[lilith]

Summary

Eliminates avoidable per-invocation AST reparsing, redundant token-tree walks, and short-lived heap allocations across archmage-macros (PR 1 of 3 in the compile-time optimization stack):

• Fast-path attribute deduplication (drop_attrs_equal_to): Skip non-#[cfg] attributes (doc comments, #[inline], lint attributes) before formatting attribute tokens to String, and short-circuit immediately when the guard is empty or no #[cfg] attribute is present.
• Guarded Self/self token walks: Gate replace_self_in_tokens and replace_self_value_in_tokens behind tokens_contain_ident in #[arcane(_self = Type)] and #[autoversion(in_trait, _self = Type)], and skip unrelated delimited groups inside replace_self_value_in_tokens.
• Direct token-stream emission (no parse_quote! / syn::parse_str on hot paths): Emit #[inline], #[target_feature(enable = ...)], #[allow(...)], and ::archmage::* token paths directly into TokenStream (InlinePolicy::tokens, colon_path_tokens), inspect _self = Type via find_token_in_self_type without cloning syn::Signature + parse_quote!, and avoid cloning function bodies in attune proof wrappers.
• Stack-first identifier & small-buffer paths: Replace Ident::to_string() keyword/registry dispatches and small per-call Vec buffers with stack-backed with_ident_str ([u8; 64]), 'static token/trait string slices, borrowed TokenTypeInfo<'a>, and safe inline SmallVec<T, N>.

Measured Impact (default opt-level = 0, 20-thread Linux x86_64)

Per-invocation allocations (expansion_tests::profile_allocations)

Macro shape	Base allocs (bytes)	PR allocs (bytes)	Allocation / byte delta
#[arcane]	302 (15,525 B)	261 (12,080 B)	−13.6% / −22.2%
#[rite]	231 (12,245 B)	182 (8,794 B)	−21.2% / −28.2%
#[attune] direct family	393 (14,854 B)	299 (10,812 B)	−23.9% / −27.2%
#[attune(all)]	1,237 (45,403 B)	942 (30,943 B)	−23.8% / −31.8%
#[attune] composed	1,601 (57,214 B)	1,282 (40,952 B)	−19.9% / −28.4%
#[magetypes]	1,859 (69,687 B)	1,700 (57,721 B)	−8.6% / −17.2%
#[autoversion]	1,272 (55,933 B)	1,062 (40,890 B)	−16.5% / −26.9%

Crate expansion (rustc -Ztime-passes, median of 5 runs)

Target	Base macro_expand_crate	PR macro_expand_crate	Delta
magetypes --features "std avx512"	0.219 s	0.193 s	−26 ms (−11.9%)
1,500 #[arcane] functions + 3,000 doc lines	0.219 s	0.193 s	−26 ms (−11.9%)
400 #[attune(_*, dispatch)] families	0.456 s	0.378 s	−78 ms (−17.1%)

Verification

• cargo test -p archmage-macros --features avx512 (164 unit + raw_snapshots byte-exact expansion tests)
• cargo test --features "std avx512" (all macrotest, trybuild compile-fail UI tests, soundness_exploits, attune_*, and magetypes suites)
• cargo clippy -p archmage-macros --all-targets --features avx512 -- -D warnings
• cargo run -p xtask -- validate-registry, validate, soundness

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 79.00%. Comparing base (a4eb0cb) to head (bf379d5).

Additional details and impacted files

@@                      Coverage Diff                       @@
##           feat/neural-vision-roadmap     #170      +/-   ##
==============================================================
- Coverage                       81.15%   79.00%   -2.16%     
==============================================================
  Files                               6        6              
  Lines                            1348     1348              
==============================================================
- Hits                             1094     1065      -29     
- Misses                            254      283      +29     

Flag	Coverage Δ
aarch64	82.96% <ø> (ø)
x86_64	76.35% <ø> (-2.77%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 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.