shakedown: wave 7: roadmap completion (non-accel / memory / GPU phases, 14/16 thermal, M1-M4 + Intel calibrations)

.github/ISSUE_TEMPLATE/defect-report.yml

@@ -1,6 +1,6 @@
 name: Defect report
 description: Share a defect Shakedown found on your Mac
-title: "Defect report — [chip/model]"
+title: "Defect report: [chip/model]"
 labels: ["defect-report"]
 body:
   - type: markdown
@@ -41,16 +41,16 @@ body:
     attributes:
       label: Which phase failed
       options:
-        - "0 — Pre-flight"
-        - "1 — Inventory"
-        - "2 — Battery"
-        - "3 — Sensors"
-        - "4 — CPU variance"
-        - "5 — Thermal load"
-        - "6 — Display"
-        - "7 — Physical inspection"
-        - "8 — Apple Diagnostics"
-        - "9 — Idle drain"
+        - "0: Pre-flight"
+        - "1: Inventory"
+        - "2: Battery"
+        - "3: Sensors"
+        - "4: CPU variance"
+        - "5: Thermal load"
+        - "6: Display"
+        - "7: Physical inspection"
+        - "8: Apple Diagnostics"
+        - "9: Idle drain"
         - Multiple
     validations:
       required: true
@@ -76,7 +76,7 @@ body:
     id: notes
     attributes:
       label: Anything else
-      description: Context that might matter — store, ambient temp, repro pattern, etc.
+      description: Context that might matter: store, ambient temp, repro pattern, etc.
     validations:
       required: false
 

.github/workflows/lint.yml

@@ -8,7 +8,7 @@ on:
 
 jobs:
   lint:
-    # Static checks only — scripts target macOS at runtime, but shellcheck,
+    # Static checks only, scripts target macOS at runtime, but shellcheck,
     # python ast.parse, and json.tool are all platform-agnostic.
     runs-on: ubuntu-latest
     steps:
@@ -29,7 +29,7 @@ jobs:
 
       - name: Syntax-check Python heredocs
         # Each script may contain one or more `python3 - <<'PYEOF' … PYEOF` blocks.
-        # Extract them with awk and run ast.parse — catches Python syntax errors
+        # Extract them with awk and run ast.parse, catches Python syntax errors
         # without executing macOS-only side effects (system_profiler, ioreg, etc.).
         run: |
           set -euo pipefail
@@ -68,9 +68,11 @@ jobs:
                           "target", "unit", "phases", "result", "result_reason",
                           "submission_safe"]
           REQUIRED_PHASES = ["0_preflight", "1_inventory", "2_battery", "3_sensors",
-                              "4_cpu_variance", "5_thermal_load", "6_display",
-                              "7_physical", "8_apple_diagnostics", "9_idle_drain",
-                              "10_race_bench", "11_ssd_test"]
+                              "4_cpu_variance", "4b_cpu_variance_noaccel",
+                              "5_thermal_load", "6_display", "7_physical",
+                              "8_apple_diagnostics", "9_idle_drain",
+                              "10_race_bench", "11_ssd_test", "12_memory_bandwidth",
+                              "13_gpu_variance"]
           FILENAME_RE = re.compile(r"^\d{4}-\d{2}-\d{2}-[a-z0-9-]+-[0-9a-f]{4}\.json$")
 
           def find_keys(obj, predicate, path=""):
@@ -136,7 +138,7 @@ jobs:
               for e in errors:
                   print(f"  - {e}", file=sys.stderr)
               sys.exit(1)
-          print(f"audited {len(files)} submission file(s) — all clean")
+          print(f"audited {len(files)} submission file(s), all clean")
           PYEOF
 
       - name: Verify markdown links resolve to existing files
@@ -147,7 +149,7 @@ jobs:
         #
         # Implemented in Python instead of bash because the previous bash
         # version used `${path//%/\\x}` parameter-expansion escapes that
-        # behave differently in zsh — false positives for contributors
+        # behave differently in zsh, false positives for contributors
         # running this check locally on default-shell macOS.
         run: |
           python3 <<'PYEOF'

CHANGELOG.md

@@ -4,6 +4,18 @@ All notable changes to this project are documented here. Format follows [Keep a
 
 ## [Unreleased]
 
+### Added: roadmap completion (wave 7)
+
+- **Phase 4b non-accelerated CPU variance** (`cpu-variance.sh WORKLOAD=blake2b`, opt-in `./run --noaccel`). Reruns the Phase 4 variance methodology with BLAKE2b, which has no dedicated CPU instruction and so exercises the integer pipelines and memory instead of the SHA engine. Catches batch defects that SHA-NI / the Apple crypto coprocessor hide. Same script and verdict logic as Phase 4 (no duplicated thresholds): real `pass`/`warn`/`fail` when run, `skipped` placeholder otherwise. Runs on the already-hot chassis after Phase 4 with a short re-warm and no cold burst. Adds ~6 min when enabled.
+- **Phase 12 memory bandwidth** (`memory-bandwidth.sh`). Multi-threaded sequential `ctypes.memmove` copy, one worker per P-core, over buffers sized to exceed the system-level cache (DRAM-bound, not cache-bound). Reports copy bandwidth, touched (read+write) bandwidth, and run-to-run spread. Pure Python stdlib with no native kernel, so it's an honest lower bound on a full STREAM triad (a single thread can't saturate Apple Silicon's multi-channel controllers, hence multi-threaded the way STREAM uses OpenMP). RAM-guarded: the working set is capped at 40% of physical memory, and it skips if it can't size cache-busting buffers. `info` verdict in v0.2. Runs cold, ~15 s.
+- **Phase 13 GPU compute variance** (`gpu-variance.sh`, opt-in `./run --gpu`). Compiles a small Metal compute kernel with `swiftc` at runtime (no binary shipped) and runs sustained FMA work, measuring per-iteration GFLOP/s and variance. The GPU is the bigger thermal contributor on Apple Silicon. This is the one phase that isn't pure bash + Python stdlib; the Metal source is inline in the script heredoc. Degrades to `skipped` (never fails the run) when `swiftc`, the Metal compiler, or a Metal device is unavailable (headless / SSH session, or an Intel Mac without a usable GPU context). `info` verdict in v0.2. The default `./run` is unchanged and stays pure-script; see [SECURITY.md](SECURITY.md).
+- **SSD conservative floor.** Phase 11 (`ssd-test.sh`) now emits `warn` (advisory, never fail) when read or write is below 500 MB/s with the page cache dropped, recorded as `ssd_floor_mb_per_s`. Every NVMe Mac since ~2016 clears > 1000 MB/s, so this flags the documented 256 GB single-NAND-die regression (M2 Air / base 13" MBP), a failing drive, or pre-NVMe storage. Calibrated chassis-family bands still land in v0.3.
+- **14" vs 16" thermal sub-classes.** Split `active-cooled-pro` into `active-cooled-pro-14` (looser steady-state and cliff bands; the 14" M5 Max throttles by design) and `active-cooled-pro-16` (the previous strict thresholds). The bare `active-cooled-pro` stays valid as an alias for the 16" table, so existing presets and prior submissions stay comparable. Auto-detect can't distinguish 14" from 16" (system_profiler doesn't expose screen size on Apple Silicon), so the precise sub-class comes only from a `--target` preset; `mbp-14-m5-pro-24` and `mbp-16-m5-max-64` now set it explicitly.
+- **Per-core pinning investigation** (`Verification/per-core-pinning.md`). Documents why macOS has no public per-core affinity API (`THREAD_AFFINITY_POLICY` is ignored on Apple Silicon, QoS only steers between the P and E clusters), why a single defective core gets averaged across workers, and why `worker_imbalance_pct_per_iter` is the chosen partial mitigation.
+- **Generation calibrations: M1-M4 and Intel/T2 era.** Added `examples/m1-2020/`, `examples/m2-2022/`, `examples/m3-2023/`, `examples/m4-2024/`, and `examples/intel-2016/`, each a documented, sourced defect landscape in the `examples/m5-2026/` format (overview, per-issue notes, sources). Highlights: M1 24" iMac display-cable lines and 14"/16" speaker crackle; the M2 256 GB single-die SSD regression; M3's mostly-clean generation (8 GB RAM and M3 Pro bandwidth are spec choices, not defects); M4's OS-gated JIT kernel panic; and the Intel/T2 era's butterfly keyboard, Flexgate, T2 audio/panics, 2018 i9 throttling, and 2015 battery recall. Each issue maps to the Shakedown phase that surfaces it. Added representative target presets (`macbook-air-m1`, `macbook-air-m2`, `mbp-14-m3-pro`, `mac-mini-m4`) and repointed `mbp-16-intel-2019` at the new `examples/intel-2016/` calibration.
+- **Schema bump to 1.2.** New phase keys (`12_memory_bandwidth`, `4b_cpu_variance_noaccel`, `13_gpu_variance`) added to the CI submission-audit required-phase list; the orchestrator always emits them (opt-in phases as `skipped` placeholders), so default-run submissions still pass. Backward compatible: pre-1.2 reports validate against the 1.1 shape.
+- **Comparability impact.** The new `info` phases don't change existing verdicts. The SSD floor can turn a previously-`info` SSD result into `warn` (a PASS-with-warns) on genuinely slow drives. The `active-cooled-pro` → `-16` aliasing preserves prior thermal verdicts exactly; a 14" run with `--target mbp-14-...` is now judged against looser bands than before (it previously used the strict 16" table).
+
 ### Added: performance benchmarks (wave 6)
 
 - **Phase 10 race benchmark (`race-bench.sh`).** Compresses a 200 MB incompressible random blob with `xz -9 -T<P-cores>` and reports wall-clock seconds plus MB/s throughput. Unlike Phase 4 (SHA-256, hardware-accelerated on Apple Silicon), LZMA is a general-purpose CPU workload that does not benefit from SHA-NI or Apple's crypto coprocessor, so the number is comparable across chassis families. Runs cold (before Phase 4) so boost headroom isn't already burned. Skipped gracefully if `xz` isn't on PATH.

CONTRIBUTING.md

@@ -29,7 +29,7 @@ A target preset lets users run `./run --target <name>` to assert the unit matche
 | `memory_gb` | Asserted against `sysctl hw.memsize` |
 | `model_must_include` | Substring asserted against `machine_model` (catches chassis size) |
 | `calibration_dir` | Path to the `examples/<generation>/` notes for failure-analysis cross-reference |
-| `thermal_chassis_class` | One of `fanless`, `active-cooled-pro`, `desktop`, `intel-laptop`, `intel-desktop`. Sets thermal thresholds and Phase 4 warmup duration. See [`targets/README.md`](targets/README.md) for the per-class definition. |
+| `thermal_chassis_class` | One of `fanless`, `active-cooled-pro` (with `active-cooled-pro-14` / `active-cooled-pro-16` sub-classes for the 14" / 16" MacBook Pro), `desktop`, `intel-laptop`, `intel-desktop`. Sets thermal thresholds and Phase 4 warmup duration. See [`targets/README.md`](targets/README.md) for the per-class definition. |
 
 Open a PR adding the JSON. Match an existing one for style.
 
@@ -52,7 +52,7 @@ The verification scripts and runbook **don't need to change**. They're calibrati
 
 ## Improving a script
 
-The 5 scripts in `Verification/scripts/` are deliberately self-contained: bash plus Python heredocs, no external dependencies beyond what ships with macOS. Keep it that way. If you need to add a third-party benchmark (Geekbench, sysbench), make it optional with a fallback to the built-in equivalent.
+The scripts in `Verification/scripts/` are deliberately self-contained: bash plus Python heredocs, no external dependencies beyond what ships with macOS. Keep it that way. If you need to add a third-party benchmark (Geekbench, sysbench), or something that genuinely needs more than the stdlib (the opt-in `--gpu` phase compiles a Metal kernel with `swiftc` at runtime), make it optional and have it skip cleanly when the dependency is absent, so the default `./run` stays pure bash plus Python stdlib.
 
 ### Validating changes to a script