Completion Experiments
The following table collects results obtained with mkbtt running normalized completion with different versions of AC-KBO. A tarball containing all input systems can be found here.
Results in one row were obtained in parallel and run on a server equipped with eight dual-core AMD Opteron processors 885 running at a clock rate of 2.6GHz with 64GB of main memory, for each setting with a timeout of 200 seconds. Each termination call had a timeout of 1.5 seconds, and the primality critical pair criterion was used.
The columns correspond to the following settings. Column 1 applies AC-RPO directly, and column 0 applies AC-KBO directly, with subterm coefficients and possibly partial precedences. Columns 2 and 3 compare these plain orders to the settings where AC-dependency pairs are computed and AC-KBO and AC-RPO are used as reduction pairs, respectively, in both cases with argument filterings. Column 4 applies a strategy dependency pairs, a dependency graph approximation followed by a computation of strongly connected components. Afterwards reduction pairs based on AC-RPO, AC-KBO and matrix interpretations remove rules as long as possible, incorporating argument filterings and usable rules. AC-KBO always used 2 input and 3 output bits to represent weights and subterm coefficients.
Legend | |
0 | ./mkbtt -ct -st -cp prime -T 1.5 -t 200 -s acrpo |
1 | ./mkbtt -ct -st -cp prime -T 1.5 -t 200 -s 'ackbo -ib 2 -ob 3 -direct -sc -nt' |
2 | ./mkbtt -ct -st -cp prime -T 1.5 -t 200 -s 'dp; (acrpo -af)*' |
3 | ./mkbtt -ct -st -cp prime -T 1.5 -t 200 -s 'dp; (ackbo -af -ib 2 -ob 3 -sc -nt)*' |
4 | ./mkbtt -ct -st -cp prime -T 1.5 -t 200 -s 'dp; acdg?; sccs?; ({ur?; matrix -ib 2 -ob 3 -dp -dim 1 -ur[2]}restore || acrpo -af || ackbo -af -ib 2 -ob 3 -sc -nt || {ur?; matrix -ib 2 -ob 3 -dp -dim 2 -ur[2]}restore)*' |