Comparison of Prolog implementations

The following Comparison of Prolog implementations provides a reference for the relative feature sets and performance of different implementations of the Prolog computer programming language. A comprehensive discussion of the most significant Prolog systems is presented in an article published in the 50-years of Prolog anniversary issue of the journal Theory and Practice of Logic Programming (TPLP).^[1]

There are Prolog implementations that are radically different, with different syntax and different semantics (e.g. Visual Prolog)^[2] and sub-communities have developed around different implementations.^[2]

Code that strictly conforms to the ISO-Prolog core language is portable across ISO-compliant implementations. However, the ISO standard for modules is an extension which was not fully adopted in most Prolog systems.^[2][1]

Factors that can adversely affect portability include: use of bounded vs. unbounded integer arithmetic, additional types such as string objects, advanced numeric types (rationals, complex), feature extensions such as Unicode, threads, and tabling.^[3] Use of libraries unavailable in other implementations and library organisation:^[2]

Currently, the way predicates are spread over the libraries and system built-ins differs enormously. [...] Fortunately, there are only few cases where we find predicates with the same name but different semantics (e.g. delete/3)

• Benchmarking issues: Odd Prolog benchmarking, Performance differences.^[14]
• Benchmarking software: older, Dobry's Benchmarks, Aquarius benchmark suite, (Bothe, 1990),^[15] (Demoen et al. 2001), benchmark descriptions
• Benchmarking results: B-Prolog, SICStus, XSB,^[16] SICStus vs Yap vs hProlog^[17]
• Benchmarking results: Survey of java prolog engines by Michael Zeising
• Benchmarking results: OpenRuleBench yearly open-source benchmark of rule engines

• Overview of Prolog Systems by Ulrich Neumerkel
• Conformity assessment I: Syntax