search

cyverse / edu

Education & Outreach
Artistic License 2.0
2 stars 2 forks source link

How do you train scientists to parallelize their code? #3

Open bjoyce3 opened 7 years ago

bjoyce3 commented 7 years ago

One of the most common issues in science is how to scale code (either new or legacy). It would be nice to have a short course or something to that effect that speaks to this need.

Thoughts

  1. Functional programming for easier parallelization
  2. Writing functional code from the get go for beginners
  3. Best practices for functional coding shell -> Python/R
  4. Assessment of code
  5. How to benchmark code to understand the resources that it uses
  6. Parallelization techniques that can be used for
jcoliver commented 7 years ago

This would be awesome. One implementation could walk through un-scalable code (or code that would be challenging to scale up, perhaps lots of hard-coded things), to transform it to something scalable. This could probably be done at least with points 1-5.