Skip to content

September 2026 Core Schema evidence

Inventory

File Contents
baseline.json Five sequential fresh-process unoptimized runs, native executable and DNA hashes, compiler version
memory-4mib.json Six alternating experimental/control runs, matching native executable hash, DNA hash per variant
app-earlier.txt Selected first Core Schema call markers and nested metrics from the earlier app run
app-slow.txt Equivalent excerpts from the slow app run
sampling.md CPU sampling windows, category shares, stack observations, and limitations

JSON durations ending in _ms are milliseconds. total_ms includes harness setup; conductor_ms is the conductor call. guest_ms spans coordinator markers; after_guest_ms ends at caller receipt; before_guest_ms is derived by subtraction. Log filenames identify original source files, not downloadable attachments. Absolute workstation paths were removed from JSON; numeric results are unchanged.

Conditions and provenance

  • Runs: September 19–20, 2026; timestamps in excerpts and baseline JSON are UTC.
  • Platform: developer macOS workstation. Hardware and background load were not recorded systematically; do not assume equivalence on another machine.
  • Dependencies inspected: Holochain 0.6.3, holochain_wasmer_host 0.0.102.
  • Controlled series: native opt-level 0, debug information level 2, debug assertions enabled; RUST_LOG=info, WASM_LOG=info; no sampler attached.
  • Test: bootstrap_performance_tests::measures_generated_core_schema_bootstrap, run as a fresh process from tests/sweetests for each measurement.
  • Build: CARGO_PROFILE_DEV_OPT_LEVEL=0 CARGO_PROFILE_TEST_OPT_LEVEL=0 cargo test --locked --manifest-path tests/sweetests/Cargo.toml --test bootstrap_performance_tests --no-run --message-format=json in the Nix environment.
  • Execute the resulting test binary with --exact measures_generated_core_schema_bootstrap --nocapture from tests/sweetests.
  • DNA input: happ/workdir/map_holons.dna. Both series check artifact hashes; the memory series changes only the integrity WASM minimum from 17 to 64 pages and repacks the DNA. Native executable SHA-256 is identical across series.
  • App launch: MAP_PROFILE=1 npm run start:info 2>&1 | tee /tmp/map-package-activation-profile.log.
  • The worktree contained uncommitted implementation changes. No clean source revision or complete patch snapshot was captured for these runs. Artifact hashes identify measured binaries but do not make them rebuildable by themselves.

These are historical measurements, not a portable benchmark fixture. Reproduction requires the MAP Holochain/Nix environment and current test setup; record a clean revision or patch plus input hashes in future experiments.

Retention limits

Compact results and selected log excerpts are retained here. Full logs, scripts, CPU stack dumps, and binaries were local under /tmp/map-bootstrap-matched-profile/, /tmp/map-post-return-windows/, and /tmp/map-memory-4mib/; those paths are provenance, not durable links. Raw traces are not archived here. The earlier optimized 18.8–18.9-second observations lack an archived complete run record and should be confirmed in a controlled experiment.