Code Parallelization with Python

Wednesday 26 Jun 2019, 13:00 → 17:00 Europe/Berlin

Seminar Room Ground Floor (CSBD)

This tutorial will introduce you to parallalisation techniques in python. Having experience programming python is strongly suggested, but not a necessity. Many of the concepts apply to other programming languages. 

In the course, I will

• provide you with a compute intensive problem
• teach how to find the slowest parts of the code
• calculate an expected improvement when parallelising the code
• parallelize the code to use all cores on one machines
• parallelize the code using cores on different machines
• (if time permits) file-based high throughput computing

In case of questions or concerns, please approach me.

self-paced learning

Feedback at the end of the course

red sticky notes

("something you liked, learned or had fun with")

• was lost a bit during the installation (lost pace of the course)
• would be good to touch on the fundamental differences in the approach for different types of parallelisation
• wonderful to know, we can scale and how to scale but how I want to know when/it I should scale
• can't really think of anything
• could have been useful to have had a high level overview of different types of parallelization methods (distributed vs. memory-based)
• struggled a bit with the parallelization exercise but more due to my lacking python knowledge

green sticky notes 

("something you liked, learned or had fun with")

• very nice concise overview, +1
• useful names introduced "multiprocessing module, MPI module, dask"
• like the format
• never were bored or tired
• snippets in /tmp ... but not right away (can't fall behind with code)
• nice content ... will be helpful for my project
• good to have context
• dask and how easy it is to use it on the cluster
• all hands-on
• helpfulness with questions and doubts