Benchmark Results

The next table gives an overview of three benchmarks:

1. the first benchmark is the well known naive reverse benchmark
2. the second benchmark (dnf) is the complete program from section 11 and appendix B
3. the third test is the NET DATALOG benchmark; NET is an automatically generated (from a constraint net) tool-selection program for an NC-program generator [BHH+91]; its task is to select a cutting tool for turning a given workpiece on a CNC-lathe machine

Since the -WAM was conceived as a didactic prototype written in higher-level LISP, not as a PROLOG product, the absolute values are not yet competitive with well known production PROLOGs. The average speed-up gained by indexing in our database-like applications, however, is a factor between 20 and 30. But even rather deterministic procedures like append and reverse produce a speed-up of at least a factor of 2.