use bitmask AND for cross set computation, skip dead anchors#18
Merged
Conversation
Add LetterBits: a precomputed bitmask of child tiles for each KWG sibling group position. Computed once per KWG (lazily on first use, invalidated on reset_for_another_kwg). Replace collect_bits iteration in gen_extension_sets with a single letter_bits table lookup via extension_bits_from_letter_bits. Eliminates all KWG child iteration from extension set computation. Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
Extend LetterBits to also precompute accepting_bits: bitmask of child tiles that complete a word (node.accepts() is true). This enables O(1) cross set child enumeration in a follow-up commit. Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
Replace iterative KWG child loops in the "right, no left" and "left, no right" cross set cases with O(1) accepting_bits table lookups. The "left and right" merge-intersect case still iterates (different algorithm, not a simple lookup). Thread letter_bits table through gen_cross_set for access during cross set computation. Lazily initialized before cross set loops. Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
Replace the sorted merge-intersect of two KWG sibling groups with an O(1) bitwise AND of their letter_bits. Only candidate tiles present in both groups need the full KWG word verification traversal. Eliminates the dual-pointer merge loop. Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
The extension set anchor skip (both left and right extension sets are zero) is not equity-based pruning — it means no word can be formed at this anchor. Apply it before the want_raw check so dead anchors are skipped in all modes. Also use && short-circuit instead of | to avoid loading the second extension set when the first is already zero. Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
Replace precomputed letter_bits/accepting_bits table lookups with on-the-fly sibling group iteration. The table was 24MB (2x 12MB for ~1.5M KWG nodes) which hurts cache and is too large for WASM targets. The on-the-fly computation is short sequential loops over contiguous siblings — cache-friendly. Keep LetterBits struct and compute_letter_bits() in kwg.rs for callers that want precomputed tables. Remove stale letter_bits comments and clone comment for accepts_alpha_cache (source of truth is the struct field). Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Test plan