|
|
|
@@ -6667,5 +6667,91 @@ and memorize frequently so we can pick up where we left off").
|
|
|
|
|
- **Seen on amd64:** after POST, `Stadium words: promotions=141`, CONSERVED; POST 539/539;
|
|
|
|
|
`word_count=377` (the two telling words); `dict_hash` changed once, for the new cell. The boot took
|
|
|
|
|
142 s against Phase 1's 153 s: no measurable cost.
|
|
|
|
|
**Corrected 2026-11-17:** "no measurable cost" was measured against Phase 1, which was already 1.5 to
|
|
|
|
|
2.4 times slower than f10d2bc8. The regression came in at 046ce0be: every fabric step walked every
|
|
|
|
|
place. See 25.
|
|
|
|
|
- **Next:** Phase 7, the loops on these records (decay, the rolling window, pipelining, inference,
|
|
|
|
|
L8).
|
|
|
|
|
|
|
|
|
|
### 25. The regression found; the fault dump; the one adaptive heartbeat (2026-11-17)
|
|
|
|
|
|
|
|
|
|
**The regression.** aarch64 had run out of its 570 s in POST since 37469f6c (990 cases). Captain Bob:
|
|
|
|
|
"find out when that happened, and actually start all over no matter how far back you need to go"; "Find
|
|
|
|
|
the TCG cost first".
|
|
|
|
|
- **Found at 046ce0be**, the mesh boot to spec. The protocol logs give the stage times, for the same
|
|
|
|
|
539-case POST:
|
|
|
|
|
|
|
|
|
|
| Arch | f10d2bc8 | 046ce0be |
|
|
|
|
|
|---|---|---|
|
|
|
|
|
| amd64 | 107 s | 184 s |
|
|
|
|
|
| aarch64 | 209 s | 499 s |
|
|
|
|
|
| riscv64 | 105 s | 163 s |
|
|
|
|
|
|
|
|
|
|
The demo's 87 s boot of 11-06 is a third baseline.
|
|
|
|
|
- **Measured** (`logs/20261117-195259/fabric-cost`): about half the boot's time is `v4_fabric_step`
|
|
|
|
|
testing every place, nine times a step, when only Hera exists.
|
|
|
|
|
- **Not a scheduler.** The rotation is ruled design (ENGINE.md 3d, ruled "A"; MESH.md ruling 7). I had
|
|
|
|
|
called it a scheduler, which was wrong.
|
|
|
|
|
- **Explained away at the time, never root-caused:**
|
|
|
|
|
- Phase 1's "about 90 s before -- the lines now go as messages";
|
|
|
|
|
- Phase 2's comparison with an already-slow Phase 1 (24, corrected);
|
|
|
|
|
- Phase 7's "the host slowed".
|
|
|
|
|
|
|
|
|
|
The cost method of record (five alternating runs, as for HEAT-OP and A2) was not used after 046ce0be.
|
|
|
|
|
- **Write-up:** `docs/v4.0.0/plans/2026-11-17-the-regression-at-046ce0be.md`. **Design:**
|
|
|
|
|
`2026-11-17-the-fabric-step-cost.md`. Mechanism A, each pass walking only the places that hold a node,
|
|
|
|
|
is ruled and follows the heartbeat.
|
|
|
|
|
|
|
|
|
|
**The CI directive** (Captain Bob, the same day):
|
|
|
|
|
- POST is as v3's;
|
|
|
|
|
- unit tests are the kernel's, with ASan and UBSan;
|
|
|
|
|
- three ISAs, one QEMU at a time, in the foreground;
|
|
|
|
|
- the EMERGENCY build is a build parameter, and its diagnostic build follows any failure;
|
|
|
|
|
- correctness before performance.
|
|
|
|
|
|
|
|
|
|
Ruled with it: **every discrepancy is put to him one at a time**, and v3's tests not reproduced as written
|
|
|
|
|
is the failure.
|
|
|
|
|
|
|
|
|
|
- **The fault dump** (`v4/system/dump.c`; 45b11aec). It prints when a node stops, in every build; a
|
|
|
|
|
handled fault keeps its one line. It shows:
|
|
|
|
|
- the fault;
|
|
|
|
|
- P, A, B;
|
|
|
|
|
- both stacks;
|
|
|
|
|
- the return stack named by word (the backtrace);
|
|
|
|
|
- the running line.
|
|
|
|
|
- **`scripts/nightly.sh diagnose`.** After any failed stage it builds the EMERGENCY configuration only,
|
|
|
|
|
keeps the artifacts, and is never the result. The protocol and the three workflows run it on failure.
|
|
|
|
|
One defect was found and fixed by its own test: it had kept an older build's files when its own build
|
|
|
|
|
failed.
|
|
|
|
|
- **`protocol-report` fixed** (68fd4854). It had ended silently, writing no PROTOCOL.md, when a boot
|
|
|
|
|
stopped before POST's line.
|
|
|
|
|
|
|
|
|
|
**The one adaptive heartbeat** (00be5375; plan `2026-11-17-one-adaptive-heartbeat.md`). Captain Bob:
|
|
|
|
|
"The adaptive heartbeat is its own thing. It has nothing to do with stadium with any interaction at all.
|
|
|
|
|
Not words, not messages, not VMs. It has its own independent RWOT and slope provider. Only one output from
|
|
|
|
|
the entire stadium is used as a load measurement for heart rate pacing. ... ONE AS IN THE NUMBER 1
|
|
|
|
|
heartbeat (the adaptive one) runs the entire system."
|
|
|
|
|
- **The kernel's heartbeat paces itself** (`heartbeat_pace_from`). Its one input is the Stadium's resident
|
|
|
|
|
heat. It keeps its own window and its own slope provider, and Loop #7 sets its own period.
|
|
|
|
|
- **Every VM ticks on it, in lockstep.** Each tick is delivered to every node, active or idle.
|
|
|
|
|
- **It is held in POST,** so v3's heat cases run as written.
|
|
|
|
|
- **The Stadium is charged as before.**
|
|
|
|
|
- **What went:**
|
|
|
|
|
- the tick by execution (Phase 7 Task 3, "By execution, as v3");
|
|
|
|
|
- the nodes' own Loop #7;
|
|
|
|
|
- the timer re-armed from Hera.
|
|
|
|
|
- **A first version was withdrawn before commit.** It put the beat on the fabric's steps, Loop #7 on
|
|
|
|
|
Hera, and the Stadium charge on the beat. It is kept as superseded in the plan.
|
|
|
|
|
|
|
|
|
|
**The protocol on cc09a88b** (`logs/20261117-205521/PROTOCOL.md`): **FAIL**.
|
|
|
|
|
|
|
|
|
|
| Stage | Result |
|
|
|
|
|
|---|---|
|
|
|
|
|
| amd64, riscv64, the stick | POST 990/990, one `dict_hash` (`0x627f392ccbaa96f1`) |
|
|
|
|
|
| test-32 | PASS |
|
|
|
|
|
| aarch64 | STOPPED at 570 s in POST: the fabric step's cost, unchanged by the heartbeat |
|
|
|
|
|
|
|
|
|
|
Pushed at Captain Bob's word (8660ab77).
|
|
|
|
|
|
|
|
|
|
**Next:** mechanism A, test first, measured by five alternating runs, then the protocol.
|
|
|
|
|
|
|
|
|
|