Parallel programming with Python exercises
On Midway3, with an interactive session with --ntasks-per-node=8 or more
module load python/anaconda-2021.05 openmpi/4.1.2+gcc-7.4.0
source activate parallel
ulimit -l unlimited
If you would like to clone the env parallel, first export
the packages installed into a text file
conda env export > my-parallel-env.yml
modify the prefix value in this text file to point to your own location,
and then create your own env:
conda env create -f my-parallel-env.yml
This repo contains the following items:
Exercises discussed in the workshop:
profiling.py: using timers and profilers (Exercise 1)test-multiprocessing.py: multiprocessing intro (Exercuse 2)map-reduce-pi.py: calculating pi with multiprocessing via map/reduce (Exercise 3)performance.py: strong scaling performance (Exercise 4)mpi-comm-ops.py: mpi4py communication between procs (Exercise 5)test-multithreading.py: multithreading examples (Exercise 6)
legacy/ : folder containing the examples from the past workshops
- mpi/ : scripts showing send/recv to solve the 1D heat equation with mpi4py
- notebooks/ : jupyter notebooks for calculating pi with multiprocessing
Extra stuffs:
test-numba.py,test-numba2.py: using numba JITmpi-calculate-pi.py`` : calculating pi usingmpi4py`
Slides:
- python-workshop.pdf