1. Via FREngine API
Process the document using the FRDocument object with MultiProcessingMode = MPM_Sequential or MultiProcessingMode = MPM_Auto (you can find out more about parallel processing in "Help" → "Guided Tour" → "Advanced Techniques" → "Parallel Processing").
2. Other ways
2.1. Process (analyze, recognize, synthesize, export) each document in a separate process.
The number of processes: the same as the number of processor cores. [The number of processes cannot exceed the number of processor cores.]
Engine loading: you can use all of the engine loading methods described in "Help" → "Guided Tour" → "Advanced Techniques" → "Programming Aspects" → "Different Ways" to Load Engine.
2.2. Process each document in a single thread.
Engine loading: every thread should create its own OutprocLoader for working with FREngine.
Using several InprocLoaders in a single process and using one OutprocLoader in different threads do not lead to parallel processing.
It is better to use FRDocument object with MultiProcessingMode = MPM_Sequential for the methods described in 2.1 and 2.2.
You can find other ways to increase processing speed in "Help" → "Guided Tour" → "Best Practices" → "Increasing Processing Speed".