docs: update incremental benchmarks (3.9.1)

[github-actions]

Automated incremental benchmark update for 3.9.1 from workflow run #667.

[greptile-apps]

Greptile Summary

This PR adds automated incremental benchmark results for codegraph v3.9.1 (570 files, measured 2026-04-06), generated by workflow run #667.

• Native full-build improved 20% (6.5s → 5.2s) while WASM held roughly flat (6.8s → 6.9s)
• The native incremental-rebuild regression (1-file: 527ms → 757ms, no-op: 6ms → 15ms) is already tracked in issue perf: investigate native engine incremental rebuild regression in v3.9.1 #882 — no new action needed here
• All percentage-change annotations in the summary table are mathematically consistent with the underlying raw JSON values (e.g. nativeBatchMs 5.1ms → 7ms = +37.25% ≈ ↑37%; jsFallbackMs 10.8ms → 12.2ms = +12.96% ≈ ↑13%)
• Import-resolution speedup ratio correctly updated from 2.1x to 1.7x to reflect the new 7ms / 12.2ms timing ratio

Confidence Score: 5/5

Safe to merge — documentation-only update with correct, internally consistent benchmark data.

No code is changed. All numeric values and percentage deltas have been verified against the raw embedded JSON. The only notable regression (native incremental rebuild) was already flagged and tracked in issue #882 before this review. No new issues found.

No files require special attention.

Important Files Changed

Filename	Overview
generated/benchmarks/INCREMENTAL-BENCHMARKS.md	Automated 3.9.1 benchmark data appended; summary table percentage deltas and embedded raw JSON are internally consistent

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[CI Workflow Run #667] --> B[Run codegraph on its own codebase]
    B --> C{Engine}
    C -->|Native Rust| D[Full: 5.2s\nNo-op: 15ms\n1-file: 757ms]
    C -->|WASM| E[Full: 6.9s\nNo-op: 14ms\n1-file: 603ms]
    D --> F[Import resolution\nNative: 7ms / JS: 12.2ms\nSpeedup: 1.7x]
    E --> F
    F --> G[Generate INCREMENTAL-BENCHMARKS.md]
    G --> H[PR 879: docs: update incremental benchmarks 3.9.1]

Loading

_{Reviews (2): Last reviewed commit: "docs: update incremental benchmarks (3.9..." | Re-trigger Greptile}

[greptile-apps]

Native incremental builds slower than WASM

The 3.9.1 data shows the native engine's 1-file rebuild (757ms) is now 25% slower than WASM (603ms). In prior releases the two engines were roughly at parity for incremental builds (3.9.0: native 527ms vs WASM 541ms). Combined with the no-op rebuild regression (6ms → 15ms, +150%), this suggests the native engine may have a performance issue introduced in 3.9.1 specifically in the incremental code path.

This PR correctly records the measured data, but it may be worth opening a follow-up issue to investigate the root cause before the next release.

[carlos-alm]

Good catch. Opened #882 to track investigating the native incremental rebuild regression before the next release. Since this PR only records the measured data (which is correct), merging as-is and addressing the root cause separately.

[carlos-alm]

@greptileai