Reference for Learners¶

Course Materials¶

All course materials are organized in the content/ directory:

Episodes: Jupyter notebooks with hands-on examples
Code Examples: Supporting Python scripts in episodes/code/
Images: Diagrams and visualizations in episodes/images/
Solutions: Example solutions in solution/
Datasets: Sample data in dataset/

HPC Resources¶

Karolina Supercomputer¶

The Karolina supercomputer is available for hands-on sessions:

Location: IT4Innovations, Czech Republic
Performance: 11.6 PFlop/s (standard), 360 PFlop/s (GPU-accelerated)
Access: Through job submission via SLURM

Installation & Setup¶

Python Environment:

conda create -n python-hpc python=3.11
conda activate python-hpc
pip install -r requirements.txt

Container-based Setup (using Apptainer/Singularity):
- See content/resources/*.def files for container definitions
- Build: apptainer build image.sif image.def
- Run: apptainer exec image.sif python script.py

Module Loading (on HPC clusters):

module load Python/3.11
module load CUDA  # for GPU computing

Key Topics Covered¶

Performance Optimization¶

Profiling: Identifying bottlenecks using cProfile, line_profiler
Benchmarking: Measuring execution time and resource usage
NumPy: Vectorization and array operations for speed

Parallel Programming¶

Dask: Task-based parallelization for distributed computing
Numba: JIT compilation for CPU optimization
Cython: C-extensions for performance-critical code
MPI: Message Passing Interface for distributed memory systems

HPC Job Management¶

SLURM: Job scheduling on supercomputers
Batch Scripts: Writing and submitting job submissions
Job Monitoring: Tracking resource usage and job status

Code Examples¶

Running Benchmarking Scripts¶

python content/episodes/code/dist.py        # Performance distribution example
python content/episodes/code/example.py     # General example

MPI Examples¶

# Run with OpenMPI
mpirun -np 4 python content/episodes/code/mpi-test.py
./content/episodes/code/hello_mpi.sh        # Simple MPI test

Numba JIT Compilation¶

from numba import jit
import numpy as np

@jit(nopython=True)
def fast_function(x):
    return x ** 2

result = fast_function(np.array([1, 2, 3]))

Dask Distributed Computing¶

import dask.array as da

# Create a Dask array
x = da.random.random((1000, 1000), chunks=(100, 100))
result = x + x.T
computed = result.compute()

Common Issues & Solutions¶

Import Errors¶

Ensure all packages are installed: pip install -r requirements.txt
Activate the correct virtual environment
Check that module paths are correctly set

MPI Not Found¶

Load the MPI module: module load OpenMPI
Install mpi4py: pip install mpi4py

Performance Not Improving¶

Profile your code first using cProfile or line_profiler
Use appropriate optimization techniques for identified bottlenecks
Consider algorithmic improvements alongside code optimization

Container Issues¶

Ensure Apptainer/Singularity is installed: apptainer --version
Check container definition files in content/resources/
Use --nv flag for GPU-enabled containers: apptainer run --nv image.sif

External Resources¶

Documentation¶

Performance & Optimization¶

HPC & Parallel Programming¶

Training Materials¶

Getting Help¶

Consult the instructor-guide.md for teaching notes
Review solution notebooks in solution/ folder
Check error messages carefully - they often indicate the exact problem
Search Python documentation and community forums for specific issues

Course Feedback¶

Your feedback helps improve this course. Please share:

What topics were most helpful?
What could be improved?
What additional topics would you like to see?

Last Updated: May 2026 License: CC BY-SA 4.0