V12.0 BASE CONTRACT

Benchmark ledger

Current performance evidence status and the minimum protocol required before comparative claims enter the manual.

There is no trustworthy cross-language benchmark dataset in the supplied v12 release material. EMPTY is better than invented evidence.

current ledger§

Claim	Status	Evidence
DByte vs C runtime	NOT PUBLISHED	none in supplied release
DByte vs Rust runtime	NOT PUBLISHED	none in supplied release
DByte vs Zig runtime	NOT PUBLISHED	none in supplied release
compiler wall time	NOT PUBLISHED AS RELEASE CONTRACT	local measurements would be machine-specific
binary size comparison	NOT PUBLISHED AS CROSS-LANGUAGE PROOF	requires identical workload and strip/link policy

minimum evidence protocol§

RECORD
  dbyte version + commit + package SHA256
  comparison compiler version + exact flags
  CPU / RAM / OS / power state
  source workload SHA256
  warmup policy
  run count
  raw per-run samples
  median + dispersion
  output correctness hash
  produced binary size + section map

REJECT
  single-run screenshots
  different algorithms
  different I/O work
  hidden compiler flags
  claims without raw samples

SOURCE POLICY: release/tag documents define contracts; examples are copied from the supplied v12.0.0 package or explicitly marked as explanatory. Unsupported claims are not upgraded by prose.