DDT
DDT is a parallel debugger with a straightforward graphical interface. After setting up DDT for our environment (more on this below) there are only a few steps to using it:
This will run DDT on the head-node. If you have access to private nodes, you should try to run DDT on those nodes. To do that, insert this step before starting DDT.
- Make sure you are running an X-Windows server on your computer.
- When logging into nyx-login, make sure you have enabled X-forwarding in your ssh client; you can confirm this is working by typing echo $DISPLAY after logging into nyx and you should see something similar to localhost:10.0, if you don't see that, you can't yet start DDT. (For more information on X-servers and X-forwarding on Windows, see CAEN's Technote Introduction to X11.)
- Compile your code with -g to include debugging symbols.
- Load the DDT module by typing: module load ddt
- Start DDT with your code by typing: ddt myprogram then when prompted select the number of MPI processes you want to run and press "Run".
For more information, see the Quickstart Guide for C Programs, Quickstart Guide for Fortran Programs, or the DDT User Guide.
- Start an interactive PBS job including X-forwarding with the command:
qsub -N debug -q myQueue -l nodes=4:ppn=2,walltime=04:00:00 -I -V -X
Running DDT for the first time
When DDT starts for the first time, you'll need to answer a few questions.
After this you will be put into DDT where you can chose the number of processes to run and press "Run"
- DDT cannot open the configuration file
- Choose "Create a new file"
- A "wizard" will start
- Click "Next"
- You will need to chose an MPI Implementation
- Unless you know otherwise, you should choose OpenMPI from the drop-down menu and press "Next"
- The next question is to chose a list of compute nodes to run on
- You should not choose any nodes; to do this select "Do not configure DDT for attaching at this time" click and "Next"
- This is the end of the wizard
- Click "Finish"
