)
Create a Batch Script
Create and open a text file called my_batch_script with a text editor like vi or emacs and paste in the contents below. The batch script is used to tell the Cori system to reserve compute node resources for your job and how it should launch your application. In this example, two nodes are requested in the Cori Haswell partition and runs 64 MPI tasks spread over the two 32-core nodes.
#!/bin/bash -l #SBATCH -p debug #SBATCH -N 2
#SBATCH -C haswell #SBATCH -t 00:10:00 #SBATCH -J my_job srun -n 64 ./helloWorld
Submit Your Job to the Queue
The sbatch command is used to submit your batch script to run your code on the Cori compute nodes.
% sbatch my_batch_script
A job number will be returned, such as 13479.
Monitor Your Job in the Queue
After you submit your job, the system scheduler will check to see if there are compute nodes available to run it. If there are compute nodes available, your job will start running. If there are not, your job will wait in the queue until there are enough resources to run your application. You can monitor your position in the queue using the squeue command:
cori% squeue -u username
For more information, see Monitoring Jobs.
Examine Your Job's Output
When your job has completed you should see a file called slurm-jobid.out that contains:
Hello from Processor 1
Hello from Processor 2
Hello from Processor 10
Hello from Processor 8
Hello from Processor 11
Hello from Processor 3
Hello from Processor 9
Hello from Processor 4
Hello from Processor 6
Hello from Processor 5
Hello from Processor 7
Hello from Processor 12
Hello from Processor 17
Hello from Processor 20
Hello from Processor 14
Hello from Processor 22
Hello from Processor 15
Hello from Processor 16
Hello from Processor 19
Hello from Processor 18
Hello from Processor 0 etc.
Accessing job information
To access detailed information about your job, use the scontrol command with your jobid number:
% scontrol show job jobid
This will give you a detailed output, such as:
Priority=1003 Nice=0 Account=(null) QOS=normal
JobState=COMPLETED Reason=None Dependency=(null)
Requeue=1 Restarts=0 BatchFlag=1 Reboot=0 ExitCode=0:0
RunTime=00:00:06 TimeLimit=00:10:00 TimeMin=N/A
PreemptTime=None SuspendTime=None SecsPreSuspend=0
NumNodes=2 NumCPUs=128 CPUs/Task=1 ReqB:S:C:T=0:0:*:*
Socks/Node=* NtasksPerN:B:S:C=0:0:*:* CoreSpec=*
MinCPUsNode=1 MinMemoryNode=122G MinTmpDiskNode=0
Features=(null) Gres=craynetwork:1 Reservation=(null)
Shared=0 Contiguous=0 Licenses=(null) Network=(null)
For more information about SLURM please see the SLURM pages.