This is an example of how to perform profiling on the Arc Unwrapping algorithms. In this example we used the so-called "LAMBDA method" developed by TUDelft, and use a wrapper function in DePSI to call the unwrapping algorithm.

• LAMBDA method: LAMBDA.py (private repo)
• Wrapper funtion lambda_estimation: lambda_estimation (private repo)
• Data needed for the example: stm_amsterdam_173p.zarr

• lambda_unwrap.ipynb: Jupyter notebook for debugging and visualization
• lambda_unwrap.py: Python script for running the unwrapping algorithm in a recursive loop
• lambda_unwrap_dask.ipynb: Jupyter notebook for debugging dask method
• lambda_unwrap_dask.py: Python script for running the same unwrapping algorithm with dask databags

We use the py-spy package to profile the unwrapping algorithm. For example, we can use the following command to profile the unwrapping algorithm without dask

py-spy record --output profile_loop_60pnts.json --idle --rate 5 --subprocesses --format speedscope python lambda_unwrap.py

In lambda_unwrap_dask.py configure:

# Configure dask scheduler
dask.config.set(scheduler="processes") 

py-spy record --output profile_dask_60pnts_processes.json --idle --rate 5 --subprocesses --format speedscope python lambda_unwrap_dask.py

# Configure dask scheduler
dask.config.set(scheduler="threads") 

py-spy record --output profile_dask_60pnts_threads.json --idle --rate 5 --subprocesses --format speedscope python lambda_unwrap_dask.py

Then you can visualize the profiling results using the speedscope web tool by uploading the corresponding JSON file.

The results for 5 points, 24 points and 60 points can be found at: this link