| commit | e1bf95210835e764c68b84b37e802689df15959f | |
| author | Eric Blake <eblake@redhat.com> | |
| Tue, 19 Dec 2023 19:18:59 +0000 (19 13:18 -0600) | ||
| committer | Eric Blake <eblake@redhat.com> | |
| Tue, 19 Dec 2023 19:28:02 +0000 (19 13:28 -0600) | ||
| tree | 9a7d06352c2f464a796a10fef41b3679478f3543 | treesnapshot (tar.gz zip) |
| parent | 8096efdcf763a309ff8653e8a5ff31f2ec86686f | commitdiff |
day 16 part 2 solved
D'oh, I typo'd the rays entering to the left. Now takes ~24s;
probably some fertile room for optimization.
For an NxN grid, each of the 4N scans visits up to 4*N*N nodes. Lots
of paths are being rescanned, for O(n^3) work. But memoizing how many
nodes were scanned after a split will only be useful if I can come up
with a convenient way to not over-count intersections with the portion
of the incoming ray before the splitter.
D'oh, I typo'd the rays entering to the left. Now takes ~24s;
probably some fertile room for optimization.
For an NxN grid, each of the 4N scans visits up to 4*N*N nodes. Lots
of paths are being rescanned, for O(n^3) work. But memoizing how many
nodes were scanned after a split will only be useful if I can come up
with a convenient way to not over-count intersections with the portion
of the incoming ray before the splitter.
| 2023/day16.m4 | diffblobblamehistory |