clpz: Reducing the number of inferences #3193
Description
In the benchmarks generously provided by @Yordando in #3150, I found the following comparison regarding the number of inferences particularly interesting as an example of something that we can possibly address in the Prolog implementation of clpz (as opposed to opportunities for improvements in the Rust engine, such as the changes in #3185):
We see a similar phenomenon for N-queens, also mentioned in the thread:
It may be possible to reduce the number of inferences that clpz needs to solve these tasks.
I suggest we start this work by trying to find smaller self-contained programs and queries that show a similar difference between different Prolog systems, i.e., examples where clpz currently needs ca. 3 times more inferences than any comparable Prolog system with a Prolog-based CLP solver.
Once we have such examples, we can think about how we can reduce the number of inferences.
Is anyone interested in trying to find such examples? I would greatly appreciate any help, and I hope that for example you @notoria, @flexoron and @haijinSk may be interested in this? Thank you a lot!