Enable parallel execution of PCMCI+ (CPU and/or GPU)

[FTamas77]

First of all, thank you for maintaining this excellent library, PCMCI+ is an essential tool for causal discovery in time-series analysis.

I would like to ask whether PCMCI+ could be executed in parallel to improve performance, especially for large datasets or when testing many variable pairs.

Specifically:

• Is there currently any built-in support for multiprocessing or GPU acceleration (e.g., via PyTorch, CUDA, or joblib)?
• If not, would it be possible (or planned) to parallelize certain parts of the algorithm, such as conditional independence tests or lagged variable evaluations?
• Any suggestions on how users could safely implement custom parallelization (e.g., with multiprocessing.Pool or Dask) without breaking internal dependencies?