day11: perf

src/day11.rs

@@ -86,6 +86,12 @@ fn part1_memo(fwd: &Edges) -> u64 {
 
 #[aoc(day11, part2, MemoDfs)]
 fn part2_memo(edges: &Edges) -> u64 {
+    // we assume all valid paths reach svr->fft->dac->out in that order. this holds on the example and the input.
+    // to handle both possible orderings, we need to sum paths fft->dac and dac->fft, and then partition the graph
+    // (just remove the node) at dac & fft before computing the sum of dac->out and fft->out.
+    //
+    // this is our performance implementation though and it's faster to not do that work. we do the posterity for the
+    // naive solution
     count_paths("svr", "fft", edges)
         * count_paths("fft", "dac", edges)
         * count_paths("dac", "out", edges)