Lab 7 & 8 - Parallel/GPU computing

Introduction¶

Welcome to the lab for the last two lectures - Parallel & GPU computing. It will entail some independent work on the software you intend to use the cluster for. This task will account for the last two labs, making it the last task you have to submit.

Task¶

You will have to choose a software to analyze and also run using some method of parallelisation on the Rocket cluster. If you don't know which software to choose, the easier options are for example mpi4py and PyTorch. For example, PyTorch could be executed with just parallel processing, with GPUs, or with both. There are good tutorials for these including steps for execution. For PyTorch you can also find an example in our documentation. If you're having doubt about choosing something, ask us.

Since the software packages you will be selecting can be very different, there is no fixed agenda for this lab. Take your time and if you happen to have trouble, ask in Slack. For more general discussions, the labs will be able to help you best.

The task will be to answer the questions found in this pdf. Most of the questions will require you to read the documentation of the software you are interested in. In the end there is also a practical section. You will set up and execute your code.

Submission¶

There is no automatic testing and the last day for submission is January 31. However, we recommend that you try to submit your report by the end of December. This way, we have enough time to check everything and fix the problems should there be any. The reports will be graded as they come in, so submit earlier to get approved earlier. You can either answer the questions at the corresponding sections by editing the .docx file or in free form, but specify which point your answering to. The submission should be done on the courses page.