[DanielLuning]

What changed

Entering a statement computed the same read set three times:

• Copies.live_after/1, through stmt_reads/1;
• the compiler's hoist/3, through Copies.reads/1;
• Copies.enter_stmt/4, through reads/1 again.

reads/1 of a statement is its stmt_reads/1, because read_counts/1 already skips a match's left side.

• Copies.block_liveness/1 returns, for each statement of a block, {live after it, its reads} from one walk. live_after/1 is now a view of it, and its tests stay.
• The compiler's enter/3 takes {live, reads} and hands the reads to hoist/4 and to the new Copies.enter_stmt/5. enter_stmt/4 stays and computes the reads itself.
• A property test in test/copies_test.exs checks two claims on generated blocks: reads(e) == stmt_reads(e), and block_liveness/1 agrees with live_after/1 and reads/1.
• Docs: docs/performance.md 6.40, and ADR-0007's cost note.

Results

• The C: mix dump_c is byte-identical for all 108 modules. The only difference is the known path-hashed defk:loop key in Defk.ReportTest.Gen.c.
• Compiler work: measured with the DEFK_COMPILE_TIME harness (defk-o08.2: compiler time measured warm in a fresh process; the chain measured again against its budget #68), two interleaved rounds against main:
  • reductions: 125.9 M → 115.0 M (−8.6%, the same without SatEx.Kernels);
  • time: 987 → 934 ms (−5.4%).
• The budget: against the compiler before the symbolic-bounds chain, the work is now +7.4% (−1.0% without SatEx.Kernels), down from +17.6%. That is under ADR-0007's +15% ceiling again.

Provisional: the laptop was on battery, with no quiet-machine check.

How it was checked

• mix ci on the laptop (macOS) passed: 2304 tests, Credo found no issues, Dialyzer reported 0 errors.

• The byte-identical mix dump_c gate, against an untouched main checkout.

• Touches a promised surface? No. No CHANGELOG.md entry.

• Touches test/stability/? No.