Change ref_from_bytes() to inline(always).

[korran]

When building my firmware for rv32imc with lto=fat, changing codegen-units from 16 to 1 influences the compiler heuristics enough to not inline this function, causing code bloat and unnecessary panic logic.

[gemini-code-assist]

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

[joshlf]

You'll want to run ./tools/update-expected-test-output.sh to update expected output and fix the failing CI job.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 91.87%. Comparing base (e38763f) to head (5a4f1e1).

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #3137   +/-   ##
=======================================
  Coverage   91.87%   91.87%           
=======================================
  Files          20       20           
  Lines        6057     6057           
=======================================
  Hits         5565     5565           
  Misses        492      492           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[korran]

hmm, those benchmark changes don't look good :(

[jswrenn]

The benchmark results are fine. :-) There's only a six instructions overhead on the dynamic padding case, and a hypothetical cycle improvement in the dynamic size case. Since dynamically padded types are unusual, not too much weight should be put in optimizing them relative to unpadded dynamically sized types.