Merge HalfKA and Threats accumulators

[anematode]

Passed STC (https://tests.stockfishchess.org/tests/view/6a2893ad7c758d82accea129):
LLR: 3.20 (-2.94,2.94) <0.00,2.00>
Total: 23328 W: 6145 L: 5838 D: 11345
Ptnml(0-2): 50, 2463, 6346, 2740, 65

Instead of repeatedly doing the sum HalfKA + threats at the end, it's profitable to simply store one accumulator per side that combines them. This also avoids an extra/load store of an accumulator, and halves the cache footprint of the accumulators.

For full refreshes, we always compute both halfka and threats simultaneously. Any threat full refresh is always a halfka refresh because it occurs when the king crosses the center line, while halfka refreshes are required for ANY king move, so we don't need a separate detection path for threats.

I get about a 2.5% speedup locally with this, but I'd appreciate other ppl's measurements.

No functional change

[github-actions]

(execution 27248272859 / attempt 1)

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: 644a7127-5ba3-47d4-902d-084bd347c993

📥 Commits

Reviewing files that changed from the base of the PR and between 6867a8f and b50e231.

📒 Files selected for processing (1)

• src/nnue/nnue_accumulator.h

💤 Files with no reviewable changes (1)

• src/nnue/nnue_accumulator.h

---

📝 Walkthrough

Walkthrough

This PR unifies NNUE accumulator handling by replacing templated PSQ/threat states with a single non-templated AccumulatorState containing DirtyPiece and DirtyThreats. AccumulatorStack APIs (latest/mut_latest/evaluate_side/find_last_usable_accumulator/forward/backward incremental updates) and storage are converted to non-templated, Color-keyed forms. A new apply_combined implements SIMD/scalar combined PSQ+threat delta application. update_accumulator_refresh_cache now refreshes PSQ and threat contributions together and the separate threat-only refresh function was removed. FeatureTransformer::transform() reads psqtAccumulation from the unified latest accumulator and omits threatAccumulation additions.

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[github-actions]

(execution 27251005329 / attempt 1)

[github-actions]

(execution 27251205882 / attempt 1)

[github-actions]

(execution 27253689613 / attempt 1)

[vondele]

==== master ====
==== Bench: 3493826 ====
1 Nodes/second : 294691755
2 Nodes/second : 294351857
3 Nodes/second : 296083491
Average (over 3):  295042367
==== merge-accumulators ====
==== Bench: 3493826 ====
1 Nodes/second : 308522687
2 Nodes/second : 307923326
3 Nodes/second : 308319161
Average (over 3):  308255058

[mstembera]

I wonder if this makes a larger L1 like say 1280 more promising?