An alternative approach to enumerating SSPiRAL Metalayouts, using the Nauty graph isomorphism solver.
tests can be run with cargo test and the benchmarks used to evaluate performance related code changes can be run with cargo bench.
I recommend using cargo build --release to build, the result will be ./target/release/nautust.
N can be set and DEBUG toggled at compile time in src/main.rs.
I believe the recursive enumeration is exhaustive. I would like to add a test using the other rust codebase to also enumerate metalatouts in order to demonstrate that.
I was hoping to be able to simply reproduce the Scores table for the full layout as the other simulators do, but the way that I am preventing redundant child metalayout visitation seemingly prevents me from calculating the number of collapsed layouts properly. In the regular algorithm the influence of a layout on a grandparent's reliability gets counted multiple times, but a predictable number that is corrected for with the normalization step. The pruning that is the core of this approach for performance reasons means the multiple counting is no longer predictable.