[kprotty-fd]

Folds the store's tags, shred_pool, and root_map into the same ring_pool. Combined with smaller ticket seq, use of fd_map_chain (over fd_map_chain_para) guarded under a rwlock (rserve only caller), and smaller bucket counts reduces store memory overhead.

[github-actions]

 ┌─ ⚡ PERF · e70f855 vs main@3a633f8 ─────────────────────────────────
 │ SUITE                               BASELINE          NEW         Δ
 │ replay tps, mainnet               12,927 tps   12,787 tps  ·  -1.09%
 │ bench tps, localnet              669,411 tps  670,314 tps  ·  +0.13%
 │ snapshot load, testnet                4.93 s       4.94 s  ·  +0.17%
+│ mem total, mainnet                150.35 GiB   148.35 GiB  ▼  -1.33%
+│ mem total, testnet                 86.66 GiB    85.66 GiB  ▼  -1.15%
 │ clean compile, firedancer             3.17 s       3.14 s  ·  -0.86%
 │ binary size, firedancer             57.92 MB     57.89 MB  ·  -0.04%
 ├─────────────────────────────────────────────────────────────────────
@@ 0 REGRESSIONS · 0 WARNINGS · 2 IMPROVED · 5 NOISE @@
 └─────────────────────────────────────────────────────────────────────

history · 4 pushes

 ┌─ HISTORY · Δ vs main, per push, newest first ──────────────────────────────────
 │ HEAD         TPS    BENCH     SNAP    MEM·M    MEM·T  COMPILE   BINARY
+│ e70f855   -1.09%   +0.13%   +0.17%   -1.33%   -1.15%   -0.86%   -0.04%
+│ 309bc85   +0.20%   +0.10%   -0.12%   -1.33%   -1.15%   +0.42%   -0.04%
+│ dd3e08d   +0.33%   -0.48%   -5.52%   -1.33%   -1.15%   -0.17%   -0.04%
+│ d8e3670   +0.03%   +0.23%  -19.98%    0.00%    0.00%   +0.66%    0.00%
 └────────────────────────────────────────────────────────────────────────────────

[Copilot]

🟡 Changes recommended

Generation tags repeat on consecutive cell reuse at supported large store sizes, undermining concurrent-read validation.

1 open issue

• Prevent same-cell generation collisions after 2^30 reservations · New

What changed in this PR

Reduces the disk store’s index memory overhead by consolidating metadata and replacing concurrent hash maps with lock-protected chain maps.

Changes:

• Combines tags and map entries into a shared ring pool with smaller generations and bucket arrays.
• Reworks insertion, eviction, and queries around a shared read/write lock.
• Updates footprint assertions and adds collision, aliasing, and slot-hint tests.

File	Description
src/​disco/​store/​test_store.c	Updates memory expectations and exercises ring and map edge cases.
src/​disco/​store/​fd_store.h	Defines the compact layout, map types, and updated API documentation.
src/​disco/​store/​fd_store.c	Implements shared-pool allocation and lock-protected disk operations.

_{Review effort: Balanced}

[Copilot]

🔵 Needs a closer look

The shared-memory layout and synchronization rewrite needs human review and targeted test execution to validate concurrent I/O and eviction behavior.

No open issues 🎉

1 resolved since last review

• Prevent same-cell generation collisions after 2^30 reservations

_{Review effort: Balanced}

[Copilot]

🔵 Needs a closer look

The shared-memory layout and synchronization redesign needs human review and runtime validation of concurrent reads, writes, and eviction.

No open issues 🎉

_{Review effort: Balanced}