Native Preemptive Threads in SWI-Prolog

Concurrency is an attractive property of a language to exploit multi-CPU hardware or perform multiple tasks concurrently. In recent years we see Prolog systems experimenting with multiple threads only sharing the database. Such systems are relatively easy to build and remain very close to standard Prolog while providing valuable extra functionality. This article describes the introduction of multiple threads in SWI-Prolog exploiting OS-native threading. We discuss the extra primitives available to the Prolog programmer as well as implementation issues. We explored speedup on multi-processor hardware and speed degradation when executing a single task.