Skip to content

refactor(core): tidy the review leftovers of #407 and #408 - #409

Merged
danielplohmann merged 1 commit into
masterfrom
refactor/review-nits-407-408
Oct 20, 2026
Merged

danielplohmann merged 1 commit into
masterfrom
refactor/review-nits-407-408

Conversation

@danielplohmann

Copy link
Copy Markdown
Owner

Follow-up to the reviews of #407 and #408. No behavior change.

  • AArch64Disassembler: drop the _getDisasmWindowBuffer override. It copied ARM's line for line, and decode_window_size = DECODE_WINDOW already selects the 32-byte block window. The only difference was the default for calls without a size, and no AArch64 code makes one (only the three x86 callers in X86Backend.py do).
  • Tarjan: remove _nonrecursive_stack, which nothing has used since the iterative rewrite.
  • RecursiveDisassembler.analyzeFunction: compute the decode window once per function, not once per block, and restore the comment on why decoding reads ahead.
  • FunctionCandidateManager.getNextGap: comment the in-image address-mask checkpoint added in fix(common): end the gap scan at the image end, not at the address mask #407.

no-changelog: an internal cleanup with no behavior change, so a report consumer has nothing to read in an entry.

  • make lint and ruff format --check .
  • ty check: same diagnostic count as master (60 errors and 328 warnings locally, none new)
  • Fast suite: 2431 passed, 2 skipped

🤖 Generated with Claude Code

- AArch64Disassembler: drop the _getDisasmWindowBuffer override. It duplicated
  ARM's, and decode_window_size already selects the 32-byte block window; the
  only difference was the default for size-less calls, and AArch64 makes none.
- Tarjan: remove _nonrecursive_stack, unused since the iterative rewrite.
- RecursiveDisassembler.analyzeFunction: compute the decode window once per
  function instead of once per block, and restore the note on why there is a
  lookahead at all.
- FunctionCandidateManager.getNextGap: comment the in-image address mask
  checkpoint added in #407.

No behavior change.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
@danielplohmann danielplohmann added the no-changelog PR touches src/smda but needs no CHANGELOG entry; skips the changelog check label Oct 20, 2026
@danielplohmann danielplohmann added this to the v5.0.1 milestone Oct 20, 2026
@github-actions

Copy link
Copy Markdown

📊 SMDA Performance Evaluation Benchmark Results

Generated on: 2026-10-20 21:13:03

Summary

Metric Result
Correctness ✅ PASS — 0 / 155 common file(s) differ
Determinism ✅ PASS — base 3 run(s), PR 3 run(s)
Verdict inconclusive — within run-to-run noise (±2.4%)
Median paired speedup +0.22% (95% CI [-0.17%, +0.60%])
Timing noise band ±2.4%

Per-side Timing Context (best of 3/3 runs per file)

Side Files Functions Median best time/file (s) Sum of best times (s) Throughput estimate (func/s)
base 155 173540 0.5975 218.66 ~794
pr 155 173540 0.6143 218.04 ~796

These context rows sum each file's best observed time across repeated runs; they are normalized comparison estimates, not single-run CI wall-clock times.

Paired per-file timing (positive speedup = PR faster):

Statistic Value
Files compared 155
Median paired speedup +0.22% (95% CI [-0.17%, +0.60%])
Mean speedup -0.20% (95% CI [-1.10%, +0.69%])
Std dev / IQR 5.74% / 4.35%
Wilcoxon signed-rank p 0.3618 (n=155)

ℹ️ base and PR are timed in the same job on the same runner, with their passes interleaved base/PR/PR/base/base/PR so that drift over the job cancels rather than landing on whichever side ran last. The noise band above is therefore run-to-run variation on one machine, which is what it is able to measure; a difference inside it is reported as inconclusive. Correctness and determinism are not timing-based and are unaffected either way.

Determinism (self-check across repeated runs)

Side Runs Files Deterministic Median timing CV
base 3 155 ✅ 2.4%
pr 3 155 ✅ 1.9%
Pairwise run matrix (individual run medians, diagnostic)

Diagnostic only: each row compares one raw PR run against one raw base run. The headline verdict above uses paired per-file best-of-runs timings.

Comparison Run PR Files Base Files Common Function Set Matches Med PR (s) Med Base (s) Med Diff (s) Speedup %
pr_0 vs base_0 155 155 155 155/155 0.6661s 0.6349s +0.0312s -4.92%
pr_0 vs base_1 155 155 155 155/155 0.6661s 0.6477s +0.0185s -2.85%
pr_0 vs base_2 155 155 155 155/155 0.6661s 0.6402s +0.0259s -4.05%
pr_1 vs base_0 155 155 155 155/155 0.6246s 0.6349s -0.0103s +1.62%
pr_1 vs base_1 155 155 155 155/155 0.6246s 0.6477s -0.0231s +3.56%
pr_1 vs base_2 155 155 155 155/155 0.6246s 0.6402s -0.0156s +2.44%
pr_2 vs base_0 155 155 155 155/155 0.6324s 0.6349s -0.0025s +0.39%
pr_2 vs base_1 155 155 155 155/155 0.6324s 0.6477s -0.0153s +2.35%
pr_2 vs base_2 155 155 155 155/155 0.6324s 0.6402s -0.0078s +1.22%

@danielplohmann
danielplohmann merged commit 14d929d into master Oct 20, 2026
26 of 27 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

no-changelog PR touches src/smda but needs no CHANGELOG entry; skips the changelog check

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant