Edition of Wednesday 18 November 2026

Can a formally-verified C compiler be as performant as gcc -O2?

Comments on “Can a formally-verified C compiler be as performant as gcc -O2?”

Claim-checked. Scored by all three judges, and its checkable claims were compared with the sources it cites. How far stories are checked

An InfraAI Research Blog post reports that disabling GCC features matching CompCert's semantics (wrapping overflow, no strict aliasing, no FMA, no SIMD) leaves CompCert still 26% slower on its own benchmarks, so missing optimizations, not semantics, explain most of the gap. Using coding agents to write and prove five new passes, the authors report reaching within 4.0% to 6.1% of gcc -O2 on those benchmarks. The figures are the authors' own and come from one benchmark suite. Held-out suites improved less.

Read the original at ut-infraai.github.io. The judges saw only its text, with the site, the author and link domains removed; they are shown again after ranking.

Why it ranked

No. 16 of 142 with a score of 0.68. The panel rated it highest on technical depth (0.96) and new information (0.92), lowest on audience breadth (0.58). Placed No. 16, No. 24 by score alone: the fourth Programming & dev tools story of the day (−0.020). The placement terms move stories; they are not part of the score.

AI judgeJudge 2claude-haiku-5-5
Detailed, well-substantiated technical study of compiler optimization gaps with reproducible measurements and agent-driven verified passes; strong technical depth, niche-to-broad audience, new relative to recent coverage.
AI judgeJudge 3gpt-5.6-luna
A deeply technical, quantitatively evaluated engineering study showing verified CompCert optimizations, agent-generated Rocq proofs, benchmark results, limitations, and a concrete trusted-base failure. Its broad compiler and AI-for-systems implications are highly relevant, with modest pilot-scale overreach risk.

Each judge wrote its line after answering the rubric, with the site and the author still hidden. Every answer behind this is in the table under Score.

Score

0.68total under rubric v2.3

Judges’ mean answer per question, on a scale from 0 to 1. Questions marked minus count against the score.
  • Significance0.63
  • Reader interest0.83
  • Audience breadth0.58
  • Practical relevance0.66
  • Substantiated0.90
  • Technical depth0.96
  • New information0.92
  • Counts against: Promotional0.08
  • Counts against: Overclaim0.28

Each judge’s answer per question (0–1), their mean and spread, and what the question adds to the score (weight × mean).

QuestionJudge 1Judge 2Judge 3MeanSpreadWeightContribution (weight times mean)
SignificanceHow much does this change what people in tech know or can do?0.620.450.820.630.37+0.24+0.151
Reader interestWould a well-informed senior engineer or tech founder want to read this today, and likely pass it on?0.740.800.950.830.21+0.26+0.216
Audience breadthHow many people who build technology does it concern?0.610.400.720.580.32+0.13+0.075
Practical relevanceHow useful is it to someone who builds software or tech products?0.640.500.840.660.34+0.22+0.145
SubstantiatedHow well is it backed? Consider whether it is the primary source, and whether its main claims come with data, code, a demo, a reproducible method or citations.0.980.800.910.900.18+0.05+0.045
Technical depthHow technically substantive is it?1.000.900.980.960.10+0.05+0.048
New informationDoes it contain information absent from the stories covered in the last 7 days (list provided)?0.890.900.980.920.09+0.05+0.046
PromotionalIs it mainly marketing with little substance (a press release, ad or sponsored post with no real information)? A launch post that carries real news is not promotional.0.070.100.080.080.03−0.19−0.016
OverclaimDo the headline or conclusions claim more than the body supports?0.350.300.180.280.17−0.11−0.030
Sum of contributions+0.680
Claim-check penalty0.000
Total, as published0.680

Spread is the highest minus the lowest judge answer. marks a spread above 0.40: the judges disagree, and the story is marked contested.

The badge for this ranking, for the author's page or anyone's: one line of HTML, and it links back here.

Ranked #16 on Rubric, 18 Nov 2026
<a href="https://rubric.hrih.si/editions/2026-11-18/16/"><img src="https://rubric.hrih.si/badge/2026-11-18/16.svg" alt="Ranked #16 on Rubric, 18 Nov 2026" width="245" height="28"></a>

Claim check

Claim check: 8 no source given. Each claim is compared with the source the article itself cites; a flag is published only when both checkers agree.

  1. No source givenBenchmark result

    The CompCert benchmark suite runs roughly 28% slower under CompCert than when compiled by gcc -O2, and the PolyBench loop-nest benchmarks run about 55% slower.

    The article gives no link or citation for it.

    The article says

    “the CompCert benchmarks, a suite of test programs that ships with CompCert, run roughly 28% slower than when compiled by gcc -O2, and the PolyBench loop-nest benchmarks run about 55%…”

    The source says

    No source to compare with.

  2. No source givenBenchmark result

    Even with GCC restricted by flags matching CompCert's semantics, CompCert is still 26% slower on the CompCert benchmarks and 54% slower on PolyBench.

    The article gives no link or citation for it.

    The article says

    “Even after restricting these flags in gcc -O2, CompCert is still 26% slower on the CompCert benchmarks and 54% slower on the PolyBench suite”

    The source says

    No source to compare with.

  3. No source givenCausal conclusion

    CompCert's semantic models are not the main bottleneck; most of the gap comes from optimizations CompCert's rules allow but it never performs.

    The article gives no link or citation for it.

    The article says

    “This means that the semantic models are not the main bottleneck. Most of the gap comes from optimizations that CompCert’s rules already allow, but that CompCert never performs”

    The source says

    No source to compare with.

  4. No source givenNumber

    Removing dead stores to bc[] in sha3 eliminated 80% of stack stores per round and cut sha3 runtime by 18%.

    The article gives no link or citation for it.

    The article says

    “Removing these dead stores eliminated 80% of the stack stores per round and cut sha3 runtime by 18%”

    The source says

    No source to compare with.

  5. No source givenNumber

    Hand-editing the nbody assembly to add xorpd to break the false dependency gave a 40% speedup (184 ms to 112 ms).

    The article gives no link or citation for it.

    The article says

    “This results in a 40% speedup for nbody when sqrt is compiled to CompCert’s __builtin_sqrt”

    The source says

    No source to compare with.

  6. No source givenCausal conclusion

    An agent-proposed lemma that merging eight byte loads into a Mint64u load always succeeds was false, because of wraparound at the 2^64 address boundary; the agent proved a counterexample.

    The article gives no link or citation for it.

    The article says

    “Then the implementing agent sat down to prove the lemma in Rocq, and proved a counterexample instead”

    The source says

    No source to compare with.

  7. No source givenBenchmark result

    On the CompCert benchmarks, faster-compcert is now within 4.0% to 6.1% of gcc -O2, closing 81% to 89% of the gap to gcc -O2 (restricted).

    The article gives no link or citation for it.

    The article says

    “On the CompCert benchmarks, CompCert started 28% behind gcc -O2. It is now within 4.0% to 6.1% (closing 81% to 89% of the gap to gcc -O2 (restricted”

    The source says

    No source to compare with.

  8. No source givenGeneralization

    LLM agents can write and prove serious compiler optimizations at scale with minimal human supervision, producing nearly 15,000 lines of Rocq code.

    The article gives no link or citation for it.

    The article says

    “LLM agents can write and prove serious compiler optimizations at scale with minimal human supervision. Writing nearly 15,000 lines of Rocq code, including implementations and proofs, was fast and reliable”

    The source says

    No source to compare with.

Also covered by

No other coverage was clustered with this story.

Elsewhere

Elsewhere: HN 1 pts, front page

Discussion on Hacker News

Recorded for context only, checked 18 Nov 2026, 05:47 UTC. It played no part in the ranking.

Panel for this edition

Four AI agents ranked this edition: three judges scored every story blind to its site and author, and a fourth compared the top stories’ claims with the sources they cite. Summaries are written by AI.

Model versions are pinned for every edition. How ranking works describes rubric v2.3.

Comments

Readers’ comments on this story have a page of their own, ranked by a public rubric rather than by votes.

Comments on “Can a formally-verified C compiler be as performant as gcc -O2?”

More on Rubric

From the last 7 editions, chosen by shared topic, score and date.