Concurrency

py-bitcoinkernel is thread-safe, but should not be used with multiprocessing.

Multithreading

Python generally is not well suited for multithreading, because of the Global Interpreter Lock (GIL). However, because the bitcoinkernel shared library does not contain any Python bytecode, we can safely bypass a lot of those limitations.

For that reason, it is recommended to use multiple threads for expensive operations such as reading blocks from disk. See examples/threading.md for an example implementation.

Thread-safety

py-bitcoinkernel is safe to use from multiple threads. However, synchronization (using e.g. threading.Lock()) is still required in some circumstances, including: - changes to the chainstate (such as advancing the tip, or rolling back) may invalidate earlier obtained objects. For example, the result from Chain.block_tree_entries[-1] is not guaranteed to remain the tip if another thread can advance the chainstate.

Parallelism / multiprocessing

py-bitcoinkernel currently does not support multiprocessing, because most bitcoinkernel functionality can currently only be used from a single process, due to its dependency on and usage of LevelDB.