feat(builder): emit validity-predicate state-load metrics

[BrianBland]

Track the accounts and storage slots read while evaluating validity predicates during a block build, emitted once per block as histograms of total reads (every read, including re-reads across flashblocks and parked-transaction rescans) and distinct locations (the predicate state footprint).

A PredicateReadRecorder wraps the builder State for the duration of each predicate evaluation and records reads into a per-block PredicateLoadTracker carried on ExecutionInfo, respecting short-circuit evaluation. This is the footprint signal a predicate-state prewarmer must cover; disk-tier cost is covered separately by reth's sync.caching miss metrics and by predicate evaluation time.

[linear]

BASE-246

[cb-heimdall]

🟡 Heimdall Review Status

Requirement	Status	More Info
Reviews	🟡 0/1	Denominator calculation Show calculation 1 if user is bot 0 1 if user is external 0 2 if repo is sensitive 0 From .codeflow.yml 1 Additional review requirements Show calculation Max 0 0 From CODEOWNERS 0 Global minimum 0 Max 1 1 1 if commit is unverified 0 Sum 1

[github-actions]

Review Summary

PR: feat(builder): emit validity-predicate state-load metrics

This PR adds per-block observability for validity-predicate state reads by introducing a PredicateReadRecorder (a Database wrapper) and a PredicateLoadTracker accumulator. The recorder wraps the builder's State during predicate evaluation, counts account and storage-slot reads (total and distinct), and emits four histogram metrics at block finalization.

Block Production Safety

This PR touches the flashblock build loop (context.rs) — a block-production-sensitive path. I reviewed it against the block production review guide and found no critical issues:

• No new error paths: PredicateReadRecorder delegates all Database methods transparently with the same error type (<State<DB> as Database>::Error). Errors propagate identically to the current code.
• No panics: Counter increments use += 1 on u64 (unreachable overflow), HashSet::len() as u64 is safe on 64-bit.
• No I/O boundary changes: Metrics are recorded in-process via the metrics crate — no new data transport.
• No concurrency issues: The recorder is created and consumed synchronously within accumulate_elapsed.
• Bounded collections: unique_accounts and unique_slots HashSet growth is bounded by gas limits and per-tx predicate cap (default 64).

Findings

No issues found. The implementation is clean:

• The Database wrapper pattern correctly delegates all four trait methods, recording only the two relevant to predicates (basic, storage).
• Short-circuit semantics are naturally preserved since reads flow through the wrapper as issued.
• The has_activity() gate prevents zero-observation dilution on ordinary blocks.
• Tests cover counting, short-circuit behavior, and the no-activity case for both the tracker and the metrics emission.

[github-actions]

Base Std historical fork tests

Fork	Result	Passed	Failed	Skipped	base/base	base-anvil	base-std
Beryl	pass	616	0	13	21b8baa1	8d0f5b8a	4658f1b7
Cobalt	pass	721	0	14	21b8baa1	9df661bc	e30b3421

View run

[github-actions]

Tip

Nice, this PR improves performance. 1 benchmark(s) faster by more than 10% beyond the noise band: frame_parse/parse_frames_8x128KiB (-30.0%).

Benchmark results (advisory)

Median time on the PR head versus the base branch, measured on the same host. Wall-clock, so a change is only flagged when it clears ±10% and the confidence intervals do not overlap. Only benchmarks past the ±10% threshold (plus new or dropped ones) are listed. This check never blocks a merge.

Benchmark	Base	Head	Δ median
frame_parse/parse_frames_8x128KiB	34.31 µs	24.03 µs	-30.0% ✅ faster

48 benchmark(s) within ±10% omitted.

View run · Re-run benchmarks