igneum-network/igneum
1
0
Fork 0
Code Issues Pull requests Releases Activity

Merge x2-1p5x-landing ee5582a5 into master (gate: green on 12354ed3, recorded by tools/ci/pre-push.sh; landed on the build mirror)

Browse source
This commit is contained in:
igneum-labs 2026-10-09 09:34:33 +00:00
commit 4fb77ea8dd
96 changed files with 45840 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

221
docs/analysis/class-v6/1p5x/x2/README.md Normal file
View file

@ -0,0 +1,221 @@
# X2: serving-mode result compaction on the frozen control (9 October 2026)
The 1.5x programme's item X2 (the brief: `docs/plans/igneum-2.0-master/1p5x/`, sha256 74ee073286d4a6e2..., its experiment
matrix row X2: "exact output equivalence incl. overflow/sentinels; wall joules and accepted throughput; no invented
percentage"). The lane: X2 (the worker lane of the programme). Branch `x2-compaction` off class-v6 755c2dbcf. Clock 09:32 to
11:30 UK. Rows in UTC; "on an island, not a network" does not apply (no row here is taken on the devnet).
## 0. The premise, read first: the compaction is already in the control
The frozen control 1a938abe4 already carries a serving-mode compaction, ON by default, in the CUDA kit worker:
`proto-cuda/nvrtc/worker.cpp` at 12cb65cc4 (review B finding F10, landed on master as 964cdf6c; the file is unchanged from
12cb65cc4 to 1a938abe4, sha256 of the text 9bfabc716b217f5e). A tiny NVRTC kernel `igneum_select` scans the bound kernel's
per-nonce output on the GPU, writes the hits (chunk index, hash word) behind an atomic counter into a 256-entry table plus 34
sentinel words (the first 32 outputs, the middle and the last, for the stale-output detector), and the host reads back the
count, the hits and the sentinels. A chunk with more than 256 hits falls back to the full read. `--readback full` or
`IGNEUM_READBACK=full` is the uncompacted path (8 bytes per nonce over the bus, the host scans). The OpenCL worker has had
the same pass since 5 October 2026. F10 measured it at stock on a 5090 and a 3090
(`docs/analysis/review-2026-10-08-b/f10/rows.md`).
So X2 adds no flag and changes no worker code (the flag exists; turning it off by default would undo a landed fix), touches
nothing in consensus, the generator or any `.rs` or Cargo file (rule 24 is not triggered), and delivers what F10 did not:
the equivalence on the three same-work contexts against the CPU references with the sentinel and overflow cases enumerated,
the 4090 same-session pairs, and the arithmetic against the yardstick. The coordinator confirmed this shape at 09:5x UK.
What the compaction does NOT remove, read from the code: the bound kernel still writes 8 bytes per nonce to device memory
(`out[g] = ...` in every pack's `kernel_bound.cu`) and the select kernel reads them back once; fusing the compare into the
emitted kernel would change the emitter's output (`igneum-pow/src/emit.rs`), which this lane does not touch. Against the
class v6 hash's own traffic (256 dependent 4-byte loads per hash under reg64, `program.json` `loads_per_hash`, each a 32-byte
sector on NVIDIA) those 16 bytes are 16 / 8,192 = 0.2 percent of the bytes a hash moves, a byte count from the code and not
an energy figure; no row here measures it.
## 1. Software, hardware, boundary
| Item | Identity |
|---|---|
| Worker | `igneum-worker-cuda` built on the pod from 1a938abe4's `proto-cuda/nvrtc/worker.cpp` and `packfile.h` by the kit recipe's Linux line (`tools/class-v5/kits-on-box.sh` line 25: `g++ -std=c++17 -O2 -Wall -Wextra -I proto-cuda/nvrtc -I /usr/local/cuda/include ... -ldl -lpthread`), Ubuntu 24.04, g++ 13.3.0; sha256 `4152c960d14210072a5c85ae675b56ae4dab12bedfb5f1af04ea2a2c6c025032`. Not a kit-lane cut: no Linux CUDA binary on build-1 carries both generator 6 and the select pass (`/srv/artefacts/packs/igneum-worker-cuda-gen6`, 804a6f7f, is generator 6 with no `igneum_select`; the F10 binary 9b36b523 refuses generator 6 packs) |
| Packs and references | the same-work contexts on build-1 `/srv/artefacts/tas/same-work-20261008-0{1,2,3}/` (job-context.json, packs/, references/), copied whole; reference sha256 read back equal to the published `.sha256` files: 01 D c49fed11 / D1 5ae1220d, 02 D 7c7be9b5 / D1 cad7849f, 03 D 873528b1 / D1 7f8e64ae |
| Objects | 01: class v6, generator 6, the frozen-object context (pack hl-v6-all, program id 0x4de7b836cc40a4ea on day D+1); 02: class v6, the chain-seed object 0x2a1d6caab4c24564 (the brief's signing id, epoch seed af89be5d, era edc4fa84, day 20730); 03: class v5, epoch 2, 0x81fbfa3aa413173f. Dataset 2^28 words (1 GiB) in every pack, as the same-work packs were exported |
| Hardware | RTX 4090 24 GB, driver 595.91.07, NVRTC 12.8, sm_89, 128 SMs, PCIe link read at idle as gen 1 x16, RunPod secure instance lyum23xjpv4re9 (the fleet lane's fk-x2-4090), 450 W board limit |
| Power boundary | the board only: nvidia-smi `power.draw` at 1 Hz. The container exposes no host rail and no wall meter, so the WALL row is BLOCKED (section 5); the worker's own CPU seconds stand in for the host side, as in F10 |
| State | stock only. `nvidia-smi -lgc 1300` and `-pl 300` answered "does not have permission" / "Insufficient Permissions" on this host (read 10:0x UK); the knee row is NOT RUN (section 5) |
## 2. Equivalence (requirement 1): EQUAL on all three contexts
The driver `tools/1p5x/x2/x2-equiv.py` runs one worker binary in `--serve` exactly as the miner drives it (job lines on
stdin, `prepare` of pack D1 before the first day-D1 job, the worker's report of a finished prepare read before switching),
writes the normalized protocol output, and builds the EXPECTED output from the CPU references alone: for each job, every
nonce with `ref[n] <= target` in nonce order, then `done <job> <count>`. A verdict is EQUAL only when the full run, the select
run and the expected file are byte-identical, which is every nonce's verdict at every target and every found nonce's digest.
Campaign: `tools/1p5x/x2/run-equiv.sh` plus `run-prod-then-pairs.sh` (the prod plan's second take, section 2.3), on the pod
at 08:45 to 08:51 UTC.
### 2.1 The known-failed tests, first
| Test | Result |
|---|---|
| The driver's self-test (`x2-equiv.py --self-test`: a fake worker over a fake hash on all three plans) | PASS: the clean fake is EQUAL on every plan; five injected faults each compare DIFFERENT: the 257th hit of a chunk dropped (the overflow taken as a full table), one digest wrong, the previous chunk's hits returned again (the counter-reset fault), the day switch missed (D's dataset after the boundary), `<` for `<=` (the target-equality edge). Log: `evidence/self-test.log` |
| The real worker's fault run (`--readback select` with `IGNEUM_READBACK_FAULT_TEST=1`: the hit counter is not reset from the second chunk on), context 01, the kf plan (four 2,048-nonce jobs at a 1/16 target on 1,024-nonce chunks) | flagged: DIFFERENT at line 64 (`found 1 1037 00cb3ac5...` where the reference has `found 1 1053 0fa540bb...`), 368 extra lines; the same plan's full and select runs EQUAL to the references. Only then did any select row count |
A finding about the fault injection itself: once the un-reset count passes 256, every later chunk takes the full-read
fallback and prints the right answer, so the counter-reset fault is invisible on a plan whose early chunks overflow (the
fake reproduces this: the stale fault reads EQUAL on the prod and fuzz plans). The kf plan exists for that reason.
### 2.2 The rows
| Context | Plan | Jobs | Nonce evaluations | Found lines | Select overflow fallbacks | Normalized output sha256, expected = full = select |
|---|---|---|---|---|---|---|
| 01 (class v6, frozen object) | prod (2^22 chunks) | 16 | 9,437,568 | 3,146,498 | 5 of 16 chunks | f278f0594dfd566d... |
| 01 | fuzz (2^10 chunks) | 38 | 1,048,576 | 438,688 | 534 of 1,039 | 5f1e591db673f964... |
| 02 (class v6, chain-seed 0x2a1d6caab4c24564) | prod | 16 | 9,437,568 | 3,146,498 | 5 of 16 | ad091a0d49a9d3a2... |
| 02 | fuzz | 35 | 1,048,576 | 446,422 | 599 of 1,042 | 8f39a1b15b16e86f... |
| 03 (class v5, epoch 2) | prod | 16 | 9,437,568 | 3,146,498 | 5 of 16 | 2aa97d5d63e5e01c... |
| 03 | fuzz | 33 | 1,048,576 | 342,786 | 365 of 1,035 | 3a7a7cc495e2d485... |
Every verdict EQUAL; full sha256s in `evidence/*.compare.json` and `evidence/norm.sha256`; the normalized outputs (up to
103 MB each, 31 files) at build-1 `/srv/artefacts/tas/x2-20261009/x2-equiv-outputs.tar.gz` (sha256 64740f73213b9deb...).
What each plan covers. **prod** (the worker's and the miner's 2^22 batch): every nonce of [0, 3,145,728) at the all-ones
target (every digest of the context through the overflow fallback); phase 1 at the sentinel targets T256 (the 256th smallest
hash of the job: exactly 256 hits, the table full, no fallback), T257 (257 hits: overflow by one, the fallback), 0, Teq (the
job's smallest hash: one hit, the `<=` edge) and Teq - 1 (zero hits); the day switch inside phase 2 at nonce 1,572,864 by
`prepare` (the worker's `prepared` line in `evidence/*.prod.*.meta.json`); phase 3 at all-ones and at its own T256. The 5
fallbacks per select run are the five all-ones and T257 jobs, and the T256 jobs went through the table, which is the 256/257
edge read on the real worker. **fuzz** (1,024-nonce chunks): 1,048,576 nonces per context in 29 to 34 seeded random jobs (seed
`x2-fuzz/<run id>`) at targets drawn from {0, all-ones, 2^k - 1 for k in 50 to 63, 1/8, 1/4, an in-job hash, that hash minus
1}; the 1/4 target puts about 256 hits in a chunk, so overflow and table chunks alternate inside one job; plus the 8,192-nonce
boundary block ([1,568,768, 1,572,864) on D, `prepare`, [1,572,864, 1,576,960) on D1) at all-ones and 1/4.
The sentinels enumerated, each with the plan that reaches it: target 0 (prod p1-zero, fuzz fz0); all-ones, every chunk over
the table (prod p1-all, p2, p3-all; fuzz fz1 and the boundary block); exactly 256 hits (prod p1-t256, p3-t256); 257 hits
(prod p1-t257); the `<=` edge (prod p1-teq; fuzz fz5, fz6); one below it (prod p1-teq-1; fuzz fz7); overflow and table chunks
in one job (fuzz fz4 at 1/4); the day switch (prod p2; fuzz boundary block); the 34 sentinel words feeding the stale-output
detector on both paths (every chunk of 64 nonces or more). Not reached here and covered by F10's emulation check
(`proto-cuda/nvrtc/emu/select-check.sh`, green on build-7 on 8 October): the high-32 rollover, the 96-nonce tail with a
32-lane launch, two queued jobs, a mismatch error then a job.
### 2.3 Finding X2-F1: identical work twice in a row restarts the worker (both paths)
The prod plan's first take sent phase 1 [0, 2^20) six times in a row at six targets. The second job of each context was
answered `error 2 worker fault: the output buffer is unchanged since the previous dispatch (signature ef40224d5958063d): the
kernel did not run; exiting 3 so the miner restarts the worker`, on the full and the select path alike (`evidence/prod-v1/`).
The stale-output detector (`worker.cpp`, the first, middle and last output words plus an FNV of the first 32, compared with
the previous chunk's) assumes two consecutive chunks never carry the same work. A job re-sent with the same prehash and nonce
range breaks that assumption and the worker exits 3. Meaning: a miner or pool that re-sends an identical job (a reconnect
replay, a duplicated line) costs a worker restart, the NVRTC compile, cache and dataset build again (about 4 s on this card,
the `prepared` timings). Not a compaction fault and not fixed here (worker code, a separate owner); the plan's second take
puts a 64-nonce job of other nonces at target 0 between repeats. Owed: the worker lane's fix (the signature keyed with the
job's prehash and nonce base, or the check skipped when the job equals the previous one) with this repeat as its known-failed
case.
## 3. Measured rows (requirement 2)
Instrument: `tools/1p5x/x2/run-pairs.sh` driving `tools/1p5x/x2/serve-pair.sh` (F10's `proto-cuda/nvrtc/serve-bench.sh` at
1a938abe4 with nvidia-smi at 1 Hz, the batch at the worker's and the miner's 2^22, and the share check added): the worker in
`--serve` fed by a stub pool exactly as the miner feeds it (one job queued behind the one in flight), 1,024 jobs of 2^22
nonces per run (4,294,967,296 hashes, about 125 to 137 s), share target `00000fffffffffff` (one share per 2^20 nonces), every
found line's hash checked against the target at the submit point; `--race off`, variant base, 88 registers, 20 blocks per
SM. Four pairs in ABBA order (full, select / select, full / full, select / select, full), 5 s apart, 08:52 to 09:11 UTC; then
the kernel timer (`--bench --batch-log2 24 --batches 250`, the instrument behind the record's E_GPU rows) with nvidia-smi at 1
Hz around it, 09:11 to 09:13 UTC. Pack: the frozen object's own, context 02 `chain-seed-d20730` (0x2a1d6caab4c24564, day
20730, 1 GiB dataset). The card held 2,760 to 2,775 MHz SM in every run (no throttle; 270 to 290 W of a 450 W limit, the
memory chain bounds the hash on Ada as the connected-state rows found). Steady = from the worker's ready line to the last
done line; production = from the process start (setup about 4 s). `worker_cpu_s` reads 0.00 because the script reads
`/proc` of the coproc's subshell, not the worker: the host-side CPU row is not read (F10 read 2.3 to 2.6 s against 2.9 to
3.1 s per 32 jobs of 2^24 on the 5090). Generated by `python3 tools/1p5x/x2/summarize.py evidence/pairs` (the file
`evidence/pairs/summary.txt`):
| run | mode | prod_mhs | steady_mhs | steady_mean_w | nj_per_hash_steady | accepted | over_target | j_per_share_steady | steady_j_per_batch | worker_cpu_s | samples |
|---|---|---|---|---|---|---|---|---|---|---|---|
| x2-full-r1 | full | 30.78 | 31.75 | 268.4 | 8453.5 | 4223 | 0 | 8.598 | 35.5 | 0.00 | 135 |
| x2-full-r2 | full | 31.14 | 32.08 | 273.1 | 8513.1 | 4223 | 0 | 8.659 | 35.7 | 0.00 | 134 |
| x2-full-r3 | full | 30.46 | 31.37 | 269.9 | 8603.8 | 4223 | 0 | 8.750 | 36.1 | 0.00 | 136 |
| x2-full-r4 | full | 30.74 | 31.65 | 270.2 | 8537.1 | 4223 | 0 | 8.683 | 35.8 | 0.00 | 136 |
| x2-select-r1 | select | 33.66 | 34.88 | 288.0 | 8256.9 | 4223 | 0 | 8.397 | 34.6 | 0.00 | 123 |
| x2-select-r2 | select | 33.78 | 34.89 | 290.0 | 8311.8 | 4223 | 0 | 8.454 | 34.9 | 0.00 | 123 |
| x2-select-r3 | select | 33.77 | 34.89 | 289.3 | 8291.8 | 4223 | 0 | 8.434 | 34.8 | 0.00 | 123 |
| x2-select-r4 | select | 33.70 | 34.78 | 287.2 | 8257.6 | 4223 | 0 | 8.398 | 34.6 | 0.00 | 124 |
| mode | runs | steady MH/s (sd) | production MH/s | steady W | nJ per hash (sd) | J per accepted share | accepted shares per run |
|---|---|---|---|---|---|---|---|
| full | 4 | 31.71 (0.29) | 30.78 | 270.4 | 8526.9 (62.2) | 8.672 | 4223 |
| select | 4 | 34.86 (0.05) | 33.73 | 288.6 | 8279.5 (27.0) | 8.421 | 4223 |
paired g (select over full, nJ per hash) per pair: 0.9767, 0.9764, 0.9637, 0.9673; mean of means 0.9710
paired rate ratio (select over full) per pair: 1.0986, 1.0876, 1.1122, 1.0989; mean of means 1.0993
kernel timer: 35.057 MH/s, 289.8 W loaded, 8267.3 nJ per hash, fingerprint 01f51b9d4805e5e6
g against the kernel timer (select serving over --bench): 1.0015; full serving over --bench: 1.0314
| r0 (the brief's current ratio) | r_new with g = select/full | r_new with g = select/kernel timer |
|---|---|---|
| 1.6 | 1.554 | 1.602 |
| 2.0 | 1.942 | 2.003 |
| 2.4 | 2.330 | 2.404 |
(a = 1: the specialist side is not touched by a host-side result path; the identity r_new = r0 * g / a, the brief.)
What the rows say. The compacted path serves 9.9 percent more accepted shares per second (paired 1.088 to 1.112) and spends
2.9 percent fewer board joules per hash and per accepted share (paired g 0.964 to 0.977): the card draws 18 W more because
it is never idle between chunks, and finishes the same work 9.9 percent sooner. The uncompacted path's cost is visible in the
worker's own transfer line: 32 MiB read back and a host scan of 2^22 words per chunk, during which the GPU waits (the
select path reads back 4 bytes, 34 sentinel words and 16 bytes per hit). The compacted path reaches the kernel timer: 34.86
against 35.06 MH/s (99.4 percent), 8,280 against 8,267 nJ per hash. Accepted shares per run are 4,223 on both paths with 0
over the target; the equivalence of section 2 is what makes a found line on either path the same share.
Beside F10 (2^24 batch, 4 Hz, pack hl-v6-all): 5090 rate +10.5 percent and J per batch -2.8 percent; 3090 +8.8 and -6.7.
The 4090 here: +9.9 and -2.9. Three cards, one direction, the joule saving 2.8 to 6.7 percent and the rate 8.8 to 10.5.
Consequences per tier (the standing rule). A home miner on any NVIDIA card running a CUDA worker without the select pass
loses about a tenth of its serving rate and 3 to 7 percent of its joules per share to the full read; the gen-6 Linux CUDA
worker on build-1 (`/srv/artefacts/packs/igneum-worker-cuda-gen6`, 804a6f7f, built from db63e1e36 before the class-v6 merge
of 964cdf6c) carries no `igneum_select` (strings read: 0), so a class v6 kit cut from 1a938abe4 is what moves those miners
(the kit lane's cut, not this lane's). AMD and Intel on the OpenCL worker have had the pass since 5 October. Apple's Metal
worker reads from unified memory, so its read-back is not a bus transfer; it is not measured here. A pool user sees the same shares sooner; the pool
sees nothing else (the found line is unchanged).
## 4. The arithmetic (requirement 3)
The brief's identity: r_new = r0 * g / a, with g and a the new-to-old GPU and specialist energies per equivalent accepted
work. The compaction is a host-side result path; nothing in it touches the specialist, so a = 1 (stated, not measured). Two
values of g, because the yardstick matters:
| g | Value (4090, stock) | Meaning |
|---|---|---|
| select serving over full serving | 0.971 (paired 0.964 to 0.977) | what a miner on an uncompacted worker gains by moving to the compacted one |
| select serving over the kernel timer | 1.0015 | what the compaction does to the E_GPU the ratio is computed from |
| full serving over the kernel timer | 1.031 | the serving loss the full read adds above the yardstick |
| r0 (the brief's ratios) | r_new, g = select/full | r_new, g = select/kernel timer |
|---|---|---|
| 1.6 (the pure DRAM board, node-for-node) | 1.554 | 1.602 |
| 2.0 (the brief's worked row) | 1.942 | 2.003 |
| 2.4 (the stored-half hybrid, the N2 SRAM die) | 2.330 | 2.404 |
Reading. The record's E_GPU rows come from the kernel timer (`igneum-worker-cuda --bench`, 250 x 2^24, nvidia-smi 1 Hz;
connected-state.md section 4, the l8off knee rows), which never includes the read-back, so the full read's 3 percent was
never in r0. Against the ratio's own yardstick X2 moves r by nothing measurable (g = 1.0015, inside the paired spread). The
brief's 1.5x target needs g = 0.75 from a ratio of 2.0 with a = 1; the compaction supplies 0.971 only against a worse
baseline that the ratio does not use. No percentage here is invented: every figure is the measured difference above, on one
card at stock. The saving the brief asked about is real for miners and zero for the ratio.
## 5. BLOCKED and NOT RUN rows
| Row | Decision | Reason | What unblocks it |
|---|---|---|---|
| Wall joules (the brief's word) | BLOCKED | a rented container has no wall meter and no host rail; every joule here is the board's nvidia-smi reading, and the host side is the worker's CPU seconds | a host with a wall meter on the card's machine |
| The knee (the 1,300 MHz lock or a power cap) | NOT RUN | no clock lock on a rented pod (`nvidia-smi -lgc` and `-pl` refused on fk-x2-4090, and on every 8 October pod by the fleet lane's record); the house rigs unreachable (the coordinator's amendment, 10:4x UK); the CUDA worker has no in-kernel throttle | a host that permits `-lgc` |
| The 5090 and other tiers on this harness | NOT RUN | one pod for this lane; the 5090 and 3090 pairs are F10's at stock (2^24 batch, 4 Hz) | another pod on the same scripts |
| The fused compare (no per-nonce write at all) | NOT RUN | it changes the emitter's output (`emit.rs`), outside this lane | a lane that owns the emitter, after X1 |
| CPU re-check of the bench's accepted shares | NOT RUN | the bench jobs use prehash = job index, which the references do not cover; the shares are counted as found lines whose hash is at or under the share target (all of them, `over_target=0`); the hash word itself is the one section 2 proves byte-identical to the CPU reference | `igneum-pow hash-bound` for the found nonces on a box |
| X2-F1 fix | NOT RUN | worker code, separate owner | the worker lane |
| Registry | NOT RUN | `tools/1p5x/x2/registry-batch-x2-1p5x-20261009.json`, both cells NOT RUN, method GPU, evidence by file; kept out of `tools/ci/batches/` so the landing writes no registry row; the steward records | the steward |
## 6. Commands
```
# on the GPU host, under /root/x2/pids (pid files; never on the Mac)
g++ -std=c++17 -O2 -Wall -Wextra -I proto-cuda/nvrtc -I /usr/local/cuda/include -o bin/igneum-worker-cuda proto-cuda/nvrtc/worker.cpp -ldl -lpthread # tree: git archive 1a938abe4
python3 tools/1p5x/x2/x2-equiv.py --self-test
tools/1p5x/x2/run-equiv.sh bin/igneum-worker-cuda ctx out/equiv pids # kf known-failed first, then prod and fuzz per context
tools/1p5x/x2/run-pairs.sh bin/igneum-worker-cuda ctx/same-work-20261008-02/packs/chain-seed-d20730 out/pairs pids 4 1024 22
python3 tools/1p5x/x2/summarize.py out/pairs # every figure of sections 3 and 4
```

31
docs/analysis/class-v6/1p5x/x2/evidence/norm.sha256 Normal file
View file

@ -0,0 +1,31 @@
5f1e591db673f964ff53a60ebb9ea0e8f4cad974642a2354b3cc27571d935340 equiv/same-work-20261008-01.fuzz.expected.norm
5f1e591db673f964ff53a60ebb9ea0e8f4cad974642a2354b3cc27571d935340 equiv/same-work-20261008-01.fuzz.full.norm
5f1e591db673f964ff53a60ebb9ea0e8f4cad974642a2354b3cc27571d935340 equiv/same-work-20261008-01.fuzz.select.norm
e3e8913fb6788dfbbba4d43de14d674fa411de019980e2fc543be4ed7c17303c equiv/same-work-20261008-01.kf.expected.norm
f7ad6ee87564f49abaee8e8ee64b62e41c6a45259c0cace233aecb898cbd78aa equiv/same-work-20261008-01.kf.fault.norm
e3e8913fb6788dfbbba4d43de14d674fa411de019980e2fc543be4ed7c17303c equiv/same-work-20261008-01.kf.full.norm
e3e8913fb6788dfbbba4d43de14d674fa411de019980e2fc543be4ed7c17303c equiv/same-work-20261008-01.kf.select.norm
f278f0594dfd566d48fb59ee6b2f2889c60afb341856cf9efd53cce0d57ad60f equiv/same-work-20261008-01.prod.expected.norm
f278f0594dfd566d48fb59ee6b2f2889c60afb341856cf9efd53cce0d57ad60f equiv/same-work-20261008-01.prod.full.norm
f278f0594dfd566d48fb59ee6b2f2889c60afb341856cf9efd53cce0d57ad60f equiv/same-work-20261008-01.prod.select.norm
8f39a1b15b16e86f09bedeaa9d2f811cd6329012a71139e93babd5199f468938 equiv/same-work-20261008-02.fuzz.expected.norm
8f39a1b15b16e86f09bedeaa9d2f811cd6329012a71139e93babd5199f468938 equiv/same-work-20261008-02.fuzz.full.norm
8f39a1b15b16e86f09bedeaa9d2f811cd6329012a71139e93babd5199f468938 equiv/same-work-20261008-02.fuzz.select.norm
ad091a0d49a9d3a2b6206873ab5c9639d62e76cb5e65006aa730e3444e0a24d7 equiv/same-work-20261008-02.prod.expected.norm
ad091a0d49a9d3a2b6206873ab5c9639d62e76cb5e65006aa730e3444e0a24d7 equiv/same-work-20261008-02.prod.full.norm
ad091a0d49a9d3a2b6206873ab5c9639d62e76cb5e65006aa730e3444e0a24d7 equiv/same-work-20261008-02.prod.select.norm
3a7a7cc495e2d4853f5c85b59309ccee637ae5f9d446b806c577ca904f54785f equiv/same-work-20261008-03.fuzz.expected.norm
3a7a7cc495e2d4853f5c85b59309ccee637ae5f9d446b806c577ca904f54785f equiv/same-work-20261008-03.fuzz.full.norm
3a7a7cc495e2d4853f5c85b59309ccee637ae5f9d446b806c577ca904f54785f equiv/same-work-20261008-03.fuzz.select.norm
2aa97d5d63e5e01cb9fafc2d2508ad5df1c576467dfc46cb72b714fea35c93a5 equiv/same-work-20261008-03.prod.expected.norm
2aa97d5d63e5e01cb9fafc2d2508ad5df1c576467dfc46cb72b714fea35c93a5 equiv/same-work-20261008-03.prod.full.norm
2aa97d5d63e5e01cb9fafc2d2508ad5df1c576467dfc46cb72b714fea35c93a5 equiv/same-work-20261008-03.prod.select.norm
52a0f79483ce948b83c4bcde39ce37b89fdf04091826e14f56260717e0d8fce6 equiv-prod-v1/same-work-20261008-01.prod.expected.norm
2f7f1de48a50650a50daf8835c6c86eba981a155b28cc4bc5a5990f74ea56a12 equiv-prod-v1/same-work-20261008-01.prod.full.norm
2f7f1de48a50650a50daf8835c6c86eba981a155b28cc4bc5a5990f74ea56a12 equiv-prod-v1/same-work-20261008-01.prod.select.norm
db89fbfc3a7a7b595555e9d150fa82b93ef2ecc1f0944f69f759b3ebd39cc685 equiv-prod-v1/same-work-20261008-02.prod.expected.norm
f8a898c426596d8a482139309d6d9a85493ff2d1da8ef099406d97db168bcc9f equiv-prod-v1/same-work-20261008-02.prod.full.norm
f8a898c426596d8a482139309d6d9a85493ff2d1da8ef099406d97db168bcc9f equiv-prod-v1/same-work-20261008-02.prod.select.norm
8e9c9955eda5f46377ffb0afd5e0558760019885d725d2547c1232c6fc939362 equiv-prod-v1/same-work-20261008-03.prod.expected.norm
40778c06a13e40b329f17d9899bf8bcded6e2597f22670e6925d5d7aa5adcc75 equiv-prod-v1/same-work-20261008-03.prod.full.norm
40778c06a13e40b329f17d9899bf8bcded6e2597f22670e6925d5d7aa5adcc75 equiv-prod-v1/same-work-20261008-03.prod.select.norm

4
docs/analysis/class-v6/1p5x/x2/evidence/pairs/kernel-bench.log Normal file
View file

@ -0,0 +1,4 @@
info igneum-worker-cuda 1.0 (4 October 2026): device 0 NVIDIA_GeForce_RTX_4090 (sm_89, 128 SMs), driver 13.2 from libcuda.so.1, NVRTC 12.8 from libnvrtc.so.12, target sm_89 (the device's architecture, listed by NVRTC)
pack /root/x2/out/pairs/pack-chain-seed-d20730 on NVIDIA_GeForce_RTX_4090: nvrtc 2574 cache 3 dataset 28 hot 0 check 1144 race 0 ms variant base class v6 (state leaves 93, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 448274a57f508cbc; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes)
warm-up dispatch (base 0): 478.32 ms; 250 timed dispatches of 16777216 nonces: mean 478.57 ms
RESULT pack=/root/x2/out/pairs/pack-chain-seed-d20730 class=mx8-erad810f22d+sh256x27+state+reg64c+fold+rw device=NVIDIA_GeForce_RTX_4090 arch=sm_89 regs=88 blocks_per_sm=20 warps=0 resident=2560 arena_mib=0 hot_mib=0 hot_slots=0 hot_fill_ms=0.00 nonces=16777216 batches=250 check=PASS fingerprint=01f51b9d4805e5e6 mhs=35.057 loads=256 bytes=1024 scratch_ops=0 time=wall

1
docs/analysis/class-v6/1p5x/x2/evidence/pairs/kernel-bench.row Normal file
View file

@ -0,0 +1 @@
kernel_mhs=35.057 loaded_mean_w=289.8 samples=120 nj_per_hash=8267.3 fingerprint=01f51b9d4805e5e6

129
docs/analysis/class-v6/1p5x/x2/evidence/pairs/kernel-bench.smi Normal file
View file

@ -0,0 +1,129 @@
2026/10/09 09:11:14.106, 46.06, 210, 40
2026/10/09 09:11:15.112, 39.67, 210, 40
2026/10/09 09:11:16.112, 36.08, 210, 40
2026/10/09 09:11:17.113, 50.73, 2520, 40
2026/10/09 09:11:18.113, 86.22, 2520, 40
2026/10/09 09:11:19.113, 85.68, 2520, 40
2026/10/09 09:11:20.113, 89.37, 2520, 39
2026/10/09 09:11:21.114, 165.06, 2775, 41
2026/10/09 09:11:22.114, 265.88, 2760, 46
2026/10/09 09:11:23.114, 288.06, 2760, 47
2026/10/09 09:11:24.114, 288.38, 2760, 47
2026/10/09 09:11:25.114, 288.50, 2760, 47
2026/10/09 09:11:26.114, 288.97, 2760, 47
2026/10/09 09:11:27.115, 288.99, 2760, 47
2026/10/09 09:11:28.115, 288.36, 2760, 48
2026/10/09 09:11:29.115, 288.95, 2760, 48
2026/10/09 09:11:30.115, 289.15, 2760, 48
2026/10/09 09:11:31.116, 288.99, 2760, 48
2026/10/09 09:11:32.116, 288.85, 2760, 48
2026/10/09 09:11:33.116, 289.21, 2760, 48
2026/10/09 09:11:34.116, 289.61, 2760, 48
2026/10/09 09:11:35.117, 289.60, 2760, 48
2026/10/09 09:11:36.117, 289.54, 2760, 49
2026/10/09 09:11:37.117, 289.36, 2760, 49
2026/10/09 09:11:38.117, 289.06, 2760, 49
2026/10/09 09:11:39.118, 289.00, 2760, 49
2026/10/09 09:11:40.118, 288.92, 2760, 49
2026/10/09 09:11:41.118, 289.17, 2760, 49
2026/10/09 09:11:42.118, 289.25, 2760, 49
2026/10/09 09:11:43.118, 289.38, 2760, 49
2026/10/09 09:11:44.119, 289.35, 2760, 49
2026/10/09 09:11:45.119, 289.08, 2760, 49
2026/10/09 09:11:46.119, 289.00, 2760, 49
2026/10/09 09:11:47.119, 289.10, 2760, 49
2026/10/09 09:11:48.119, 289.25, 2760, 49
2026/10/09 09:11:49.120, 289.37, 2760, 50
2026/10/09 09:11:50.120, 289.52, 2760, 50
2026/10/09 09:11:51.120, 289.84, 2760, 50
2026/10/09 09:11:52.120, 289.88, 2760, 50
2026/10/09 09:11:53.120, 290.05, 2760, 50
2026/10/09 09:11:54.121, 289.83, 2760, 50
2026/10/09 09:11:55.121, 289.86, 2760, 50
2026/10/09 09:11:56.121, 290.22, 2760, 50
2026/10/09 09:11:57.121, 290.16, 2760, 50
2026/10/09 09:11:58.122, 289.94, 2760, 50
2026/10/09 09:11:59.122, 290.63, 2760, 50
2026/10/09 09:12:00.122, 290.45, 2760, 50
2026/10/09 09:12:01.122, 290.58, 2760, 50
2026/10/09 09:12:02.122, 291.19, 2760, 50
2026/10/09 09:12:03.123, 291.22, 2760, 50
2026/10/09 09:12:04.123, 291.30, 2760, 50
2026/10/09 09:12:05.123, 291.62, 2760, 50
2026/10/09 09:12:06.123, 291.70, 2760, 50
2026/10/09 09:12:07.124, 291.68, 2760, 51
2026/10/09 09:12:08.124, 291.31, 2760, 51
2026/10/09 09:12:09.124, 291.66, 2760, 51
2026/10/09 09:12:10.124, 291.41, 2760, 51
2026/10/09 09:12:11.125, 290.95, 2760, 51
2026/10/09 09:12:12.125, 290.96, 2760, 51
2026/10/09 09:12:13.125, 290.98, 2760, 51
2026/10/09 09:12:14.125, 291.01, 2760, 51
2026/10/09 09:12:15.126, 291.61, 2760, 51
2026/10/09 09:12:16.126, 291.45, 2760, 51
2026/10/09 09:12:17.126, 291.76, 2760, 51
2026/10/09 09:12:18.126, 291.73, 2760, 51
2026/10/09 09:12:19.127, 291.65, 2760, 51
2026/10/09 09:12:20.127, 291.78, 2760, 51
2026/10/09 09:12:21.127, 291.79, 2760, 51
2026/10/09 09:12:22.128, 291.68, 2760, 51
2026/10/09 09:12:23.128, 291.47, 2760, 51
2026/10/09 09:12:24.128, 291.69, 2760, 51
2026/10/09 09:12:25.128, 291.60, 2760, 51
2026/10/09 09:12:26.129, 291.90, 2760, 51
2026/10/09 09:12:27.129, 291.80, 2760, 51
2026/10/09 09:12:28.129, 291.42, 2760, 51
2026/10/09 09:12:29.129, 291.95, 2760, 51
2026/10/09 09:12:30.129, 291.47, 2760, 51
2026/10/09 09:12:31.130, 291.26, 2760, 51
2026/10/09 09:12:32.130, 291.78, 2760, 51
2026/10/09 09:12:33.130, 291.73, 2760, 51
2026/10/09 09:12:34.130, 291.73, 2760, 51
2026/10/09 09:12:35.131, 292.38, 2760, 51
2026/10/09 09:12:36.131, 292.33, 2760, 51
2026/10/09 09:12:37.131, 292.52, 2760, 51
2026/10/09 09:12:38.131, 292.28, 2760, 52
2026/10/09 09:12:39.132, 292.36, 2760, 52
2026/10/09 09:12:40.132, 292.54, 2760, 52
2026/10/09 09:12:41.132, 292.07, 2760, 52
2026/10/09 09:12:42.132, 291.84, 2760, 52
2026/10/09 09:12:43.132, 291.90, 2760, 52
2026/10/09 09:12:44.133, 292.01, 2760, 52
2026/10/09 09:12:45.133, 292.71, 2760, 52
2026/10/09 09:12:46.133, 291.98, 2760, 52
2026/10/09 09:12:47.133, 292.16, 2760, 52
2026/10/09 09:12:48.133, 292.44, 2760, 52
2026/10/09 09:12:49.134, 292.08, 2760, 52
2026/10/09 09:12:50.134, 292.25, 2760, 52
2026/10/09 09:12:51.134, 291.91, 2760, 52
2026/10/09 09:12:52.134, 291.98, 2760, 52
2026/10/09 09:12:53.134, 292.34, 2760, 52
2026/10/09 09:12:54.135, 292.46, 2760, 52
2026/10/09 09:12:55.135, 292.54, 2760, 52
2026/10/09 09:12:56.135, 292.18, 2760, 52
2026/10/09 09:12:57.135, 292.31, 2760, 52
2026/10/09 09:12:58.136, 292.52, 2760, 52
2026/10/09 09:12:59.136, 292.52, 2760, 52
2026/10/09 09:13:00.136, 292.69, 2760, 52
2026/10/09 09:13:01.136, 292.01, 2760, 52
2026/10/09 09:13:02.137, 292.35, 2760, 52
2026/10/09 09:13:03.137, 292.79, 2760, 52
2026/10/09 09:13:04.137, 292.60, 2760, 52
2026/10/09 09:13:05.137, 292.39, 2760, 52
2026/10/09 09:13:06.137, 291.99, 2760, 52
2026/10/09 09:13:07.138, 291.90, 2760, 52
2026/10/09 09:13:08.138, 292.00, 2760, 52
2026/10/09 09:13:09.138, 292.40, 2760, 52
2026/10/09 09:13:10.138, 292.42, 2760, 52
2026/10/09 09:13:11.138, 292.40, 2760, 52
2026/10/09 09:13:12.139, 292.14, 2760, 52
2026/10/09 09:13:13.139, 292.12, 2760, 52
2026/10/09 09:13:14.139, 291.81, 2760, 52
2026/10/09 09:13:15.139, 291.77, 2760, 52
2026/10/09 09:13:16.139, 292.57, 2760, 52
2026/10/09 09:13:17.139, 292.16, 2760, 52
2026/10/09 09:13:18.140, 291.56, 2760, 52
2026/10/09 09:13:19.140, 291.96, 2760, 52
2026/10/09 09:13:20.140, 292.07, 2760, 52
2026/10/09 09:13:21.140, 244.38, 2760, 46
2026/10/09 09:13:22.140, 107.67, 2760, 45

27
docs/analysis/class-v6/1p5x/x2/evidence/pairs/summary.txt Normal file
View file

@ -0,0 +1,27 @@
| run | mode | prod_mhs | steady_mhs | steady_mean_w | nj_per_hash_steady | accepted | over_target | j_per_share_steady | steady_j_per_batch | worker_cpu_s | samples |
|---|---|---|---|---|---|---|---|---|---|---|---|
| x2-full-r1 | full | 30.78 | 31.75 | 268.4 | 8453.5 | 4223 | 0 | 8.598 | 35.5 | 0.00 | 135 |
| x2-full-r2 | full | 31.14 | 32.08 | 273.1 | 8513.1 | 4223 | 0 | 8.659 | 35.7 | 0.00 | 134 |
| x2-full-r3 | full | 30.46 | 31.37 | 269.9 | 8603.8 | 4223 | 0 | 8.750 | 36.1 | 0.00 | 136 |
| x2-full-r4 | full | 30.74 | 31.65 | 270.2 | 8537.1 | 4223 | 0 | 8.683 | 35.8 | 0.00 | 136 |
| x2-select-r1 | select | 33.66 | 34.88 | 288.0 | 8256.9 | 4223 | 0 | 8.397 | 34.6 | 0.00 | 123 |
| x2-select-r2 | select | 33.78 | 34.89 | 290.0 | 8311.8 | 4223 | 0 | 8.454 | 34.9 | 0.00 | 123 |
| x2-select-r3 | select | 33.77 | 34.89 | 289.3 | 8291.8 | 4223 | 0 | 8.434 | 34.8 | 0.00 | 123 |
| x2-select-r4 | select | 33.70 | 34.78 | 287.2 | 8257.6 | 4223 | 0 | 8.398 | 34.6 | 0.00 | 124 |
| mode | runs | steady MH/s (sd) | production MH/s | steady W | nJ per hash (sd) | J per accepted share | accepted shares per run |
|---|---|---|---|---|---|---|---|
| full | 4 | 31.71 (0.29) | 30.78 | 270.4 | 8526.9 (62.2) | 8.672 | 4223 |
| select | 4 | 34.86 (0.05) | 33.73 | 288.6 | 8279.5 (27.0) | 8.421 | 4223 |
paired g (select over full, nJ per hash) per pair: 0.9767, 0.9764, 0.9637, 0.9673; mean of means 0.9710
paired rate ratio (select over full) per pair: 1.0986, 1.0876, 1.1122, 1.0989; mean of means 1.0993
kernel timer: 35.057 MH/s, 289.8 W loaded, 8267.3 nJ per hash, fingerprint 01f51b9d4805e5e6
g against the kernel timer (select serving over --bench): 1.0015; full serving over --bench: 1.0314
| r0 (the brief's current ratio) | r_new with g = select/full | r_new with g = select/kernel timer |
|---|---|---|
| 1.6 | 1.554 | 1.602 |
| 2.0 | 1.942 | 2.003 |
| 2.4 | 2.330 | 2.404 |
(a = 1: the specialist side is not touched by a host-side result path; the identity r_new = r0 * g / a, the brief.)

3
docs/analysis/class-v6/1p5x/x2/evidence/pairs/x2-full-r1.row Normal file
View file

@ -0,0 +1,3 @@
card=NVIDIA_GeForce_RTX_4090 mode=full jobs=1024 batch_log2=22 hashes=4294967296 found=4223 setup_s=4.238 prod_wall_s=139.523 prod_mhs=30.78 prod_mean_w=261.1 prod_j_per_batch=35.6 prod_mh_per_w=0.1179 samples=140 steady_wall_s=135.285 steady_mhs=31.75 steady_mean_w=268.4 steady_j_per_batch=35.5 steady_mh_per_w=0.1183 samples=135 accepted=4223 over_target=0 nj_per_hash_steady=8453.5 j_per_share_steady=8.598 smi_ms=1000 worker_cpu_s=0.00 worker_sha=4152c960d1421007
ready: ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/af89be5ddbadb6f6b4aee28ac8f249713be5d4c12621e3cea7f83ceada3c66b3/day/69676e65756d2d6461792ffa50000000000000 dataset-log2 28 batch 4194304 regs 88 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback full worker 1.0 (4 October 2026)
transfers: transfers per chunk: down 33554432 B (full, 0 full fallbacks); mean per chunk: kernel+sync 119.79 ms, select 0.000 ms, read-back 6.433 ms, scan 5.746 ms; 1024 jobs 1024 chunks

142
docs/analysis/class-v6/1p5x/x2/evidence/pairs/x2-full-r1.smi Normal file
View file

@ -0,0 +1,142 @@
2026/10/09 08:52:20.540, 30.54, 210, 25
2026/10/09 08:52:21.547, 30.47, 210, 25
2026/10/09 08:52:22.547, 45.67, 2520, 26
2026/10/09 08:52:23.547, 83.23, 2520, 26
2026/10/09 08:52:24.547, 83.04, 2520, 27
2026/10/09 08:52:25.547, 87.12, 2520, 27
2026/10/09 08:52:26.548, 148.98, 2790, 32
2026/10/09 08:52:27.548, 266.91, 2790, 34
2026/10/09 08:52:28.548, 268.11, 2790, 34
2026/10/09 08:52:29.548, 269.74, 2790, 35
2026/10/09 08:52:30.548, 269.69, 2790, 35
2026/10/09 08:52:31.549, 267.72, 2775, 35
2026/10/09 08:52:32.550, 267.85, 2775, 36
2026/10/09 08:52:33.550, 268.28, 2775, 36
2026/10/09 08:52:34.551, 269.81, 2775, 37
2026/10/09 08:52:35.551, 270.04, 2775, 37
2026/10/09 08:52:36.551, 269.30, 2775, 36
2026/10/09 08:52:37.551, 270.31, 2775, 38
2026/10/09 08:52:38.551, 269.85, 2775, 38
2026/10/09 08:52:39.552, 270.02, 2775, 38
2026/10/09 08:52:40.552, 270.30, 2775, 38
2026/10/09 08:52:41.552, 268.29, 2775, 39
2026/10/09 08:52:42.552, 271.12, 2775, 39
2026/10/09 08:52:43.553, 270.18, 2775, 39
2026/10/09 08:52:44.553, 271.80, 2775, 39
2026/10/09 08:52:45.553, 271.85, 2775, 39
2026/10/09 08:52:46.553, 269.38, 2775, 39
2026/10/09 08:52:47.553, 265.64, 2775, 40
2026/10/09 08:52:48.554, 272.40, 2775, 40
2026/10/09 08:52:49.554, 272.97, 2775, 41
2026/10/09 08:52:50.554, 271.62, 2775, 41
2026/10/09 08:52:51.554, 272.04, 2775, 41
2026/10/09 08:52:52.554, 270.56, 2775, 41
2026/10/09 08:52:53.554, 272.13, 2775, 41
2026/10/09 08:52:54.555, 272.28, 2775, 41
2026/10/09 08:52:55.555, 270.70, 2775, 42
2026/10/09 08:52:56.556, 272.18, 2775, 42
2026/10/09 08:52:57.556, 272.93, 2775, 41
2026/10/09 08:52:58.556, 273.05, 2775, 41
2026/10/09 08:52:59.556, 272.07, 2775, 42
2026/10/09 08:53:00.556, 267.60, 2775, 42
2026/10/09 08:53:01.557, 267.03, 2775, 43
2026/10/09 08:53:02.557, 264.08, 2775, 42
2026/10/09 08:53:03.557, 266.77, 2775, 43
2026/10/09 08:53:04.557, 265.22, 2775, 43
2026/10/09 08:53:05.558, 266.39, 2775, 43
2026/10/09 08:53:06.558, 269.08, 2775, 43
2026/10/09 08:53:07.558, 267.74, 2775, 42
2026/10/09 08:53:08.558, 265.73, 2775, 43
2026/10/09 08:53:09.558, 265.49, 2775, 44
2026/10/09 08:53:10.559, 268.00, 2775, 44
2026/10/09 08:53:11.559, 266.07, 2775, 43
2026/10/09 08:53:12.559, 262.55, 2775, 44
2026/10/09 08:53:13.559, 267.61, 2775, 44
2026/10/09 08:53:14.560, 265.51, 2775, 44
2026/10/09 08:53:15.560, 266.86, 2775, 44
2026/10/09 08:53:16.560, 267.00, 2775, 44
2026/10/09 08:53:17.560, 268.22, 2775, 44
2026/10/09 08:53:18.560, 267.46, 2775, 44
2026/10/09 08:53:19.561, 267.45, 2775, 45
2026/10/09 08:53:20.561, 268.42, 2775, 44
2026/10/09 08:53:21.561, 267.91, 2775, 45
2026/10/09 08:53:22.561, 268.11, 2775, 44
2026/10/09 08:53:23.561, 269.76, 2775, 45
2026/10/09 08:53:24.562, 269.96, 2775, 45
2026/10/09 08:53:25.562, 269.33, 2775, 45
2026/10/09 08:53:26.562, 269.21, 2775, 45
2026/10/09 08:53:27.562, 268.87, 2775, 45
2026/10/09 08:53:28.562, 268.83, 2775, 45
2026/10/09 08:53:29.562, 268.79, 2775, 46
2026/10/09 08:53:30.563, 268.98, 2775, 45
2026/10/09 08:53:31.563, 268.23, 2760, 46
2026/10/09 08:53:32.563, 270.23, 2760, 44
2026/10/09 08:53:33.563, 267.15, 2760, 46
2026/10/09 08:53:34.563, 269.11, 2760, 44
2026/10/09 08:53:35.564, 269.95, 2760, 46
2026/10/09 08:53:36.564, 269.08, 2760, 44
2026/10/09 08:53:37.564, 270.01, 2760, 46
2026/10/09 08:53:38.564, 268.91, 2760, 44
2026/10/09 08:53:39.564, 267.24, 2760, 46
2026/10/09 08:53:40.564, 271.86, 2760, 45
2026/10/09 08:53:41.565, 270.68, 2760, 46
2026/10/09 08:53:42.565, 269.88, 2760, 44
2026/10/09 08:53:43.565, 265.21, 2760, 46
2026/10/09 08:53:44.565, 270.68, 2760, 47
2026/10/09 08:53:45.565, 269.00, 2760, 47
2026/10/09 08:53:46.565, 271.45, 2760, 47
2026/10/09 08:53:47.565, 268.02, 2760, 47
2026/10/09 08:53:48.565, 270.02, 2760, 47
2026/10/09 08:53:49.565, 267.69, 2760, 47
2026/10/09 08:53:50.566, 269.66, 2760, 47
2026/10/09 08:53:51.566, 267.02, 2760, 47
2026/10/09 08:53:52.566, 271.11, 2760, 47
2026/10/09 08:53:53.566, 267.69, 2760, 47
2026/10/09 08:53:54.566, 267.08, 2760, 47
2026/10/09 08:53:55.566, 266.10, 2760, 47
2026/10/09 08:53:56.567, 274.16, 2760, 46
2026/10/09 08:53:57.567, 267.51, 2760, 47
2026/10/09 08:53:58.567, 271.89, 2760, 45
2026/10/09 08:53:59.567, 269.58, 2760, 47
2026/10/09 08:54:00.567, 268.92, 2760, 46
2026/10/09 08:54:01.567, 267.82, 2760, 47
2026/10/09 08:54:02.568, 270.53, 2760, 46
2026/10/09 08:54:03.568, 270.64, 2760, 48
2026/10/09 08:54:04.568, 269.70, 2760, 47
2026/10/09 08:54:05.568, 269.34, 2760, 48
2026/10/09 08:54:06.568, 269.92, 2760, 46
2026/10/09 08:54:07.569, 269.09, 2760, 48
2026/10/09 08:54:08.569, 270.27, 2760, 46
2026/10/09 08:54:09.569, 268.85, 2760, 48
2026/10/09 08:54:10.569, 270.02, 2760, 47
2026/10/09 08:54:11.569, 268.22, 2760, 48
2026/10/09 08:54:12.570, 269.71, 2760, 48
2026/10/09 08:54:13.570, 267.21, 2760, 48
2026/10/09 08:54:14.570, 269.69, 2760, 48
2026/10/09 08:54:15.570, 267.52, 2760, 48
2026/10/09 08:54:16.570, 270.91, 2760, 48
2026/10/09 08:54:17.571, 268.89, 2760, 48
2026/10/09 08:54:18.571, 271.63, 2760, 48
2026/10/09 08:54:19.571, 269.82, 2760, 48
2026/10/09 08:54:20.571, 272.63, 2760, 48
2026/10/09 08:54:21.572, 268.91, 2760, 48
2026/10/09 08:54:22.572, 271.44, 2760, 48
2026/10/09 08:54:23.572, 269.83, 2760, 48
2026/10/09 08:54:24.572, 271.77, 2760, 48
2026/10/09 08:54:25.572, 267.96, 2760, 48
2026/10/09 08:54:26.572, 270.92, 2760, 49
2026/10/09 08:54:27.573, 268.28, 2760, 48
2026/10/09 08:54:28.573, 272.15, 2760, 49
2026/10/09 08:54:29.573, 269.73, 2760, 48
2026/10/09 08:54:30.573, 270.32, 2760, 49
2026/10/09 08:54:31.574, 268.03, 2760, 48
2026/10/09 08:54:32.574, 271.94, 2760, 49
2026/10/09 08:54:33.574, 269.47, 2760, 48
2026/10/09 08:54:34.574, 271.02, 2760, 49
2026/10/09 08:54:35.574, 268.35, 2760, 48
2026/10/09 08:54:36.574, 271.08, 2760, 49
2026/10/09 08:54:37.575, 268.37, 2760, 49
2026/10/09 08:54:38.575, 270.39, 2760, 49
2026/10/09 08:54:39.575, 269.01, 2760, 48
2026/10/09 08:54:40.575, 270.82, 2760, 49
2026/10/09 08:54:41.575, 251.37, 2760, 44

5257
docs/analysis/class-v6/1p5x/x2/evidence/pairs/x2-full-r1.worker.log Normal file
View file

File diff suppressed because it is too large Load diff

3
docs/analysis/class-v6/1p5x/x2/evidence/pairs/x2-full-r2.row Normal file
View file

@ -0,0 +1,3 @@
card=NVIDIA_GeForce_RTX_4090 mode=full jobs=1024 batch_log2=22 hashes=4294967296 found=4223 setup_s=4.005 prod_wall_s=137.908 prod_mhs=31.14 prod_mean_w=267.3 prod_j_per_batch=36.0 prod_mh_per_w=0.1165 samples=138 steady_wall_s=133.903 steady_mhs=32.08 steady_mean_w=273.1 steady_j_per_batch=35.7 steady_mh_per_w=0.1175 samples=134 accepted=4223 over_target=0 nj_per_hash_steady=8513.1 j_per_share_steady=8.659 smi_ms=1000 worker_cpu_s=0.00 worker_sha=4152c960d1421007
ready: ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/af89be5ddbadb6f6b4aee28ac8f249713be5d4c12621e3cea7f83ceada3c66b3/day/69676e65756d2d6461792ffa50000000000000 dataset-log2 28 batch 4194304 regs 88 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback full worker 1.0 (4 October 2026)
transfers: transfers per chunk: down 33554432 B (full, 0 full fallbacks); mean per chunk: kernel+sync 119.81 ms, select 0.000 ms, read-back 5.661 ms, scan 5.148 ms; 1024 jobs 1024 chunks

140
docs/analysis/class-v6/1p5x/x2/evidence/pairs/x2-full-r2.smi Normal file
View file

@ -0,0 +1,140 @@
2026/10/09 08:59:19.764, 43.64, 210, 41
2026/10/09 08:59:20.770, 37.96, 210, 41
2026/10/09 08:59:21.770, 73.96, 2520, 41
2026/10/09 08:59:22.770, 87.36, 2520, 41
2026/10/09 08:59:23.771, 88.89, 2520, 41
2026/10/09 08:59:24.771, 87.73, 2520, 40
2026/10/09 08:59:25.771, 250.12, 2760, 46
2026/10/09 08:59:26.771, 266.11, 2760, 47
2026/10/09 08:59:27.772, 269.29, 2760, 47
2026/10/09 08:59:28.772, 267.14, 2760, 48
2026/10/09 08:59:29.772, 268.56, 2760, 46
2026/10/09 08:59:30.772, 269.14, 2760, 48
2026/10/09 08:59:31.773, 274.80, 2760, 48
2026/10/09 08:59:32.773, 275.35, 2760, 48
2026/10/09 08:59:33.773, 275.62, 2760, 48
2026/10/09 08:59:34.773, 276.48, 2760, 48
2026/10/09 08:59:35.773, 277.81, 2760, 48
2026/10/09 08:59:36.773, 276.80, 2760, 48
2026/10/09 08:59:37.774, 276.76, 2760, 49
2026/10/09 08:59:38.774, 274.29, 2760, 49
2026/10/09 08:59:39.774, 277.00, 2760, 48
2026/10/09 08:59:40.774, 278.54, 2760, 49
2026/10/09 08:59:41.774, 275.21, 2760, 48
2026/10/09 08:59:42.775, 273.17, 2760, 49
2026/10/09 08:59:43.775, 278.83, 2760, 49
2026/10/09 08:59:44.775, 277.56, 2760, 49
2026/10/09 08:59:45.775, 276.35, 2760, 48
2026/10/09 08:59:46.775, 276.63, 2760, 49
2026/10/09 08:59:47.776, 275.16, 2760, 49
2026/10/09 08:59:48.776, 276.64, 2760, 49
2026/10/09 08:59:49.776, 277.37, 2760, 49
2026/10/09 08:59:50.776, 276.12, 2760, 49
2026/10/09 08:59:51.776, 274.44, 2760, 49
2026/10/09 08:59:52.776, 278.67, 2760, 49
2026/10/09 08:59:53.776, 278.99, 2760, 49
2026/10/09 08:59:54.777, 277.74, 2760, 49
2026/10/09 08:59:55.777, 276.41, 2760, 50
2026/10/09 08:59:56.777, 277.59, 2760, 49
2026/10/09 08:59:57.777, 278.42, 2760, 49
2026/10/09 08:59:58.777, 278.40, 2760, 49
2026/10/09 08:59:59.778, 277.53, 2760, 49
2026/10/09 09:00:00.778, 278.25, 2760, 50
2026/10/09 09:00:01.778, 277.84, 2760, 50
2026/10/09 09:00:02.778, 279.25, 2760, 50
2026/10/09 09:00:03.779, 279.76, 2760, 50
2026/10/09 09:00:04.779, 278.89, 2760, 49
2026/10/09 09:00:05.779, 276.09, 2760, 50
2026/10/09 09:00:06.779, 276.98, 2760, 50
2026/10/09 09:00:07.779, 279.90, 2760, 50
2026/10/09 09:00:08.780, 279.15, 2760, 50
2026/10/09 09:00:09.780, 271.14, 2760, 50
2026/10/09 09:00:10.780, 271.12, 2760, 49
2026/10/09 09:00:11.780, 271.34, 2760, 50
2026/10/09 09:00:12.780, 271.80, 2760, 49
2026/10/09 09:00:13.781, 266.54, 2760, 50
2026/10/09 09:00:14.781, 271.71, 2760, 48
2026/10/09 09:00:15.781, 270.75, 2760, 50
2026/10/09 09:00:16.781, 273.28, 2760, 49
2026/10/09 09:00:17.781, 269.96, 2760, 50
2026/10/09 09:00:18.781, 270.51, 2760, 48
2026/10/09 09:00:19.782, 271.05, 2760, 50
2026/10/09 09:00:20.782, 270.49, 2760, 50
2026/10/09 09:00:21.782, 274.32, 2760, 50
2026/10/09 09:00:22.782, 271.95, 2760, 49
2026/10/09 09:00:23.783, 276.32, 2760, 50
2026/10/09 09:00:24.783, 276.05, 2760, 50
2026/10/09 09:00:25.783, 272.14, 2760, 50
2026/10/09 09:00:26.783, 271.11, 2760, 50
2026/10/09 09:00:27.783, 271.23, 2760, 50
2026/10/09 09:00:28.784, 272.54, 2760, 50
2026/10/09 09:00:29.784, 272.16, 2760, 50
2026/10/09 09:00:30.784, 272.93, 2760, 50
2026/10/09 09:00:31.784, 273.94, 2760, 50
2026/10/09 09:00:32.784, 273.90, 2760, 50
2026/10/09 09:00:33.785, 273.38, 2760, 48
2026/10/09 09:00:34.786, 277.19, 2760, 50
2026/10/09 09:00:35.786, 275.67, 2760, 50
2026/10/09 09:00:36.786, 278.14, 2760, 49
2026/10/09 09:00:37.789, 278.08, 2760, 50
2026/10/09 09:00:38.790, 276.65, 2760, 50
2026/10/09 09:00:39.790, 277.18, 2760, 50
2026/10/09 09:00:40.790, 279.41, 2760, 50
2026/10/09 09:00:41.791, 279.42, 2760, 50
2026/10/09 09:00:42.791, 278.47, 2760, 50
2026/10/09 09:00:43.792, 275.61, 2760, 50
2026/10/09 09:00:44.792, 277.53, 2760, 50
2026/10/09 09:00:45.792, 277.70, 2760, 50
2026/10/09 09:00:46.795, 281.01, 2760, 49
2026/10/09 09:00:47.795, 278.59, 2760, 50
2026/10/09 09:00:48.797, 277.35, 2760, 50
2026/10/09 09:00:49.797, 278.25, 2760, 50
2026/10/09 09:00:50.797, 280.28, 2760, 50
2026/10/09 09:00:51.798, 279.47, 2760, 49
2026/10/09 09:00:52.798, 278.68, 2760, 51
2026/10/09 09:00:53.798, 279.03, 2760, 51
2026/10/09 09:00:54.800, 279.01, 2760, 50
2026/10/09 09:00:55.800, 280.40, 2760, 50
2026/10/09 09:00:56.800, 279.44, 2760, 50
2026/10/09 09:00:57.801, 278.99, 2760, 51
2026/10/09 09:00:58.801, 277.81, 2760, 51
2026/10/09 09:00:59.801, 274.94, 2760, 50
2026/10/09 09:01:00.801, 271.38, 2760, 51
2026/10/09 09:01:01.802, 274.13, 2760, 51
2026/10/09 09:01:02.802, 276.82, 2760, 50
2026/10/09 09:01:03.802, 274.14, 2760, 51
2026/10/09 09:01:04.802, 271.94, 2760, 50
2026/10/09 09:01:05.802, 274.88, 2760, 51
2026/10/09 09:01:06.803, 270.42, 2760, 50
2026/10/09 09:01:07.803, 274.53, 2760, 51
2026/10/09 09:01:08.803, 270.30, 2760, 50
2026/10/09 09:01:09.803, 273.16, 2760, 51
2026/10/09 09:01:10.804, 271.37, 2760, 50
2026/10/09 09:01:11.804, 275.66, 2760, 50
2026/10/09 09:01:12.804, 274.66, 2760, 50
2026/10/09 09:01:13.804, 274.17, 2760, 50
2026/10/09 09:01:14.805, 276.41, 2760, 50
2026/10/09 09:01:15.805, 270.87, 2760, 50
2026/10/09 09:01:16.805, 275.81, 2760, 50
2026/10/09 09:01:17.805, 272.57, 2760, 50
2026/10/09 09:01:18.805, 271.96, 2760, 50
2026/10/09 09:01:19.806, 271.33, 2760, 50
2026/10/09 09:01:20.806, 271.43, 2760, 49
2026/10/09 09:01:21.806, 271.50, 2760, 50
2026/10/09 09:01:22.806, 271.99, 2760, 49
2026/10/09 09:01:23.807, 271.64, 2760, 50
2026/10/09 09:01:24.807, 271.47, 2760, 49
2026/10/09 09:01:25.807, 271.04, 2760, 50
2026/10/09 09:01:26.807, 272.60, 2760, 49
2026/10/09 09:01:27.807, 271.10, 2760, 50
2026/10/09 09:01:28.808, 272.94, 2760, 50
2026/10/09 09:01:29.808, 274.18, 2760, 50
2026/10/09 09:01:30.808, 271.37, 2760, 50
2026/10/09 09:01:31.808, 271.38, 2760, 50
2026/10/09 09:01:32.809, 271.23, 2760, 48
2026/10/09 09:01:33.809, 272.70, 2760, 50
2026/10/09 09:01:34.809, 270.80, 2760, 48
2026/10/09 09:01:35.809, 270.57, 2760, 50
2026/10/09 09:01:36.809, 270.28, 2760, 51
2026/10/09 09:01:37.810, 268.51, 2760, 50
2026/10/09 09:01:38.810, 260.61, 2760, 46

5257
docs/analysis/class-v6/1p5x/x2/evidence/pairs/x2-full-r2.worker.log Normal file
View file

File diff suppressed because it is too large Load diff

3
docs/analysis/class-v6/1p5x/x2/evidence/pairs/x2-full-r3.row Normal file
View file

@ -0,0 +1,3 @@
card=NVIDIA_GeForce_RTX_4090 mode=full jobs=1024 batch_log2=22 hashes=4294967296 found=4223 setup_s=4.100 prod_wall_s=141.006 prod_mhs=30.46 prod_mean_w=262.8 prod_j_per_batch=36.2 prod_mh_per_w=0.1159 samples=141 steady_wall_s=136.906 steady_mhs=31.37 steady_mean_w=269.9 steady_j_per_batch=36.1 steady_mh_per_w=0.1162 samples=136 accepted=4223 over_target=0 nj_per_hash_steady=8603.8 j_per_share_steady=8.750 smi_ms=1000 worker_cpu_s=0.00 worker_sha=4152c960d1421007
ready: ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/af89be5ddbadb6f6b4aee28ac8f249713be5d4c12621e3cea7f83ceada3c66b3/day/69676e65756d2d6461792ffa50000000000000 dataset-log2 28 batch 4194304 regs 88 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback full worker 1.0 (4 October 2026)
transfers: transfers per chunk: down 33554432 B (full, 0 full fallbacks); mean per chunk: kernel+sync 119.80 ms, select 0.000 ms, read-back 8.203 ms, scan 5.549 ms; 1024 jobs 1024 chunks

144
docs/analysis/class-v6/1p5x/x2/evidence/pairs/x2-full-r3.smi Normal file
View file

@ -0,0 +1,144 @@
2026/10/09 09:01:45.776, 40.33, 210, 40
2026/10/09 09:01:46.783, 35.33, 210, 39
2026/10/09 09:01:47.783, 54.14, 2520, 40
2026/10/09 09:01:48.784, 84.70, 2520, 40
2026/10/09 09:01:49.784, 87.41, 2520, 40
2026/10/09 09:01:50.784, 90.21, 2520, 39
2026/10/09 09:01:51.784, 220.66, 2775, 46
2026/10/09 09:01:52.784, 268.13, 2760, 46
2026/10/09 09:01:53.785, 268.84, 2760, 47
2026/10/09 09:01:54.785, 270.12, 2760, 47
2026/10/09 09:01:55.785, 266.54, 2760, 47
2026/10/09 09:01:56.785, 265.58, 2760, 46
2026/10/09 09:01:57.786, 267.10, 2760, 47
2026/10/09 09:01:58.786, 267.89, 2760, 46
2026/10/09 09:01:59.786, 267.24, 2760, 47
2026/10/09 09:02:00.786, 271.65, 2760, 46
2026/10/09 09:02:01.787, 269.01, 2760, 47
2026/10/09 09:02:02.787, 253.90, 2760, 47
2026/10/09 09:02:03.787, 270.11, 2760, 48
2026/10/09 09:02:04.787, 266.46, 2760, 47
2026/10/09 09:02:05.788, 268.82, 2760, 48
2026/10/09 09:02:06.788, 268.61, 2760, 47
2026/10/09 09:02:07.789, 271.86, 2760, 48
2026/10/09 09:02:08.789, 269.28, 2760, 48
2026/10/09 09:02:09.789, 269.18, 2760, 48
2026/10/09 09:02:10.789, 269.39, 2760, 48
2026/10/09 09:02:11.790, 269.63, 2760, 48
2026/10/09 09:02:12.790, 267.38, 2760, 48
2026/10/09 09:02:13.790, 275.61, 2760, 48
2026/10/09 09:02:14.790, 270.16, 2760, 48
2026/10/09 09:02:15.790, 269.68, 2760, 47
2026/10/09 09:02:16.791, 269.80, 2760, 48
2026/10/09 09:02:17.791, 273.93, 2760, 48
2026/10/09 09:02:18.791, 269.80, 2760, 48
2026/10/09 09:02:19.791, 269.18, 2760, 48
2026/10/09 09:02:20.791, 269.68, 2760, 48
2026/10/09 09:02:21.792, 270.30, 2760, 48
2026/10/09 09:02:22.792, 270.76, 2760, 49
2026/10/09 09:02:23.792, 270.15, 2760, 48
2026/10/09 09:02:24.792, 270.67, 2760, 49
2026/10/09 09:02:25.792, 270.31, 2760, 48
2026/10/09 09:02:26.793, 269.45, 2760, 49
2026/10/09 09:02:27.793, 268.52, 2760, 48
2026/10/09 09:02:28.793, 270.55, 2760, 49
2026/10/09 09:02:29.793, 268.88, 2760, 49
2026/10/09 09:02:30.793, 271.43, 2760, 49
2026/10/09 09:02:31.794, 269.86, 2760, 49
2026/10/09 09:02:32.794, 270.83, 2760, 49
2026/10/09 09:02:33.794, 269.42, 2760, 49
2026/10/09 09:02:34.794, 271.53, 2760, 49
2026/10/09 09:02:35.794, 269.46, 2760, 49
2026/10/09 09:02:36.795, 271.12, 2760, 49
2026/10/09 09:02:37.795, 268.79, 2760, 49
2026/10/09 09:02:38.795, 270.62, 2760, 49
2026/10/09 09:02:39.796, 270.08, 2760, 49
2026/10/09 09:02:40.796, 270.69, 2760, 49
2026/10/09 09:02:41.796, 270.19, 2760, 49
2026/10/09 09:02:42.796, 270.93, 2760, 49
2026/10/09 09:02:43.796, 266.50, 2760, 47
2026/10/09 09:02:44.797, 275.51, 2760, 49
2026/10/09 09:02:45.797, 270.53, 2760, 49
2026/10/09 09:02:46.797, 273.10, 2760, 49
2026/10/09 09:02:47.797, 273.92, 2760, 49
2026/10/09 09:02:48.797, 273.39, 2760, 49
2026/10/09 09:02:49.798, 272.48, 2760, 49
2026/10/09 09:02:50.798, 272.67, 2760, 49
2026/10/09 09:02:51.799, 277.55, 2760, 49
2026/10/09 09:02:52.799, 278.97, 2760, 49
2026/10/09 09:02:53.799, 277.76, 2760, 48
2026/10/09 09:02:54.799, 231.03, 2760, 49
2026/10/09 09:02:55.800, 177.83, 2760, 49
2026/10/09 09:02:56.800, 275.73, 2760, 48
2026/10/09 09:02:57.800, 259.11, 2760, 49
2026/10/09 09:02:58.800, 277.54, 2760, 49
2026/10/09 09:02:59.800, 275.70, 2760, 49
2026/10/09 09:03:00.801, 271.41, 2760, 49
2026/10/09 09:03:01.801, 271.92, 2760, 48
2026/10/09 09:03:02.801, 277.60, 2760, 49
2026/10/09 09:03:03.801, 278.35, 2760, 49
2026/10/09 09:03:04.801, 177.28, 2760, 43
2026/10/09 09:03:05.802, 222.23, 2760, 49
2026/10/09 09:03:06.802, 276.97, 2760, 49
2026/10/09 09:03:07.802, 277.18, 2760, 49
2026/10/09 09:03:08.802, 277.03, 2760, 49
2026/10/09 09:03:09.802, 277.30, 2760, 49
2026/10/09 09:03:10.803, 277.23, 2760, 49
2026/10/09 09:03:11.803, 272.37, 2760, 49
2026/10/09 09:03:12.803, 277.38, 2760, 49
2026/10/09 09:03:13.803, 280.21, 2760, 49
2026/10/09 09:03:14.803, 278.21, 2760, 49
2026/10/09 09:03:15.804, 278.60, 2760, 49
2026/10/09 09:03:16.804, 277.48, 2760, 50
2026/10/09 09:03:17.804, 278.17, 2760, 50
2026/10/09 09:03:18.804, 275.38, 2760, 49
2026/10/09 09:03:19.804, 280.10, 2760, 49
2026/10/09 09:03:20.805, 275.29, 2760, 50
2026/10/09 09:03:21.805, 278.37, 2760, 50
2026/10/09 09:03:22.805, 279.24, 2760, 50
2026/10/09 09:03:23.805, 277.31, 2760, 49
2026/10/09 09:03:24.805, 228.04, 2760, 49
2026/10/09 09:03:25.806, 277.83, 2760, 49
2026/10/09 09:03:26.806, 271.42, 2760, 50
2026/10/09 09:03:27.806, 272.81, 2760, 49
2026/10/09 09:03:28.806, 272.29, 2760, 50
2026/10/09 09:03:29.806, 273.03, 2760, 49
2026/10/09 09:03:30.807, 272.48, 2760, 50
2026/10/09 09:03:31.807, 271.74, 2760, 49
2026/10/09 09:03:32.807, 271.24, 2760, 50
2026/10/09 09:03:33.807, 271.99, 2760, 49
2026/10/09 09:03:34.808, 277.48, 2760, 49
2026/10/09 09:03:35.808, 276.23, 2760, 50
2026/10/09 09:03:36.808, 277.36, 2760, 50
2026/10/09 09:03:37.808, 279.18, 2760, 50
2026/10/09 09:03:38.808, 279.06, 2760, 50
2026/10/09 09:03:39.809, 275.63, 2760, 50
2026/10/09 09:03:40.809, 276.58, 2760, 50
2026/10/09 09:03:41.809, 277.23, 2760, 50
2026/10/09 09:03:42.809, 278.42, 2760, 50
2026/10/09 09:03:43.809, 277.19, 2760, 50
2026/10/09 09:03:44.809, 278.62, 2760, 50
2026/10/09 09:03:45.810, 278.46, 2760, 50
2026/10/09 09:03:46.810, 278.76, 2760, 50
2026/10/09 09:03:47.810, 277.82, 2760, 49
2026/10/09 09:03:48.810, 271.36, 2760, 50
2026/10/09 09:03:49.811, 272.60, 2760, 49
2026/10/09 09:03:50.811, 271.71, 2760, 50
2026/10/09 09:03:51.811, 273.16, 2760, 50
2026/10/09 09:03:52.811, 270.83, 2760, 50
2026/10/09 09:03:53.812, 274.01, 2760, 50
2026/10/09 09:03:54.812, 274.69, 2760, 49
2026/10/09 09:03:55.812, 271.38, 2760, 50
2026/10/09 09:03:56.812, 277.30, 2760, 49
2026/10/09 09:03:57.812, 272.29, 2760, 50
2026/10/09 09:03:58.813, 272.25, 2760, 50
2026/10/09 09:03:59.813, 270.46, 2760, 50
2026/10/09 09:04:00.813, 271.41, 2760, 50
2026/10/09 09:04:01.813, 269.86, 2760, 50
2026/10/09 09:04:02.814, 273.04, 2760, 50
2026/10/09 09:04:03.814, 271.18, 2760, 50
2026/10/09 09:04:04.814, 274.47, 2760, 50
2026/10/09 09:04:05.814, 271.48, 2760, 50
2026/10/09 09:04:06.815, 272.30, 2760, 50
2026/10/09 09:04:07.815, 266.81, 2760, 50
2026/10/09 09:04:08.815, 115.79, 2760, 44

5257
docs/analysis/class-v6/1p5x/x2/evidence/pairs/x2-full-r3.worker.log Normal file
View file

File diff suppressed because it is too large Load diff

3
docs/analysis/class-v6/1p5x/x2/evidence/pairs/x2-full-r4.row Normal file
View file

@ -0,0 +1,3 @@
card=NVIDIA_GeForce_RTX_4090 mode=full jobs=1024 batch_log2=22 hashes=4294967296 found=4223 setup_s=4.010 prod_wall_s=139.724 prod_mhs=30.74 prod_mean_w=264.4 prod_j_per_batch=36.1 prod_mh_per_w=0.1163 samples=140 steady_wall_s=135.714 steady_mhs=31.65 steady_mean_w=270.2 steady_j_per_batch=35.8 steady_mh_per_w=0.1171 samples=136 accepted=4223 over_target=0 nj_per_hash_steady=8537.1 j_per_share_steady=8.683 smi_ms=1000 worker_cpu_s=0.00 worker_sha=4152c960d1421007
ready: ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/af89be5ddbadb6f6b4aee28ac8f249713be5d4c12621e3cea7f83ceada3c66b3/day/69676e65756d2d6461792ffa50000000000000 dataset-log2 28 batch 4194304 regs 88 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback full worker 1.0 (4 October 2026)
transfers: transfers per chunk: down 33554432 B (full, 0 full fallbacks); mean per chunk: kernel+sync 119.79 ms, select 0.000 ms, read-back 6.631 ms, scan 5.962 ms; 1024 jobs 1024 chunks

142
docs/analysis/class-v6/1p5x/x2/evidence/pairs/x2-full-r4.smi Normal file
View file

@ -0,0 +1,142 @@
2026/10/09 09:08:46.093, 48.60, 210, 41
2026/10/09 09:08:47.100, 42.60, 210, 41
2026/10/09 09:08:48.100, 51.20, 2520, 41
2026/10/09 09:08:49.100, 86.16, 2520, 41
2026/10/09 09:08:50.100, 85.30, 2520, 40
2026/10/09 09:08:51.101, 89.17, 2520, 40
2026/10/09 09:08:52.101, 192.27, 2760, 46
2026/10/09 09:08:53.101, 267.11, 2760, 47
2026/10/09 09:08:54.101, 264.91, 2760, 47
2026/10/09 09:08:55.102, 267.42, 2760, 47
2026/10/09 09:08:56.102, 269.58, 2760, 47
2026/10/09 09:08:57.102, 272.19, 2760, 48
2026/10/09 09:08:58.103, 268.53, 2760, 47
2026/10/09 09:08:59.103, 271.72, 2760, 48
2026/10/09 09:09:00.103, 274.45, 2760, 48
2026/10/09 09:09:01.103, 270.29, 2760, 46
2026/10/09 09:09:02.103, 268.87, 2760, 48
2026/10/09 09:09:03.104, 271.18, 2760, 47
2026/10/09 09:09:04.104, 268.76, 2760, 48
2026/10/09 09:09:05.104, 270.62, 2760, 47
2026/10/09 09:09:06.107, 267.92, 2760, 48
2026/10/09 09:09:07.107, 271.64, 2760, 47
2026/10/09 09:09:08.107, 273.08, 2760, 48
2026/10/09 09:09:09.107, 271.49, 2760, 48
2026/10/09 09:09:10.108, 273.38, 2760, 49
2026/10/09 09:09:11.108, 269.46, 2760, 48
2026/10/09 09:09:12.108, 271.59, 2760, 49
2026/10/09 09:09:13.108, 270.44, 2760, 48
2026/10/09 09:09:14.109, 272.31, 2760, 49
2026/10/09 09:09:15.109, 270.35, 2760, 48
2026/10/09 09:09:16.109, 270.74, 2760, 49
2026/10/09 09:09:17.109, 271.14, 2760, 49
2026/10/09 09:09:18.110, 271.72, 2760, 49
2026/10/09 09:09:19.110, 269.78, 2760, 49
2026/10/09 09:09:20.110, 272.89, 2760, 49
2026/10/09 09:09:21.110, 269.07, 2760, 49
2026/10/09 09:09:22.111, 269.73, 2760, 49
2026/10/09 09:09:23.111, 269.46, 2760, 49
2026/10/09 09:09:24.111, 272.11, 2760, 49
2026/10/09 09:09:25.111, 269.24, 2760, 49
2026/10/09 09:09:26.112, 272.83, 2760, 49
2026/10/09 09:09:27.112, 275.31, 2760, 49
2026/10/09 09:09:28.112, 273.48, 2760, 49
2026/10/09 09:09:29.112, 272.01, 2760, 49
2026/10/09 09:09:30.112, 270.89, 2760, 49
2026/10/09 09:09:31.113, 276.31, 2760, 49
2026/10/09 09:09:32.113, 268.75, 2760, 48
2026/10/09 09:09:33.113, 268.81, 2760, 49
2026/10/09 09:09:34.113, 270.31, 2760, 49
2026/10/09 09:09:35.114, 272.23, 2760, 49
2026/10/09 09:09:36.114, 271.67, 2760, 49
2026/10/09 09:09:37.114, 270.68, 2760, 49
2026/10/09 09:09:38.114, 271.38, 2760, 49
2026/10/09 09:09:39.115, 273.43, 2760, 50
2026/10/09 09:09:40.115, 272.61, 2760, 49
2026/10/09 09:09:41.115, 272.01, 2760, 49
2026/10/09 09:09:42.115, 270.87, 2760, 49
2026/10/09 09:09:43.116, 272.57, 2760, 50
2026/10/09 09:09:44.116, 270.46, 2760, 49
2026/10/09 09:09:45.116, 271.03, 2760, 50
2026/10/09 09:09:46.116, 269.90, 2760, 49
2026/10/09 09:09:47.116, 271.55, 2760, 50
2026/10/09 09:09:48.117, 271.51, 2760, 49
2026/10/09 09:09:49.117, 270.86, 2760, 50
2026/10/09 09:09:50.117, 269.92, 2760, 49
2026/10/09 09:09:51.118, 271.20, 2760, 50
2026/10/09 09:09:52.118, 271.03, 2760, 49
2026/10/09 09:09:53.118, 273.02, 2760, 50
2026/10/09 09:09:54.118, 270.91, 2760, 49
2026/10/09 09:09:55.119, 270.29, 2760, 50
2026/10/09 09:09:56.119, 270.75, 2760, 49
2026/10/09 09:09:57.119, 272.93, 2760, 50
2026/10/09 09:09:58.119, 273.64, 2760, 49
2026/10/09 09:09:59.119, 271.74, 2760, 50
2026/10/09 09:10:00.120, 271.17, 2760, 50
2026/10/09 09:10:01.120, 276.41, 2760, 50
2026/10/09 09:10:02.120, 270.75, 2760, 50
2026/10/09 09:10:03.120, 272.01, 2760, 49
2026/10/09 09:10:04.120, 268.45, 2760, 50
2026/10/09 09:10:05.121, 272.12, 2760, 49
2026/10/09 09:10:06.121, 268.86, 2760, 50
2026/10/09 09:10:07.121, 272.10, 2760, 49
2026/10/09 09:10:08.121, 270.48, 2760, 50
2026/10/09 09:10:09.121, 273.45, 2760, 50
2026/10/09 09:10:10.122, 273.44, 2760, 50
2026/10/09 09:10:11.122, 271.70, 2760, 49
2026/10/09 09:10:12.122, 272.21, 2760, 50
2026/10/09 09:10:13.122, 271.68, 2760, 49
2026/10/09 09:10:14.123, 271.94, 2760, 50
2026/10/09 09:10:15.123, 272.64, 2760, 50
2026/10/09 09:10:16.123, 271.32, 2760, 50
2026/10/09 09:10:17.123, 270.38, 2760, 49
2026/10/09 09:10:18.124, 274.93, 2760, 50
2026/10/09 09:10:19.124, 275.93, 2760, 50
2026/10/09 09:10:20.124, 278.30, 2760, 50
2026/10/09 09:10:21.124, 280.29, 2760, 50
2026/10/09 09:10:22.124, 276.02, 2760, 50
2026/10/09 09:10:23.125, 277.16, 2760, 50
2026/10/09 09:10:24.125, 277.49, 2760, 50
2026/10/09 09:10:25.125, 279.40, 2760, 50
2026/10/09 09:10:26.125, 278.73, 2760, 50
2026/10/09 09:10:27.125, 277.20, 2760, 50
2026/10/09 09:10:28.126, 277.27, 2760, 50
2026/10/09 09:10:29.126, 280.17, 2760, 50
2026/10/09 09:10:30.126, 276.92, 2760, 50
2026/10/09 09:10:31.126, 274.90, 2760, 50
2026/10/09 09:10:32.126, 279.08, 2760, 50
2026/10/09 09:10:33.127, 278.64, 2760, 50
2026/10/09 09:10:34.127, 270.81, 2760, 50
2026/10/09 09:10:35.127, 269.88, 2760, 50
2026/10/09 09:10:36.127, 270.61, 2760, 50
2026/10/09 09:10:37.127, 270.97, 2760, 50
2026/10/09 09:10:38.128, 272.65, 2760, 50
2026/10/09 09:10:39.128, 270.78, 2760, 50
2026/10/09 09:10:40.128, 272.94, 2760, 50
2026/10/09 09:10:41.128, 269.43, 2760, 50
2026/10/09 09:10:42.129, 272.37, 2760, 50
2026/10/09 09:10:43.129, 271.11, 2760, 50
2026/10/09 09:10:44.129, 272.56, 2760, 51
2026/10/09 09:10:45.129, 269.55, 2760, 50
2026/10/09 09:10:46.130, 276.54, 2760, 50
2026/10/09 09:10:47.130, 274.40, 2760, 50
2026/10/09 09:10:48.130, 274.00, 2760, 50
2026/10/09 09:10:49.130, 271.61, 2760, 50
2026/10/09 09:10:50.130, 274.79, 2760, 50
2026/10/09 09:10:51.131, 271.59, 2760, 50
2026/10/09 09:10:52.131, 271.37, 2760, 50
2026/10/09 09:10:53.131, 271.76, 2760, 50
2026/10/09 09:10:54.131, 272.44, 2760, 50
2026/10/09 09:10:55.132, 270.05, 2760, 50
2026/10/09 09:10:56.132, 271.91, 2760, 50
2026/10/09 09:10:57.132, 269.70, 2760, 50
2026/10/09 09:10:58.132, 272.08, 2760, 50
2026/10/09 09:10:59.132, 269.64, 2760, 50
2026/10/09 09:11:00.133, 271.48, 2760, 49
2026/10/09 09:11:01.133, 270.46, 2760, 50
2026/10/09 09:11:02.133, 271.80, 2760, 49
2026/10/09 09:11:03.133, 271.23, 2760, 50
2026/10/09 09:11:04.133, 272.28, 2760, 49
2026/10/09 09:11:05.134, 275.50, 2760, 51
2026/10/09 09:11:06.134, 273.62, 2760, 50
2026/10/09 09:11:07.134, 267.10, 2760, 49

5257
docs/analysis/class-v6/1p5x/x2/evidence/pairs/x2-full-r4.worker.log Normal file
View file

File diff suppressed because it is too large Load diff

3
docs/analysis/class-v6/1p5x/x2/evidence/pairs/x2-select-r1.row Normal file
View file

@ -0,0 +1,3 @@
card=NVIDIA_GeForce_RTX_4090 mode=select jobs=1024 batch_log2=22 hashes=4294967296 found=4223 setup_s=4.465 prod_wall_s=127.590 prod_mhs=33.66 prod_mean_w=279.5 prod_j_per_batch=34.8 prod_mh_per_w=0.1204 samples=128 steady_wall_s=123.125 steady_mhs=34.88 steady_mean_w=288.0 steady_j_per_batch=34.6 steady_mh_per_w=0.1211 samples=123 accepted=4223 over_target=0 nj_per_hash_steady=8256.9 j_per_share_steady=8.397 smi_ms=1000 worker_cpu_s=0.00 worker_sha=4152c960d1421007
ready: ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/af89be5ddbadb6f6b4aee28ac8f249713be5d4c12621e3cea7f83ceada3c66b3/day/69676e65756d2d6461792ffa50000000000000 dataset-log2 28 batch 4194304 regs 88 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback select worker 1.0 (4 October 2026)
transfers: transfers per chunk: down 342 B (select, 0 full fallbacks); mean per chunk: kernel+sync 119.76 ms, select 0.161 ms, read-back 0.198 ms, scan 0.008 ms; 1024 jobs 1024 chunks

130
docs/analysis/class-v6/1p5x/x2/evidence/pairs/x2-select-r1.smi Normal file
View file

@ -0,0 +1,130 @@
2026/10/09 08:54:48.491, 42.31, 210, 39
2026/10/09 08:54:49.504, 36.65, 210, 38
2026/10/09 08:54:50.504, 47.61, 2520, 39
2026/10/09 08:54:51.504, 85.52, 2520, 39
2026/10/09 08:54:52.504, 85.55, 2520, 38
2026/10/09 08:54:53.505, 89.41, 2520, 38
2026/10/09 08:54:54.505, 114.60, 2775, 43
2026/10/09 08:54:55.505, 281.13, 2775, 45
2026/10/09 08:54:56.506, 286.29, 2775, 45
2026/10/09 08:54:57.506, 286.60, 2775, 46
2026/10/09 08:54:58.506, 286.12, 2760, 46
2026/10/09 08:54:59.506, 286.28, 2760, 46
2026/10/09 08:55:00.507, 286.65, 2760, 46
2026/10/09 08:55:01.507, 285.75, 2760, 46
2026/10/09 08:55:02.507, 286.23, 2760, 46
2026/10/09 08:55:03.507, 286.75, 2760, 47
2026/10/09 08:55:04.507, 287.24, 2760, 47
2026/10/09 08:55:05.508, 287.42, 2760, 47
2026/10/09 08:55:06.508, 288.01, 2760, 47
2026/10/09 08:55:07.508, 287.94, 2760, 47
2026/10/09 08:55:08.508, 286.65, 2760, 47
2026/10/09 08:55:09.508, 286.43, 2760, 47
2026/10/09 08:55:10.508, 286.93, 2760, 47
2026/10/09 08:55:11.509, 288.09, 2760, 48
2026/10/09 08:55:12.509, 287.40, 2760, 48
2026/10/09 08:55:13.509, 287.36, 2760, 48
2026/10/09 08:55:14.509, 287.72, 2760, 48
2026/10/09 08:55:15.509, 288.10, 2760, 48
2026/10/09 08:55:16.510, 288.64, 2760, 48
2026/10/09 08:55:17.510, 287.92, 2760, 48
2026/10/09 08:55:18.510, 288.29, 2760, 48
2026/10/09 08:55:19.510, 288.70, 2760, 48
2026/10/09 08:55:20.511, 288.62, 2760, 48
2026/10/09 08:55:21.511, 289.16, 2760, 48
2026/10/09 08:55:22.511, 288.77, 2760, 48
2026/10/09 08:55:23.511, 289.04, 2760, 48
2026/10/09 08:55:24.511, 288.48, 2760, 49
2026/10/09 08:55:25.512, 288.80, 2760, 49
2026/10/09 08:55:26.512, 288.63, 2760, 49
2026/10/09 08:55:27.512, 288.83, 2760, 49
2026/10/09 08:55:28.512, 288.81, 2760, 49
2026/10/09 08:55:29.512, 288.97, 2760, 49
2026/10/09 08:55:30.513, 288.92, 2760, 49
2026/10/09 08:55:31.513, 288.83, 2760, 49
2026/10/09 08:55:32.513, 289.32, 2760, 49
2026/10/09 08:55:33.513, 289.20, 2760, 49
2026/10/09 08:55:34.514, 289.13, 2760, 49
2026/10/09 08:55:35.514, 289.06, 2760, 49
2026/10/09 08:55:36.514, 289.03, 2760, 49
2026/10/09 08:55:37.514, 289.23, 2760, 49
2026/10/09 08:55:38.514, 289.83, 2760, 49
2026/10/09 08:55:39.515, 289.76, 2760, 49
2026/10/09 08:55:40.515, 289.39, 2760, 49
2026/10/09 08:55:41.515, 289.59, 2760, 50
2026/10/09 08:55:42.515, 289.90, 2760, 50
2026/10/09 08:55:43.515, 289.95, 2760, 50
2026/10/09 08:55:44.516, 289.68, 2760, 50
2026/10/09 08:55:45.516, 289.62, 2760, 50
2026/10/09 08:55:46.516, 290.47, 2760, 50
2026/10/09 08:55:47.516, 290.23, 2760, 50
2026/10/09 08:55:48.516, 290.11, 2760, 50
2026/10/09 08:55:49.517, 290.14, 2760, 50
2026/10/09 08:55:50.517, 289.96, 2760, 50
2026/10/09 08:55:51.517, 289.99, 2760, 50
2026/10/09 08:55:52.517, 289.92, 2760, 50
2026/10/09 08:55:53.517, 290.20, 2760, 50
2026/10/09 08:55:54.518, 290.35, 2760, 50
2026/10/09 08:55:55.518, 289.38, 2760, 50
2026/10/09 08:55:56.518, 289.65, 2760, 50
2026/10/09 08:55:57.518, 289.79, 2760, 50
2026/10/09 08:55:58.519, 290.11, 2760, 50
2026/10/09 08:55:59.519, 290.22, 2760, 50
2026/10/09 08:56:00.519, 290.59, 2760, 50
2026/10/09 08:56:01.519, 290.39, 2760, 50
2026/10/09 08:56:02.519, 289.80, 2760, 50
2026/10/09 08:56:03.520, 290.12, 2760, 50
2026/10/09 08:56:04.520, 290.18, 2760, 50
2026/10/09 08:56:05.520, 290.65, 2760, 50
2026/10/09 08:56:06.520, 290.63, 2760, 50
2026/10/09 08:56:07.520, 290.66, 2760, 50
2026/10/09 08:56:08.521, 290.36, 2760, 51
2026/10/09 08:56:09.521, 290.08, 2760, 51
2026/10/09 08:56:10.521, 289.06, 2760, 51
2026/10/09 08:56:11.521, 288.95, 2760, 51
2026/10/09 08:56:12.521, 290.25, 2760, 51
2026/10/09 08:56:13.522, 290.21, 2760, 51
2026/10/09 08:56:14.522, 290.04, 2760, 51
2026/10/09 08:56:15.522, 289.89, 2760, 51
2026/10/09 08:56:16.522, 290.12, 2760, 51
2026/10/09 08:56:17.522, 290.35, 2760, 51
2026/10/09 08:56:18.523, 289.20, 2760, 51
2026/10/09 08:56:19.523, 289.27, 2760, 51
2026/10/09 08:56:20.523, 290.30, 2760, 51
2026/10/09 08:56:21.523, 291.06, 2760, 51
2026/10/09 08:56:22.523, 291.29, 2760, 51
2026/10/09 08:56:23.524, 290.65, 2760, 51
2026/10/09 08:56:24.524, 290.69, 2760, 51
2026/10/09 08:56:25.524, 290.79, 2760, 51
2026/10/09 08:56:26.524, 289.90, 2760, 51
2026/10/09 08:56:27.524, 290.38, 2760, 51
2026/10/09 08:56:28.525, 291.15, 2760, 51
2026/10/09 08:56:29.525, 290.25, 2760, 51
2026/10/09 08:56:30.525, 290.68, 2760, 51
2026/10/09 08:56:31.525, 290.37, 2760, 51
2026/10/09 08:56:32.525, 290.05, 2760, 51
2026/10/09 08:56:33.526, 290.67, 2760, 51
2026/10/09 08:56:34.526, 290.63, 2760, 51
2026/10/09 08:56:35.526, 291.04, 2760, 51
2026/10/09 08:56:36.526, 291.01, 2760, 51
2026/10/09 08:56:37.527, 290.50, 2760, 51
2026/10/09 08:56:38.527, 290.62, 2760, 51
2026/10/09 08:56:39.527, 290.85, 2760, 51
2026/10/09 08:56:40.527, 290.54, 2760, 51
2026/10/09 08:56:41.527, 290.46, 2760, 51
2026/10/09 08:56:42.528, 290.48, 2760, 51
2026/10/09 08:56:43.528, 290.63, 2760, 51
2026/10/09 08:56:44.528, 290.48, 2760, 51
2026/10/09 08:56:45.528, 291.11, 2760, 51
2026/10/09 08:56:46.529, 291.37, 2760, 51
2026/10/09 08:56:47.529, 291.30, 2760, 51
2026/10/09 08:56:48.529, 291.26, 2760, 51
2026/10/09 08:56:49.529, 290.99, 2760, 51
2026/10/09 08:56:50.529, 290.97, 2760, 51
2026/10/09 08:56:51.530, 290.49, 2760, 51
2026/10/09 08:56:52.530, 290.57, 2760, 51
2026/10/09 08:56:53.530, 290.66, 2760, 51
2026/10/09 08:56:54.530, 291.03, 2760, 51
2026/10/09 08:56:55.531, 290.34, 2760, 51
2026/10/09 08:56:56.531, 291.44, 2760, 51
2026/10/09 08:56:57.531, 264.60, 2760, 46

5257
docs/analysis/class-v6/1p5x/x2/evidence/pairs/x2-select-r1.worker.log Normal file
View file

File diff suppressed because it is too large Load diff

3
docs/analysis/class-v6/1p5x/x2/evidence/pairs/x2-select-r2.row Normal file
View file

@ -0,0 +1,3 @@
card=NVIDIA_GeForce_RTX_4090 mode=select jobs=1024 batch_log2=22 hashes=4294967296 found=4223 setup_s=4.041 prod_wall_s=127.155 prod_mhs=33.78 prod_mean_w=281.5 prod_j_per_batch=35.0 prod_mh_per_w=0.1200 samples=128 steady_wall_s=123.114 steady_mhs=34.89 steady_mean_w=290.0 steady_j_per_batch=34.9 steady_mh_per_w=0.1203 samples=123 accepted=4223 over_target=0 nj_per_hash_steady=8311.8 j_per_share_steady=8.454 smi_ms=1000 worker_cpu_s=0.00 worker_sha=4152c960d1421007
ready: ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/af89be5ddbadb6f6b4aee28ac8f249713be5d4c12621e3cea7f83ceada3c66b3/day/69676e65756d2d6461792ffa50000000000000 dataset-log2 28 batch 4194304 regs 88 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback select worker 1.0 (4 October 2026)
transfers: transfers per chunk: down 342 B (select, 0 full fallbacks); mean per chunk: kernel+sync 119.76 ms, select 0.142 ms, read-back 0.200 ms, scan 0.008 ms; 1024 jobs 1024 chunks

130
docs/analysis/class-v6/1p5x/x2/evidence/pairs/x2-select-r2.smi Normal file
View file

@ -0,0 +1,130 @@
2026/10/09 08:57:04.291, 43.37, 210, 41
2026/10/09 08:57:05.297, 37.35, 210, 40
2026/10/09 08:57:06.297, 54.78, 2520, 41
2026/10/09 08:57:07.298, 85.57, 2520, 41
2026/10/09 08:57:08.298, 88.14, 2520, 40
2026/10/09 08:57:09.298, 90.74, 2520, 40
2026/10/09 08:57:10.298, 238.95, 2760, 47
2026/10/09 08:57:11.298, 287.73, 2760, 47
2026/10/09 08:57:12.299, 287.92, 2760, 47
2026/10/09 08:57:13.299, 288.22, 2760, 48
2026/10/09 08:57:14.299, 288.50, 2760, 48
2026/10/09 08:57:15.299, 288.53, 2760, 48
2026/10/09 08:57:16.300, 288.30, 2760, 48
2026/10/09 08:57:17.300, 288.55, 2760, 48
2026/10/09 08:57:18.300, 288.53, 2760, 48
2026/10/09 08:57:19.300, 288.08, 2760, 48
2026/10/09 08:57:20.300, 287.88, 2760, 49
2026/10/09 08:57:21.301, 288.58, 2760, 49
2026/10/09 08:57:22.301, 289.67, 2760, 49
2026/10/09 08:57:23.301, 289.51, 2760, 49
2026/10/09 08:57:24.301, 289.55, 2760, 49
2026/10/09 08:57:25.301, 289.42, 2760, 49
2026/10/09 08:57:26.301, 289.40, 2760, 49
2026/10/09 08:57:27.302, 289.13, 2760, 49
2026/10/09 08:57:28.302, 289.39, 2760, 49
2026/10/09 08:57:29.302, 289.71, 2760, 49
2026/10/09 08:57:30.302, 289.77, 2760, 49
2026/10/09 08:57:31.302, 289.40, 2760, 49
2026/10/09 08:57:32.302, 289.86, 2760, 49
2026/10/09 08:57:33.303, 289.52, 2760, 50
2026/10/09 08:57:34.303, 289.32, 2760, 50
2026/10/09 08:57:35.303, 289.63, 2760, 50
2026/10/09 08:57:36.303, 289.65, 2760, 50
2026/10/09 08:57:37.303, 289.27, 2760, 50
2026/10/09 08:57:38.304, 289.15, 2760, 50
2026/10/09 08:57:39.304, 290.16, 2760, 50
2026/10/09 08:57:40.304, 289.35, 2760, 50
2026/10/09 08:57:41.304, 290.01, 2760, 50
2026/10/09 08:57:42.304, 289.80, 2760, 50
2026/10/09 08:57:43.305, 290.64, 2760, 50
2026/10/09 08:57:44.305, 290.66, 2760, 50
2026/10/09 08:57:45.305, 290.47, 2760, 50
2026/10/09 08:57:46.305, 290.35, 2760, 50
2026/10/09 08:57:47.305, 290.53, 2760, 50
2026/10/09 08:57:48.306, 290.34, 2760, 50
2026/10/09 08:57:49.306, 289.97, 2760, 50
2026/10/09 08:57:50.306, 289.87, 2760, 50
2026/10/09 08:57:51.306, 289.86, 2760, 50
2026/10/09 08:57:52.306, 290.66, 2760, 50
2026/10/09 08:57:53.307, 290.26, 2760, 50
2026/10/09 08:57:54.307, 290.52, 2760, 50
2026/10/09 08:57:55.307, 290.12, 2760, 51
2026/10/09 08:57:56.307, 290.63, 2760, 51
2026/10/09 08:57:57.307, 290.57, 2760, 50
2026/10/09 08:57:58.308, 288.60, 2760, 51
2026/10/09 08:57:59.308, 290.34, 2760, 51
2026/10/09 08:58:00.308, 290.98, 2760, 51
2026/10/09 08:58:01.309, 291.02, 2760, 51
2026/10/09 08:58:02.309, 290.73, 2760, 51
2026/10/09 08:58:03.309, 290.92, 2760, 51
2026/10/09 08:58:04.309, 290.75, 2760, 51
2026/10/09 08:58:05.310, 290.38, 2760, 51
2026/10/09 08:58:06.310, 290.19, 2760, 51
2026/10/09 08:58:07.310, 290.68, 2760, 51
2026/10/09 08:58:08.310, 290.88, 2760, 51
2026/10/09 08:58:09.311, 290.61, 2760, 51
2026/10/09 08:58:10.311, 290.69, 2760, 51
2026/10/09 08:58:11.311, 290.67, 2760, 51
2026/10/09 08:58:12.311, 290.86, 2760, 51
2026/10/09 08:58:13.312, 290.58, 2760, 51
2026/10/09 08:58:14.312, 290.67, 2760, 51
2026/10/09 08:58:15.312, 290.85, 2760, 51
2026/10/09 08:58:16.312, 290.23, 2760, 51
2026/10/09 08:58:17.313, 290.90, 2760, 51
2026/10/09 08:58:18.313, 290.91, 2760, 51
2026/10/09 08:58:19.313, 290.74, 2760, 51
2026/10/09 08:58:20.313, 290.69, 2760, 51
2026/10/09 08:58:21.313, 290.86, 2760, 51
2026/10/09 08:58:22.314, 290.93, 2760, 51
2026/10/09 08:58:23.314, 291.05, 2760, 51
2026/10/09 08:58:24.314, 291.16, 2760, 51
2026/10/09 08:58:25.314, 290.88, 2760, 51
2026/10/09 08:58:26.314, 291.01, 2760, 51
2026/10/09 08:58:27.315, 290.85, 2760, 51
2026/10/09 08:58:28.315, 291.21, 2760, 51
2026/10/09 08:58:29.315, 291.38, 2760, 51
2026/10/09 08:58:30.316, 291.03, 2760, 51
2026/10/09 08:58:31.316, 291.05, 2760, 51
2026/10/09 08:58:32.316, 291.02, 2760, 52
2026/10/09 08:58:33.316, 290.71, 2760, 52
2026/10/09 08:58:34.316, 290.43, 2760, 51
2026/10/09 08:58:35.316, 289.99, 2760, 52
2026/10/09 08:58:36.317, 290.21, 2760, 52
2026/10/09 08:58:37.317, 290.23, 2760, 52
2026/10/09 08:58:38.317, 290.24, 2760, 52
2026/10/09 08:58:39.317, 290.71, 2760, 52
2026/10/09 08:58:40.317, 290.89, 2760, 52
2026/10/09 08:58:41.318, 290.68, 2760, 52
2026/10/09 08:58:42.318, 291.02, 2760, 52
2026/10/09 08:58:43.318, 290.83, 2760, 52
2026/10/09 08:58:44.318, 291.05, 2760, 52
2026/10/09 08:58:45.318, 291.52, 2760, 52
2026/10/09 08:58:46.319, 291.56, 2760, 52
2026/10/09 08:58:47.319, 291.40, 2760, 52
2026/10/09 08:58:48.319, 291.61, 2760, 52
2026/10/09 08:58:49.319, 291.38, 2760, 52
2026/10/09 08:58:50.319, 291.18, 2760, 52
2026/10/09 08:58:51.320, 290.96, 2760, 52
2026/10/09 08:58:52.320, 291.10, 2760, 52
2026/10/09 08:58:53.320, 291.06, 2760, 52
2026/10/09 08:58:54.320, 291.56, 2760, 52
2026/10/09 08:58:55.321, 291.08, 2760, 52
2026/10/09 08:58:56.321, 291.49, 2760, 52
2026/10/09 08:58:57.321, 291.82, 2760, 52
2026/10/09 08:58:58.321, 291.87, 2760, 52
2026/10/09 08:58:59.321, 291.57, 2760, 52
2026/10/09 08:59:00.321, 291.37, 2760, 52
2026/10/09 08:59:01.322, 291.48, 2760, 52
2026/10/09 08:59:02.322, 291.90, 2760, 52
2026/10/09 08:59:03.322, 291.97, 2760, 52
2026/10/09 08:59:04.322, 291.98, 2760, 52
2026/10/09 08:59:05.323, 291.81, 2760, 52
2026/10/09 08:59:06.323, 291.60, 2760, 52
2026/10/09 08:59:07.323, 291.63, 2760, 52
2026/10/09 08:59:08.323, 291.86, 2760, 52
2026/10/09 08:59:09.323, 291.78, 2760, 52
2026/10/09 08:59:10.324, 291.52, 2760, 52
2026/10/09 08:59:11.324, 291.64, 2760, 52
2026/10/09 08:59:12.324, 291.52, 2760, 52
2026/10/09 08:59:13.324, 148.19, 2760, 45

5257
docs/analysis/class-v6/1p5x/x2/evidence/pairs/x2-select-r2.worker.log Normal file
View file

File diff suppressed because it is too large Load diff

3
docs/analysis/class-v6/1p5x/x2/evidence/pairs/x2-select-r3.row Normal file
View file

@ -0,0 +1,3 @@
card=NVIDIA_GeForce_RTX_4090 mode=select jobs=1024 batch_log2=22 hashes=4294967296 found=4223 setup_s=4.067 prod_wall_s=127.175 prod_mhs=33.77 prod_mean_w=280.8 prod_j_per_batch=34.9 prod_mh_per_w=0.1203 samples=128 steady_wall_s=123.107 steady_mhs=34.89 steady_mean_w=289.3 steady_j_per_batch=34.8 steady_mh_per_w=0.1206 samples=123 accepted=4223 over_target=0 nj_per_hash_steady=8291.8 j_per_share_steady=8.434 smi_ms=1000 worker_cpu_s=0.00 worker_sha=4152c960d1421007
ready: ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/af89be5ddbadb6f6b4aee28ac8f249713be5d4c12621e3cea7f83ceada3c66b3/day/69676e65756d2d6461792ffa50000000000000 dataset-log2 28 batch 4194304 regs 88 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback select worker 1.0 (4 October 2026)
transfers: transfers per chunk: down 342 B (select, 0 full fallbacks); mean per chunk: kernel+sync 119.76 ms, select 0.147 ms, read-back 0.189 ms, scan 0.009 ms; 1024 jobs 1024 chunks

130
docs/analysis/class-v6/1p5x/x2/evidence/pairs/x2-select-r3.smi Normal file
View file

@ -0,0 +1,130 @@
2026/10/09 09:04:14.899, 48.76, 210, 40
2026/10/09 09:04:15.904, 42.46, 210, 40
2026/10/09 09:04:16.904, 54.87, 2520, 40
2026/10/09 09:04:17.904, 84.82, 2520, 40
2026/10/09 09:04:18.904, 87.76, 2520, 39
2026/10/09 09:04:19.904, 89.72, 2520, 39
2026/10/09 09:04:20.905, 222.88, 2775, 46
2026/10/09 09:04:21.905, 286.65, 2760, 46
2026/10/09 09:04:22.905, 286.91, 2760, 47
2026/10/09 09:04:23.905, 287.32, 2760, 47
2026/10/09 09:04:24.905, 287.44, 2760, 47
2026/10/09 09:04:25.905, 287.60, 2760, 47
2026/10/09 09:04:26.906, 287.79, 2760, 47
2026/10/09 09:04:27.906, 287.96, 2760, 47
2026/10/09 09:04:28.906, 288.30, 2760, 48
2026/10/09 09:04:29.906, 287.69, 2760, 48
2026/10/09 09:04:30.906, 287.48, 2760, 48
2026/10/09 09:04:31.907, 287.43, 2760, 48
2026/10/09 09:04:32.907, 288.22, 2760, 48
2026/10/09 09:04:33.907, 288.42, 2760, 48
2026/10/09 09:04:34.907, 288.41, 2760, 48
2026/10/09 09:04:35.907, 288.15, 2760, 48
2026/10/09 09:04:36.908, 287.89, 2760, 48
2026/10/09 09:04:37.908, 287.44, 2760, 48
2026/10/09 09:04:38.908, 288.52, 2760, 49
2026/10/09 09:04:39.908, 288.28, 2760, 48
2026/10/09 09:04:40.908, 288.32, 2760, 49
2026/10/09 09:04:41.909, 288.69, 2760, 49
2026/10/09 09:04:42.909, 288.70, 2760, 49
2026/10/09 09:04:43.909, 288.60, 2760, 49
2026/10/09 09:04:44.909, 288.77, 2760, 49
2026/10/09 09:04:45.910, 289.64, 2760, 49
2026/10/09 09:04:46.910, 289.48, 2760, 49
2026/10/09 09:04:47.910, 289.59, 2760, 49
2026/10/09 09:04:48.910, 289.46, 2760, 49
2026/10/09 09:04:49.911, 289.35, 2760, 49
2026/10/09 09:04:50.911, 289.64, 2760, 49
2026/10/09 09:04:51.911, 289.61, 2760, 49
2026/10/09 09:04:52.911, 289.65, 2760, 49
2026/10/09 09:04:53.911, 289.59, 2760, 49
2026/10/09 09:04:54.912, 289.69, 2760, 50
2026/10/09 09:04:55.912, 289.47, 2760, 50
2026/10/09 09:04:56.912, 289.22, 2760, 50
2026/10/09 09:04:57.912, 289.80, 2760, 50
2026/10/09 09:04:58.912, 289.83, 2760, 50
2026/10/09 09:04:59.913, 289.71, 2760, 50
2026/10/09 09:05:00.913, 290.00, 2760, 50
2026/10/09 09:05:01.913, 290.13, 2760, 50
2026/10/09 09:05:02.913, 289.75, 2760, 50
2026/10/09 09:05:03.913, 288.27, 2760, 50
2026/10/09 09:05:04.914, 289.55, 2760, 50
2026/10/09 09:05:05.914, 289.89, 2760, 50
2026/10/09 09:05:06.914, 290.35, 2760, 50
2026/10/09 09:05:07.914, 290.35, 2760, 50
2026/10/09 09:05:08.914, 290.32, 2760, 50
2026/10/09 09:05:09.915, 290.39, 2760, 50
2026/10/09 09:05:10.915, 289.82, 2760, 50
2026/10/09 09:05:11.915, 289.72, 2760, 50
2026/10/09 09:05:12.915, 288.72, 2760, 50
2026/10/09 09:05:13.915, 289.57, 2760, 50
2026/10/09 09:05:14.916, 289.82, 2760, 50
2026/10/09 09:05:15.916, 290.28, 2760, 50
2026/10/09 09:05:16.916, 290.31, 2760, 50
2026/10/09 09:05:17.916, 290.17, 2760, 50
2026/10/09 09:05:18.917, 290.23, 2760, 50
2026/10/09 09:05:19.917, 290.01, 2760, 51
2026/10/09 09:05:20.917, 290.30, 2760, 51
2026/10/09 09:05:21.918, 290.05, 2760, 51
2026/10/09 09:05:22.919, 290.46, 2760, 51
2026/10/09 09:05:23.919, 290.11, 2760, 51
2026/10/09 09:05:24.919, 288.57, 2760, 51
2026/10/09 09:05:25.920, 290.10, 2760, 51
2026/10/09 09:05:26.920, 289.63, 2760, 51
2026/10/09 09:05:27.920, 289.60, 2760, 51
2026/10/09 09:05:28.920, 289.39, 2760, 51
2026/10/09 09:05:29.920, 290.67, 2760, 51
2026/10/09 09:05:30.921, 290.54, 2760, 51
2026/10/09 09:05:31.921, 290.12, 2760, 51
2026/10/09 09:05:32.921, 290.22, 2760, 51
2026/10/09 09:05:33.921, 290.17, 2760, 51
2026/10/09 09:05:34.922, 289.95, 2760, 51
2026/10/09 09:05:35.922, 289.98, 2760, 51
2026/10/09 09:05:36.922, 290.30, 2760, 51
2026/10/09 09:05:37.922, 290.22, 2760, 51
2026/10/09 09:05:38.922, 290.47, 2760, 51
2026/10/09 09:05:39.923, 290.92, 2760, 51
2026/10/09 09:05:40.923, 290.81, 2760, 51
2026/10/09 09:05:41.923, 290.47, 2760, 51
2026/10/09 09:05:42.923, 290.46, 2760, 51
2026/10/09 09:05:43.923, 290.53, 2760, 51
2026/10/09 09:05:44.924, 290.53, 2760, 51
2026/10/09 09:05:45.924, 291.22, 2760, 51
2026/10/09 09:05:46.924, 291.36, 2760, 51
2026/10/09 09:05:47.924, 291.02, 2760, 51
2026/10/09 09:05:48.924, 291.05, 2760, 51
2026/10/09 09:05:49.925, 290.76, 2760, 51
2026/10/09 09:05:50.925, 290.75, 2760, 51
2026/10/09 09:05:51.925, 290.26, 2760, 51
2026/10/09 09:05:52.925, 289.95, 2760, 51
2026/10/09 09:05:53.926, 289.96, 2760, 51
2026/10/09 09:05:54.926, 290.65, 2760, 51
2026/10/09 09:05:55.926, 290.92, 2760, 51
2026/10/09 09:05:56.927, 290.37, 2760, 51
2026/10/09 09:05:57.927, 290.81, 2760, 51
2026/10/09 09:05:58.927, 290.69, 2760, 51
2026/10/09 09:05:59.927, 290.45, 2760, 51
2026/10/09 09:06:00.927, 290.52, 2760, 51
2026/10/09 09:06:01.928, 290.82, 2760, 51
2026/10/09 09:06:02.928, 290.64, 2760, 51
2026/10/09 09:06:03.928, 290.74, 2760, 51
2026/10/09 09:06:04.928, 291.16, 2760, 51
2026/10/09 09:06:05.928, 291.35, 2760, 51
2026/10/09 09:06:06.929, 291.37, 2760, 51
2026/10/09 09:06:07.929, 291.70, 2760, 51
2026/10/09 09:06:08.929, 291.63, 2760, 51
2026/10/09 09:06:09.929, 291.39, 2760, 51
2026/10/09 09:06:10.929, 291.13, 2760, 51
2026/10/09 09:06:11.929, 290.65, 2760, 51
2026/10/09 09:06:12.930, 291.24, 2760, 52
2026/10/09 09:06:13.930, 291.12, 2760, 52
2026/10/09 09:06:14.930, 291.26, 2760, 52
2026/10/09 09:06:15.930, 291.19, 2760, 52
2026/10/09 09:06:16.930, 290.99, 2760, 52
2026/10/09 09:06:17.931, 289.77, 2760, 52
2026/10/09 09:06:18.931, 290.39, 2760, 52
2026/10/09 09:06:19.931, 290.64, 2760, 52
2026/10/09 09:06:20.931, 290.93, 2760, 52
2026/10/09 09:06:21.932, 291.03, 2760, 52
2026/10/09 09:06:22.932, 291.53, 2760, 52
2026/10/09 09:06:23.932, 176.32, 2760, 45

5257
docs/analysis/class-v6/1p5x/x2/evidence/pairs/x2-select-r3.worker.log Normal file
View file

File diff suppressed because it is too large Load diff

3
docs/analysis/class-v6/1p5x/x2/evidence/pairs/x2-select-r4.row Normal file
View file

@ -0,0 +1,3 @@
card=NVIDIA_GeForce_RTX_4090 mode=select jobs=1024 batch_log2=22 hashes=4294967296 found=4223 setup_s=3.970 prod_wall_s=127.457 prod_mhs=33.70 prod_mean_w=280.5 prod_j_per_batch=34.9 prod_mh_per_w=0.1201 samples=128 steady_wall_s=123.487 steady_mhs=34.78 steady_mean_w=287.2 steady_j_per_batch=34.6 steady_mh_per_w=0.1211 samples=124 accepted=4223 over_target=0 nj_per_hash_steady=8257.6 j_per_share_steady=8.398 smi_ms=1000 worker_cpu_s=0.00 worker_sha=4152c960d1421007
ready: ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/af89be5ddbadb6f6b4aee28ac8f249713be5d4c12621e3cea7f83ceada3c66b3/day/69676e65756d2d6461792ffa50000000000000 dataset-log2 28 batch 4194304 regs 88 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback select worker 1.0 (4 October 2026)
transfers: transfers per chunk: down 342 B (select, 0 full fallbacks); mean per chunk: kernel+sync 119.76 ms, select 0.428 ms, read-back 0.196 ms, scan 0.008 ms; 1024 jobs 1024 chunks

130
docs/analysis/class-v6/1p5x/x2/evidence/pairs/x2-select-r4.smi Normal file
View file

@ -0,0 +1,130 @@
2026/10/09 09:06:30.330, 43.10, 210, 40
2026/10/09 09:06:31.338, 36.20, 210, 40
2026/10/09 09:06:32.338, 72.60, 2520, 41
2026/10/09 09:06:33.338, 86.35, 2520, 41
2026/10/09 09:06:34.339, 88.61, 2520, 40
2026/10/09 09:06:35.339, 85.50, 2520, 40
2026/10/09 09:06:36.339, 246.34, 2760, 47
2026/10/09 09:06:37.339, 287.22, 2760, 47
2026/10/09 09:06:38.340, 287.41, 2760, 47
2026/10/09 09:06:39.340, 286.68, 2760, 48
2026/10/09 09:06:40.340, 287.94, 2760, 48
2026/10/09 09:06:41.340, 287.66, 2760, 48
2026/10/09 09:06:42.340, 286.96, 2760, 48
2026/10/09 09:06:43.341, 288.18, 2760, 48
2026/10/09 09:06:44.341, 287.74, 2760, 48
2026/10/09 09:06:45.341, 287.61, 2760, 48
2026/10/09 09:06:46.341, 287.92, 2760, 48
2026/10/09 09:06:47.342, 287.86, 2760, 49
2026/10/09 09:06:48.342, 288.05, 2760, 49
2026/10/09 09:06:49.342, 287.85, 2760, 49
2026/10/09 09:06:50.342, 287.85, 2760, 49
2026/10/09 09:06:51.342, 287.73, 2760, 49
2026/10/09 09:06:52.343, 287.94, 2760, 49
2026/10/09 09:06:53.343, 288.28, 2760, 49
2026/10/09 09:06:54.343, 288.35, 2760, 49
2026/10/09 09:06:55.343, 289.17, 2760, 49
2026/10/09 09:06:56.344, 289.32, 2760, 49
2026/10/09 09:06:57.344, 289.35, 2760, 49
2026/10/09 09:06:58.344, 289.09, 2760, 49
2026/10/09 09:06:59.344, 288.91, 2760, 49
2026/10/09 09:07:00.345, 289.24, 2760, 49
2026/10/09 09:07:01.345, 289.90, 2760, 49
2026/10/09 09:07:02.345, 290.05, 2760, 50
2026/10/09 09:07:03.345, 290.09, 2760, 50
2026/10/09 09:07:04.345, 289.90, 2760, 50
2026/10/09 09:07:05.346, 289.67, 2760, 50
2026/10/09 09:07:06.346, 289.64, 2760, 50
2026/10/09 09:07:07.346, 289.58, 2760, 50
2026/10/09 09:07:08.346, 288.11, 2760, 50
2026/10/09 09:07:09.346, 273.31, 2760, 50
2026/10/09 09:07:10.347, 289.66, 2760, 50
2026/10/09 09:07:11.347, 289.07, 2760, 50
2026/10/09 09:07:12.347, 290.07, 2760, 50
2026/10/09 09:07:13.347, 289.72, 2760, 50
2026/10/09 09:07:14.347, 289.39, 2760, 50
2026/10/09 09:07:15.348, 261.93, 2760, 50
2026/10/09 09:07:16.348, 289.57, 2760, 50
2026/10/09 09:07:17.348, 289.11, 2760, 50
2026/10/09 09:07:18.348, 288.75, 2760, 50
2026/10/09 09:07:19.348, 289.74, 2760, 50
2026/10/09 09:07:20.348, 290.37, 2760, 50
2026/10/09 09:07:21.349, 290.27, 2760, 50
2026/10/09 09:07:22.349, 289.96, 2760, 50
2026/10/09 09:07:23.349, 290.61, 2760, 50
2026/10/09 09:07:24.349, 290.35, 2760, 50
2026/10/09 09:07:25.349, 290.05, 2760, 50
2026/10/09 09:07:26.350, 289.29, 2760, 51
2026/10/09 09:07:27.350, 288.93, 2760, 51
2026/10/09 09:07:28.350, 290.18, 2760, 51
2026/10/09 09:07:29.350, 290.11, 2760, 51
2026/10/09 09:07:30.350, 289.72, 2760, 51
2026/10/09 09:07:31.351, 288.96, 2760, 51
2026/10/09 09:07:32.351, 289.16, 2760, 51
2026/10/09 09:07:33.351, 289.38, 2760, 51
2026/10/09 09:07:34.351, 290.03, 2760, 51
2026/10/09 09:07:35.351, 289.74, 2760, 51
2026/10/09 09:07:36.352, 289.46, 2760, 51
2026/10/09 09:07:37.352, 289.91, 2760, 51
2026/10/09 09:07:38.352, 289.81, 2760, 51
2026/10/09 09:07:39.352, 289.53, 2760, 51
2026/10/09 09:07:40.352, 289.55, 2760, 51
2026/10/09 09:07:41.353, 289.54, 2760, 51
2026/10/09 09:07:42.353, 289.95, 2760, 51
2026/10/09 09:07:43.353, 289.73, 2760, 51
2026/10/09 09:07:44.353, 289.90, 2760, 51
2026/10/09 09:07:45.353, 290.21, 2760, 51
2026/10/09 09:07:46.353, 290.23, 2760, 51
2026/10/09 09:07:47.354, 290.85, 2760, 51
2026/10/09 09:07:48.354, 290.55, 2760, 51
2026/10/09 09:07:49.354, 290.43, 2760, 51
2026/10/09 09:07:50.354, 290.41, 2760, 51
2026/10/09 09:07:51.354, 290.09, 2760, 51
2026/10/09 09:07:52.355, 289.08, 2760, 51
2026/10/09 09:07:53.355, 290.65, 2760, 51
2026/10/09 09:07:54.355, 290.55, 2760, 51
2026/10/09 09:07:55.355, 275.14, 2760, 51
2026/10/09 09:07:56.355, 289.65, 2760, 51
2026/10/09 09:07:57.356, 289.93, 2760, 51
2026/10/09 09:07:58.356, 290.65, 2760, 51
2026/10/09 09:07:59.356, 290.93, 2760, 51
2026/10/09 09:08:00.356, 290.75, 2760, 51
2026/10/09 09:08:01.356, 291.05, 2760, 51
2026/10/09 09:08:02.357, 290.93, 2760, 51
2026/10/09 09:08:03.357, 290.65, 2760, 51
2026/10/09 09:08:04.357, 290.62, 2760, 51
2026/10/09 09:08:05.357, 290.72, 2760, 51
2026/10/09 09:08:06.357, 290.60, 2760, 51
2026/10/09 09:08:07.358, 290.71, 2760, 51
2026/10/09 09:08:08.358, 289.42, 2760, 51
2026/10/09 09:08:09.358, 290.10, 2760, 51
2026/10/09 09:08:10.358, 290.76, 2760, 51
2026/10/09 09:08:11.358, 290.32, 2760, 51
2026/10/09 09:08:12.359, 290.59, 2760, 51
2026/10/09 09:08:13.359, 290.76, 2760, 51
2026/10/09 09:08:14.359, 290.71, 2760, 51
2026/10/09 09:08:15.359, 291.16, 2760, 51
2026/10/09 09:08:16.360, 290.92, 2760, 51
2026/10/09 09:08:17.360, 290.89, 2760, 52
2026/10/09 09:08:18.360, 290.86, 2760, 51
2026/10/09 09:08:19.360, 290.56, 2760, 52
2026/10/09 09:08:20.360, 290.18, 2760, 52
2026/10/09 09:08:21.361, 290.49, 2760, 52
2026/10/09 09:08:22.361, 290.16, 2760, 52
2026/10/09 09:08:23.361, 290.18, 2760, 52
2026/10/09 09:08:24.361, 290.26, 2760, 52
2026/10/09 09:08:25.362, 290.56, 2760, 52
2026/10/09 09:08:26.362, 290.67, 2760, 52
2026/10/09 09:08:27.362, 290.38, 2760, 52
2026/10/09 09:08:28.362, 289.80, 2760, 52
2026/10/09 09:08:29.363, 290.32, 2760, 52
2026/10/09 09:08:30.363, 290.32, 2760, 52
2026/10/09 09:08:31.363, 290.53, 2760, 52
2026/10/09 09:08:32.363, 290.40, 2760, 52
2026/10/09 09:08:33.363, 290.35, 2760, 52
2026/10/09 09:08:34.364, 290.34, 2760, 52
2026/10/09 09:08:35.364, 290.40, 2760, 52
2026/10/09 09:08:36.364, 290.26, 2760, 52
2026/10/09 09:08:37.364, 290.37, 2760, 52
2026/10/09 09:08:38.364, 290.26, 2760, 52
2026/10/09 09:08:39.365, 200.92, 2760, 45

5257
docs/analysis/class-v6/1p5x/x2/evidence/pairs/x2-select-r4.worker.log Normal file
View file

File diff suppressed because it is too large Load diff

32
docs/analysis/class-v6/1p5x/x2/evidence/prod-v1/same-work-20261008-01.prod.compare.json Normal file
View file

@ -0,0 +1,32 @@
{
"ctx": "same-work-20261008-01",
"shape": "prod",
"expected_sha256": "52a0f79483ce948b83c4bcde39ce37b89fdf04091826e14f56260717e0d8fce6",
"modes": {
"fault": "NOT RUN",
"full": {
"sha256": "2f7f1de48a50650a50daf8835c6c86eba981a155b28cc4bc5a5990f74ea56a12",
"equal_to_expected": false,
"first_difference": {
"line": 1048578,
"expected": "found 2 5379 0010510133e9ba59",
"got": "error 2 error"
},
"lines_missing": 2097931,
"lines_extra": 1
},
"select": {
"sha256": "2f7f1de48a50650a50daf8835c6c86eba981a155b28cc4bc5a5990f74ea56a12",
"equal_to_expected": false,
"first_difference": {
"line": 1048578,
"expected": "found 2 5379 0010510133e9ba59",
"got": "error 2 error"
},
"lines_missing": 2097931,
"lines_extra": 1
}
},
"known_failed_flagged": null,
"verdict": "DIFFERENT"
}

24
docs/analysis/class-v6/1p5x/x2/evidence/prod-v1/same-work-20261008-01.prod.expected.meta.json Normal file
View file

@ -0,0 +1,24 @@
{
"ctx": "same-work-20261008-01",
"shape": "prod",
"jobs": 10,
"nonces": 9437184,
"found_lines": 3146498,
"labels": {
"p1-all": 1,
"p1-t256": 1,
"p1-t257": 1,
"p1-zero": 1,
"p1-teq": 1,
"p1-teq-1": 1,
"p2a-all": 1,
"p2b-all": 1,
"p3-all": 1,
"p3-t256": 1
},
"norm_sha256": "52a0f79483ce948b83c4bcde39ce37b89fdf04091826e14f56260717e0d8fce6",
"references": {
"D": "c49fed11d7279fee13c10a904ab21dd177c5ffad7974c049067ad72c23250b6f",
"D1": "5ae1220d24bed2677e0a9440f90393ae769da099d760dd90edbeef0bc864a914"
}
}

6
docs/analysis/class-v6/1p5x/x2/evidence/prod-v1/same-work-20261008-01.prod.full.log Normal file
View file

@ -0,0 +1,6 @@
info igneum-worker-cuda 1.0 (4 October 2026): device 0 NVIDIA_GeForce_RTX_4090 (sm_89, 128 SMs), driver 13.2 from libcuda.so.1, NVRTC 12.8 from libnvrtc.so.12, target sm_89 (the device's architecture, listed by NVRTC)
ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/edc4fa844da9dc98d37e965176f6558a31560e40502ab3ae5491b21aaaabfb07/day/69676e65756d2d6461792ffa50000000000000 dataset-log2 28 batch 4194304 regs 93 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback full worker 1.0 (4 October 2026)
info readback full (per dispatch of 4194304 nonces: 8 bytes per nonce come back and the host scans them)
info first pack /root/x2/ctx/same-work-20261008-01/packs/hl-v6-all: nvrtc 2354 cache 2 dataset 29 hot 0 check 1105 race 0 ms variant base class v6 (state leaves 93, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 448274a57f508cbc; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes)
done 1 1048576 1454.31
error 2 worker fault: the output buffer is unchanged since the previous dispatch (signature ef40224d5958063d): the kernel did not run; exiting 3 so the miner restarts the worker

6
docs/analysis/class-v6/1p5x/x2/evidence/prod-v1/same-work-20261008-01.prod.select.log Normal file
View file

@ -0,0 +1,6 @@
info igneum-worker-cuda 1.0 (4 October 2026): device 0 NVIDIA_GeForce_RTX_4090 (sm_89, 128 SMs), driver 13.2 from libcuda.so.1, NVRTC 12.8 from libnvrtc.so.12, target sm_89 (the device's architecture, listed by NVRTC)
ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/edc4fa844da9dc98d37e965176f6558a31560e40502ab3ae5491b21aaaabfb07/day/69676e65756d2d6461792ffa50000000000000 dataset-log2 28 batch 4194304 regs 93 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback select worker 1.0 (4 October 2026)
info readback select (per dispatch of 4194304 nonces: the hits and 34 sentinel words come back; the GPU scans)
info first pack /root/x2/ctx/same-work-20261008-01/packs/hl-v6-all: nvrtc 2382 cache 2 dataset 29 hot 0 check 1100 race 0 ms variant base class v6 (state leaves 93, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 448274a57f508cbc; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes)
done 1 1048576 1476.27
error 2 worker fault: the output buffer is unchanged since the previous dispatch (signature ef40224d5958063d): the kernel did not run; exiting 3 so the miner restarts the worker

32
docs/analysis/class-v6/1p5x/x2/evidence/prod-v1/same-work-20261008-02.prod.compare.json Normal file
View file

@ -0,0 +1,32 @@
{
"ctx": "same-work-20261008-02",
"shape": "prod",
"expected_sha256": "db89fbfc3a7a7b595555e9d150fa82b93ef2ecc1f0944f69f759b3ebd39cc685",
"modes": {
"fault": "NOT RUN",
"full": {
"sha256": "f8a898c426596d8a482139309d6d9a85493ff2d1da8ef099406d97db168bcc9f",
"equal_to_expected": false,
"first_difference": {
"line": 1048578,
"expected": "found 2 511 000c4b134df33d22",
"got": "error 2 error"
},
"lines_missing": 2097931,
"lines_extra": 1
},
"select": {
"sha256": "f8a898c426596d8a482139309d6d9a85493ff2d1da8ef099406d97db168bcc9f",
"equal_to_expected": false,
"first_difference": {
"line": 1048578,
"expected": "found 2 511 000c4b134df33d22",
"got": "error 2 error"
},
"lines_missing": 2097931,
"lines_extra": 1
}
},
"known_failed_flagged": null,
"verdict": "DIFFERENT"
}

24
docs/analysis/class-v6/1p5x/x2/evidence/prod-v1/same-work-20261008-02.prod.expected.meta.json Normal file
View file

@ -0,0 +1,24 @@
{
"ctx": "same-work-20261008-02",
"shape": "prod",
"jobs": 10,
"nonces": 9437184,
"found_lines": 3146498,
"labels": {
"p1-all": 1,
"p1-t256": 1,
"p1-t257": 1,
"p1-zero": 1,
"p1-teq": 1,
"p1-teq-1": 1,
"p2a-all": 1,
"p2b-all": 1,
"p3-all": 1,
"p3-t256": 1
},
"norm_sha256": "db89fbfc3a7a7b595555e9d150fa82b93ef2ecc1f0944f69f759b3ebd39cc685",
"references": {
"D": "7c7be9b549b191ffd628aa4722c6222e1752eac51d674bc0eefe518a67cedf48",
"D1": "cad7849f8df4e7d8faa6c0ffdb5561e90fbd1274c465cfd1efc61d8015122753"
}
}

6
docs/analysis/class-v6/1p5x/x2/evidence/prod-v1/same-work-20261008-02.prod.full.log Normal file
View file

@ -0,0 +1,6 @@
info igneum-worker-cuda 1.0 (4 October 2026): device 0 NVIDIA_GeForce_RTX_4090 (sm_89, 128 SMs), driver 13.2 from libcuda.so.1, NVRTC 12.8 from libnvrtc.so.12, target sm_89 (the device's architecture, listed by NVRTC)
ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/af89be5ddbadb6f6b4aee28ac8f249713be5d4c12621e3cea7f83ceada3c66b3/day/69676e65756d2d6461792ffa50000000000000 dataset-log2 28 batch 4194304 regs 88 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback full worker 1.0 (4 October 2026)
info readback full (per dispatch of 4194304 nonces: 8 bytes per nonce come back and the host scans them)
info first pack /root/x2/ctx/same-work-20261008-02/packs/chain-seed-d20730: nvrtc 2536 cache 3 dataset 29 hot 0 check 1100 race 0 ms variant base class v6 (state leaves 93, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 448274a57f508cbc; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes)
done 1 1048576 1496.81
error 2 worker fault: the output buffer is unchanged since the previous dispatch (signature 4afda524233fb274): the kernel did not run; exiting 3 so the miner restarts the worker

6
docs/analysis/class-v6/1p5x/x2/evidence/prod-v1/same-work-20261008-02.prod.select.log Normal file
View file

@ -0,0 +1,6 @@
info igneum-worker-cuda 1.0 (4 October 2026): device 0 NVIDIA_GeForce_RTX_4090 (sm_89, 128 SMs), driver 13.2 from libcuda.so.1, NVRTC 12.8 from libnvrtc.so.12, target sm_89 (the device's architecture, listed by NVRTC)
ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/af89be5ddbadb6f6b4aee28ac8f249713be5d4c12621e3cea7f83ceada3c66b3/day/69676e65756d2d6461792ffa50000000000000 dataset-log2 28 batch 4194304 regs 88 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback select worker 1.0 (4 October 2026)
info readback select (per dispatch of 4194304 nonces: the hits and 34 sentinel words come back; the GPU scans)
info first pack /root/x2/ctx/same-work-20261008-02/packs/chain-seed-d20730: nvrtc 2548 cache 3 dataset 29 hot 0 check 1111 race 0 ms variant base class v6 (state leaves 93, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 448274a57f508cbc; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes)
done 1 1048576 1489.00
error 2 worker fault: the output buffer is unchanged since the previous dispatch (signature 4afda524233fb274): the kernel did not run; exiting 3 so the miner restarts the worker

32
docs/analysis/class-v6/1p5x/x2/evidence/prod-v1/same-work-20261008-03.prod.compare.json Normal file
View file

@ -0,0 +1,32 @@
{
"ctx": "same-work-20261008-03",
"shape": "prod",
"expected_sha256": "8e9c9955eda5f46377ffb0afd5e0558760019885d725d2547c1232c6fc939362",
"modes": {
"fault": "NOT RUN",
"full": {
"sha256": "40778c06a13e40b329f17d9899bf8bcded6e2597f22670e6925d5d7aa5adcc75",
"equal_to_expected": false,
"first_difference": {
"line": 1048578,
"expected": "found 2 2348 0003dfe8b4a7fee3",
"got": "error 2 error"
},
"lines_missing": 2097931,
"lines_extra": 1
},
"select": {
"sha256": "40778c06a13e40b329f17d9899bf8bcded6e2597f22670e6925d5d7aa5adcc75",
"equal_to_expected": false,
"first_difference": {
"line": 1048578,
"expected": "found 2 2348 0003dfe8b4a7fee3",
"got": "error 2 error"
},
"lines_missing": 2097931,
"lines_extra": 1
}
},
"known_failed_flagged": null,
"verdict": "DIFFERENT"
}

24
docs/analysis/class-v6/1p5x/x2/evidence/prod-v1/same-work-20261008-03.prod.expected.meta.json Normal file
View file

@ -0,0 +1,24 @@
{
"ctx": "same-work-20261008-03",
"shape": "prod",
"jobs": 10,
"nonces": 9437184,
"found_lines": 3146498,
"labels": {
"p1-all": 1,
"p1-t256": 1,
"p1-t257": 1,
"p1-zero": 1,
"p1-teq": 1,
"p1-teq-1": 1,
"p2a-all": 1,
"p2b-all": 1,
"p3-all": 1,
"p3-t256": 1
},
"norm_sha256": "8e9c9955eda5f46377ffb0afd5e0558760019885d725d2547c1232c6fc939362",
"references": {
"D": "873528b18ad679ca9a748d52d45c4e1b6723ba703875742fb01e1cbfca169783",
"D1": "7f8e64ae899ed5452dab6a01a2b34d66f5a11ef1ca8c65a9b93ff0be2a88250c"
}
}

6
docs/analysis/class-v6/1p5x/x2/evidence/prod-v1/same-work-20261008-03.prod.full.log Normal file
View file

@ -0,0 +1,6 @@
info igneum-worker-cuda 1.0 (4 October 2026): device 0 NVIDIA_GeForce_RTX_4090 (sm_89, 128 SMs), driver 13.2 from libcuda.so.1, NVRTC 12.8 from libnvrtc.so.12, target sm_89 (the device's architecture, listed by NVRTC)
ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/3c3fab434f7fd894ce6db0b09f298c3865726253e4fc68fd77acd163ccea8221/day/69676e65756d2d6461792ffe50000000000000 dataset-log2 28 batch 4194304 regs 31 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback full worker 1.0 (4 October 2026)
info readback full (per dispatch of 4194304 nonces: 8 bytes per nonce come back and the host scans them)
info first pack /root/x2/ctx/same-work-20261008-03/packs/v5-epoch2-d20734: nvrtc 633 cache 3 dataset 28 hot 0 check 1211 race 0 ms variant base class v5 (state leaves 224, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 5cc75af0f884bd4c; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes)
done 1 1048576 1415.81
error 2 worker fault: the output buffer is unchanged since the previous dispatch (signature bcaf7511107f96a1): the kernel did not run; exiting 3 so the miner restarts the worker

6
docs/analysis/class-v6/1p5x/x2/evidence/prod-v1/same-work-20261008-03.prod.select.log Normal file
View file

@ -0,0 +1,6 @@
info igneum-worker-cuda 1.0 (4 October 2026): device 0 NVIDIA_GeForce_RTX_4090 (sm_89, 128 SMs), driver 13.2 from libcuda.so.1, NVRTC 12.8 from libnvrtc.so.12, target sm_89 (the device's architecture, listed by NVRTC)
ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/3c3fab434f7fd894ce6db0b09f298c3865726253e4fc68fd77acd163ccea8221/day/69676e65756d2d6461792ffe50000000000000 dataset-log2 28 batch 4194304 regs 31 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback select worker 1.0 (4 October 2026)
info readback select (per dispatch of 4194304 nonces: the hits and 34 sentinel words come back; the GPU scans)
info first pack /root/x2/ctx/same-work-20261008-03/packs/v5-epoch2-d20734: nvrtc 621 cache 2 dataset 36 hot 0 check 1174 race 0 ms variant base class v5 (state leaves 224, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 5cc75af0f884bd4c; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes)
done 1 1048576 1431.08
error 2 worker fault: the output buffer is unchanged since the previous dispatch (signature bcaf7511107f96a1): the kernel did not run; exiting 3 so the miner restarts the worker

58
docs/analysis/class-v6/1p5x/x2/evidence/run-equiv.log Normal file
View file

@ -0,0 +1,58 @@
== 08:45:49 known-failed first: kf plan on same-work-20261008-01, fault then select
{"ctx": "same-work-20261008-01", "shape": "kf", "expected_sha256": "e3e8913fb6788dfbbba4d43de14d674fa411de019980e2fc543be4ed7c17303c", "modes": {"fault": {"sha256": "f7ad6ee87564f49abaee8e8ee64b62e41c6a45259c0cace233aecb898cbd78aa", "equal_to_expected": false, "first_difference": {"line": 64, "expected": "found 1 1053 0fa540bb2045b271", "got": "found 1 1037 00cb3ac51a2ec48e"}, "lines_missing": 0, "lines_extra": 368}, "full": {"sha256": "e3e8913fb6788dfbbba4d43de14d674fa411de019980e2fc543be4ed7c17303c", "equal_to_expected": true, "first_difference": null, "lines_missing": 0, "lines_extra": 0}, "select": {"sha256": "e3e8913fb6788dfbbba4d43de14d674fa411de019980e2fc543be4ed7c17303c", "equal_to_expected": true, "first_difference": null, "lines_missing": 0, "lines_extra": 0}}, "known_failed_flagged": true, "verdict": "EQUAL"}
known-failed: flagged {'line': 64, 'expected': 'found 1 1053 0fa540bb2045b271', 'got': 'found 1 1037 00cb3ac51a2ec48e'} | kf full/select: EQUAL
== 08:46:13 same-work-20261008-01 prod: expected, full, select
Traceback (most recent call last):
File "/root/x2/tools/x2-equiv.py", line 416, in <module>
sys.exit({'run': run, 'expect': expect, 'compare': compare}[a.cmd](a))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/root/x2/tools/x2-equiv.py", line 222, in run
p.stdin.write(f"job {seq} {ctx['prehash']} {t:016x} {s} {c} {pk['epoch']} {pk['day']} {ctx['extra']}\n")
BrokenPipeError: [Errno 32] Broken pipe
Traceback (most recent call last):
File "/root/x2/tools/x2-equiv.py", line 416, in <module>
sys.exit({'run': run, 'expect': expect, 'compare': compare}[a.cmd](a))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/root/x2/tools/x2-equiv.py", line 222, in run
p.stdin.write(f"job {seq} {ctx['prehash']} {t:016x} {s} {c} {pk['epoch']} {pk['day']} {ctx['extra']}\n")
BrokenPipeError: [Errno 32] Broken pipe
same-work-20261008-01 prod DIFFERENT {'fault': 'NOT RUN', 'full': (False, '2f7f1de48a50650a'), 'select': (False, '2f7f1de48a50650a')} expected 52a0f79483ce948b
== 08:46:47 same-work-20261008-01 fuzz: expected, full, select
same-work-20261008-01 fuzz EQUAL {'fault': 'NOT RUN', 'full': (True, '5f1e591db673f964'), 'select': (True, '5f1e591db673f964')} expected 5f1e591db673f964
== 08:47:14 same-work-20261008-02 prod: expected, full, select
Traceback (most recent call last):
File "/root/x2/tools/x2-equiv.py", line 416, in <module>
sys.exit({'run': run, 'expect': expect, 'compare': compare}[a.cmd](a))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/root/x2/tools/x2-equiv.py", line 222, in run
p.stdin.write(f"job {seq} {ctx['prehash']} {t:016x} {s} {c} {pk['epoch']} {pk['day']} {ctx['extra']}\n")
BrokenPipeError: [Errno 32] Broken pipe
Traceback (most recent call last):
File "/root/x2/tools/x2-equiv.py", line 416, in <module>
sys.exit({'run': run, 'expect': expect, 'compare': compare}[a.cmd](a))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/root/x2/tools/x2-equiv.py", line 222, in run
p.stdin.write(f"job {seq} {ctx['prehash']} {t:016x} {s} {c} {pk['epoch']} {pk['day']} {ctx['extra']}\n")
BrokenPipeError: [Errno 32] Broken pipe
same-work-20261008-02 prod DIFFERENT {'fault': 'NOT RUN', 'full': (False, 'f8a898c426596d8a'), 'select': (False, 'f8a898c426596d8a')} expected db89fbfc3a7a7b59
== 08:47:48 same-work-20261008-02 fuzz: expected, full, select
same-work-20261008-02 fuzz EQUAL {'fault': 'NOT RUN', 'full': (True, '8f39a1b15b16e86f'), 'select': (True, '8f39a1b15b16e86f')} expected 8f39a1b15b16e86f
== 08:48:16 same-work-20261008-03 prod: expected, full, select
Traceback (most recent call last):
File "/root/x2/tools/x2-equiv.py", line 416, in <module>
sys.exit({'run': run, 'expect': expect, 'compare': compare}[a.cmd](a))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/root/x2/tools/x2-equiv.py", line 222, in run
p.stdin.write(f"job {seq} {ctx['prehash']} {t:016x} {s} {c} {pk['epoch']} {pk['day']} {ctx['extra']}\n")
BrokenPipeError: [Errno 32] Broken pipe
Traceback (most recent call last):
File "/root/x2/tools/x2-equiv.py", line 416, in <module>
sys.exit({'run': run, 'expect': expect, 'compare': compare}[a.cmd](a))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/root/x2/tools/x2-equiv.py", line 222, in run
p.stdin.write(f"job {seq} {ctx['prehash']} {t:016x} {s} {c} {pk['epoch']} {pk['day']} {ctx['extra']}\n")
BrokenPipeError: [Errno 32] Broken pipe
same-work-20261008-03 prod DIFFERENT {'fault': 'NOT RUN', 'full': (False, '40778c06a13e40b3'), 'select': (False, '40778c06a13e40b3')} expected 8e9c9955eda5f463
== 08:48:45 same-work-20261008-03 fuzz: expected, full, select
same-work-20261008-03 fuzz EQUAL {'fault': 'NOT RUN', 'full': (True, '3a7a7cc495e2d485'), 'select': (True, '3a7a7cc495e2d485')} expected 3a7a7cc495e2d485
== 08:49:04 end, fail=1

23
docs/analysis/class-v6/1p5x/x2/evidence/run-pairs.log Normal file
View file

@ -0,0 +1,23 @@
== 08:52:20 idle card before the rows
name, driver_version, power.limit [W], power.draw [W], clocks.current.sm [MHz], clocks.current.memory [MHz], temperature.gpu, utilization.gpu [%], pcie.link.gen.current, pcie.link.width.current
NVIDIA GeForce RTX 4090, 595.91.07, 450.00 W, 30.54 W, 210 MHz, 405 MHz, 25, 0 %, 1, 16
== 08:52:20 pair 1 full
card=NVIDIA_GeForce_RTX_4090 mode=full jobs=1024 batch_log2=22 hashes=4294967296 found=4223 setup_s=4.238 prod_wall_s=139.523 prod_mhs=30.78 prod_mean_w=261.1 prod_j_per_batch=35.6 prod_mh_per_w=0.1179 samples=140 steady_wall_s=135.285 steady_mhs=31.75 steady_mean_w=268.4 steady_j_per_batch=35.5 steady_mh_per_w=0.1183 samples=135 accepted=4223 over_target=0 nj_per_hash_steady=8453.5 j_per_share_steady=8.598 smi_ms=1000 worker_cpu_s=0.00 worker_sha=4152c960d1421007
== 08:54:48 pair 1 select
card=NVIDIA_GeForce_RTX_4090 mode=select jobs=1024 batch_log2=22 hashes=4294967296 found=4223 setup_s=4.465 prod_wall_s=127.590 prod_mhs=33.66 prod_mean_w=279.5 prod_j_per_batch=34.8 prod_mh_per_w=0.1204 samples=128 steady_wall_s=123.125 steady_mhs=34.88 steady_mean_w=288.0 steady_j_per_batch=34.6 steady_mh_per_w=0.1211 samples=123 accepted=4223 over_target=0 nj_per_hash_steady=8256.9 j_per_share_steady=8.397 smi_ms=1000 worker_cpu_s=0.00 worker_sha=4152c960d1421007
== 08:57:04 pair 2 select
card=NVIDIA_GeForce_RTX_4090 mode=select jobs=1024 batch_log2=22 hashes=4294967296 found=4223 setup_s=4.041 prod_wall_s=127.155 prod_mhs=33.78 prod_mean_w=281.5 prod_j_per_batch=35.0 prod_mh_per_w=0.1200 samples=128 steady_wall_s=123.114 steady_mhs=34.89 steady_mean_w=290.0 steady_j_per_batch=34.9 steady_mh_per_w=0.1203 samples=123 accepted=4223 over_target=0 nj_per_hash_steady=8311.8 j_per_share_steady=8.454 smi_ms=1000 worker_cpu_s=0.00 worker_sha=4152c960d1421007
== 08:59:19 pair 2 full
card=NVIDIA_GeForce_RTX_4090 mode=full jobs=1024 batch_log2=22 hashes=4294967296 found=4223 setup_s=4.005 prod_wall_s=137.908 prod_mhs=31.14 prod_mean_w=267.3 prod_j_per_batch=36.0 prod_mh_per_w=0.1165 samples=138 steady_wall_s=133.903 steady_mhs=32.08 steady_mean_w=273.1 steady_j_per_batch=35.7 steady_mh_per_w=0.1175 samples=134 accepted=4223 over_target=0 nj_per_hash_steady=8513.1 j_per_share_steady=8.659 smi_ms=1000 worker_cpu_s=0.00 worker_sha=4152c960d1421007
== 09:01:45 pair 3 full
card=NVIDIA_GeForce_RTX_4090 mode=full jobs=1024 batch_log2=22 hashes=4294967296 found=4223 setup_s=4.100 prod_wall_s=141.006 prod_mhs=30.46 prod_mean_w=262.8 prod_j_per_batch=36.2 prod_mh_per_w=0.1159 samples=141 steady_wall_s=136.906 steady_mhs=31.37 steady_mean_w=269.9 steady_j_per_batch=36.1 steady_mh_per_w=0.1162 samples=136 accepted=4223 over_target=0 nj_per_hash_steady=8603.8 j_per_share_steady=8.750 smi_ms=1000 worker_cpu_s=0.00 worker_sha=4152c960d1421007
== 09:04:14 pair 3 select
card=NVIDIA_GeForce_RTX_4090 mode=select jobs=1024 batch_log2=22 hashes=4294967296 found=4223 setup_s=4.067 prod_wall_s=127.175 prod_mhs=33.77 prod_mean_w=280.8 prod_j_per_batch=34.9 prod_mh_per_w=0.1203 samples=128 steady_wall_s=123.107 steady_mhs=34.89 steady_mean_w=289.3 steady_j_per_batch=34.8 steady_mh_per_w=0.1206 samples=123 accepted=4223 over_target=0 nj_per_hash_steady=8291.8 j_per_share_steady=8.434 smi_ms=1000 worker_cpu_s=0.00 worker_sha=4152c960d1421007
== 09:06:30 pair 4 select
card=NVIDIA_GeForce_RTX_4090 mode=select jobs=1024 batch_log2=22 hashes=4294967296 found=4223 setup_s=3.970 prod_wall_s=127.457 prod_mhs=33.70 prod_mean_w=280.5 prod_j_per_batch=34.9 prod_mh_per_w=0.1201 samples=128 steady_wall_s=123.487 steady_mhs=34.78 steady_mean_w=287.2 steady_j_per_batch=34.6 steady_mh_per_w=0.1211 samples=124 accepted=4223 over_target=0 nj_per_hash_steady=8257.6 j_per_share_steady=8.398 smi_ms=1000 worker_cpu_s=0.00 worker_sha=4152c960d1421007
== 09:08:46 pair 4 full
card=NVIDIA_GeForce_RTX_4090 mode=full jobs=1024 batch_log2=22 hashes=4294967296 found=4223 setup_s=4.010 prod_wall_s=139.724 prod_mhs=30.74 prod_mean_w=264.4 prod_j_per_batch=36.1 prod_mh_per_w=0.1163 samples=140 steady_wall_s=135.714 steady_mhs=31.65 steady_mean_w=270.2 steady_j_per_batch=35.8 steady_mh_per_w=0.1171 samples=136 accepted=4223 over_target=0 nj_per_hash_steady=8537.1 j_per_share_steady=8.683 smi_ms=1000 worker_cpu_s=0.00 worker_sha=4152c960d1421007
== 09:11:14 kernel timer (--bench 250 x 2^24, nvidia-smi 1 Hz)
RESULT pack=/root/x2/out/pairs/pack-chain-seed-d20730 class=mx8-erad810f22d+sh256x27+state+reg64c+fold+rw device=NVIDIA_GeForce_RTX_4090 arch=sm_89 regs=88 blocks_per_sm=20 warps=0 resident=2560 arena_mib=0 hot_mib=0 hot_slots=0 hot_fill_ms=0.00 nonces=16777216 batches=250 check=PASS fingerprint=01f51b9d4805e5e6 mhs=35.057 loads=256 bytes=1024 scratch_ops=0 time=wall
kernel_mhs=35.057 loaded_mean_w=289.8 samples=120 nj_per_hash=8267.3 fingerprint=01f51b9d4805e5e6
== 09:13:22 end

7
docs/analysis/class-v6/1p5x/x2/evidence/run-prod-then-pairs.log Normal file
View file

@ -0,0 +1,7 @@
== 08:49:30 same-work-20261008-01 prod
same-work-20261008-01 prod EQUAL {'fault': 'NOT RUN', 'full': (True, 'f278f0594dfd566d', None), 'select': (True, 'f278f0594dfd566d', None)} expected f278f0594dfd566d
== 08:50:18 same-work-20261008-02 prod
same-work-20261008-02 prod EQUAL {'fault': 'NOT RUN', 'full': (True, 'ad091a0d49a9d3a2', None), 'select': (True, 'ad091a0d49a9d3a2', None)} expected ad091a0d49a9d3a2
== 08:51:04 same-work-20261008-03 prod
same-work-20261008-03 prod EQUAL {'fault': 'NOT RUN', 'full': (True, '2aa97d5d63e5e01c', None), 'select': (True, '2aa97d5d63e5e01c', None)} expected 2aa97d5d63e5e01c
== 08:51:42 pairs

24
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-01.fuzz.compare.json Normal file
View file

@ -0,0 +1,24 @@
{
"ctx": "same-work-20261008-01",
"shape": "fuzz",
"expected_sha256": "5f1e591db673f964ff53a60ebb9ea0e8f4cad974642a2354b3cc27571d935340",
"modes": {
"fault": "NOT RUN",
"full": {
"sha256": "5f1e591db673f964ff53a60ebb9ea0e8f4cad974642a2354b3cc27571d935340",
"equal_to_expected": true,
"first_difference": null,
"lines_missing": 0,
"lines_extra": 0
},
"select": {
"sha256": "5f1e591db673f964ff53a60ebb9ea0e8f4cad974642a2354b3cc27571d935340",
"equal_to_expected": true,
"first_difference": null,
"lines_missing": 0,
"lines_extra": 0
}
},
"known_failed_flagged": null,
"verdict": "EQUAL"
}

26
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-01.fuzz.expected.meta.json Normal file
View file

@ -0,0 +1,26 @@
{
"ctx": "same-work-20261008-01",
"shape": "fuzz",
"jobs": 38,
"nonces": 1048576,
"found_lines": 438688,
"labels": {
"fz3": 4,
"fz1": 5,
"fz0": 7,
"fz5": 4,
"fz4": 4,
"fz2": 2,
"fz6": 4,
"fz7": 4,
"blk-d-all": 1,
"blk-d-q4": 1,
"blk-d1-all": 1,
"blk-d1-q4": 1
},
"norm_sha256": "5f1e591db673f964ff53a60ebb9ea0e8f4cad974642a2354b3cc27571d935340",
"references": {
"D": "c49fed11d7279fee13c10a904ab21dd177c5ffad7974c049067ad72c23250b6f",
"D1": "5ae1220d24bed2677e0a9440f90393ae769da099d760dd90edbeef0bc864a914"
}
}

57
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-01.fuzz.full.log Normal file
View file

@ -0,0 +1,57 @@
info igneum-worker-cuda 1.0 (4 October 2026): device 0 NVIDIA_GeForce_RTX_4090 (sm_89, 128 SMs), driver 13.2 from libcuda.so.1, NVRTC 12.8 from libnvrtc.so.12, target sm_89 (the device's architecture, listed by NVRTC)
ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/edc4fa844da9dc98d37e965176f6558a31560e40502ab3ae5491b21aaaabfb07/day/69676e65756d2d6461792ffa50000000000000 dataset-log2 28 batch 1024 regs 93 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback full worker 1.0 (4 October 2026)
info readback full (per dispatch of 1024 nonces: 8 bytes per nonce come back and the host scans them)
info first pack /root/x2/ctx/same-work-20261008-01/packs/hl-v6-all: nvrtc 2372 cache 2 dataset 29 hot 0 check 1144 race 0 ms variant base class v6 (state leaves 93, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 448274a57f508cbc; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes)
done 1 6048 1.90
done 2 63040 101.73
done 3 24768 6.07
done 4 20448 25.13
done 5 6560 2.62
done 6 3552 2.43
done 7 54080 12.48
done 8 63136 88.31
done 9 8096 1.86
done 10 60320 14.59
done 11 60288 26.75
done 12 4320 5.34
done 13 27456 12.25
done 14 43648 24.34
done 15 24992 6.14
done 16 60896 16.70
done 17 29888 27.06
done 18 1696 1.64
done 19 4096 5.20
done 20 4096 1.99
info prepare started for epoch edc4fa844da9dc98 day 69676e65756d2d6461792ffb50000000000000 from /root/x2/ctx/same-work-20261008-01/packs/hl-v6-all-d20731 (NVRTC sm_89 in the background)
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
prepared edc4fa844da9dc98d37e965176f6558a31560e40502ab3ae5491b21aaaabfb07 69676e65756d2d6461792ffb50000000000000 4116.0 nvrtc 2661 cache 3 dataset 28 hot 0 check 1197 race 0 ms variant base class v6 (state leaves 93, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 5aa17d87ef2c2856; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes) resident 2 programs 2 datasets
info ignored: noop
info switched to the prepared pair epoch edc4fa844da9dc98 day 69676e65756d2d6461792ffb50000000000000 (class v6) in 0.00 ms
done 21 4096 3.27
info dropped the previous pair (its program, cache and dataset)
done 22 4096 1.21
done 23 1920 0.48
done 24 21024 10.40
done 25 58176 80.62
done 26 16928 4.76
done 27 9216 18.84
done 28 61344 17.54
done 29 59360 90.47
done 30 33952 12.25
done 31 23968 32.20
done 32 29184 6.85
done 33 34944 10.29
done 34 2592 0.71
done 35 18816 20.09
done 36 55392 23.00
done 37 35904 10.80
done 38 6240 1.79
info transfers per chunk: down 8074 B (full, 0 full fallbacks); mean per chunk: kernel+sync 0.24 ms, select 0.000 ms, read-back 0.048 ms, scan 0.350 ms; 38 jobs 1039 chunks

19
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-01.fuzz.full.meta.json Normal file
View file

@ -0,0 +1,19 @@
{
"ctx": "same-work-20261008-01",
"object_id": "0x4de7b836cc40a4ea",
"shape": "fuzz",
"mode": "full",
"batch_log2": 10,
"jobs": 38,
"nonces": 1048576,
"worker": "/root/x2/bin/igneum-worker-cuda",
"worker_sha256": "4152c960d14210072a5c85ae675b56ae4dab12bedfb5f1af04ea2a2c6c025032",
"ready": "ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/edc4fa844da9dc98d37e965176f6558a31560e40502ab3ae5491b21aaaabfb07/day/69676e65756d2d6461792ffa50000000000000 dataset-log2 28 batch 1024 regs 93 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback full worker 1.0 (4 October 2026)",
"prepared": "prepared edc4fa844da9dc98d37e965176f6558a31560e40502ab3ae5491b21aaaabfb07 69676e65756d2d6461792ffb50000000000000 4116.0 nvrtc 2661 cache 3 dataset 28 hot 0 check 1197 race 0 ms variant base class v6 (state leaves 93, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 5aa17d87ef2c2856; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes) resident",
"transfers": "info transfers per chunk: down 8074 B (full, 0 full fallbacks); mean per chunk: kernel+sync 0.24 ms, select 0.000 ms, read-back 0.048 ms, scan 0.350 ms; 38 jobs 1039 chunks",
"errors": [],
"norm_sha256": "5f1e591db673f964ff53a60ebb9ea0e8f4cad974642a2354b3cc27571d935340",
"seconds": 9.3,
"exit": 0,
"at": "2026-10-09T08:47:02Z"
}

57
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-01.fuzz.select.log Normal file
View file

@ -0,0 +1,57 @@
info igneum-worker-cuda 1.0 (4 October 2026): device 0 NVIDIA_GeForce_RTX_4090 (sm_89, 128 SMs), driver 13.2 from libcuda.so.1, NVRTC 12.8 from libnvrtc.so.12, target sm_89 (the device's architecture, listed by NVRTC)
ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/edc4fa844da9dc98d37e965176f6558a31560e40502ab3ae5491b21aaaabfb07/day/69676e65756d2d6461792ffa50000000000000 dataset-log2 28 batch 1024 regs 93 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback select worker 1.0 (4 October 2026)
info readback select (per dispatch of 1024 nonces: the hits and 34 sentinel words come back; the GPU scans)
info first pack /root/x2/ctx/same-work-20261008-01/packs/hl-v6-all: nvrtc 2620 cache 2 dataset 29 hot 0 check 1168 race 0 ms variant base class v6 (state leaves 93, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 448274a57f508cbc; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes)
done 1 6048 2.45
done 2 63040 86.67
done 3 24768 10.14
done 4 20448 16.31
done 5 6560 2.62
done 6 3552 2.06
done 7 54080 15.71
done 8 63136 97.68
done 9 8096 2.66
done 10 60320 18.69
done 11 60288 41.87
done 12 4320 6.34
done 13 27456 14.70
done 14 43648 19.91
done 15 24992 8.61
done 16 60896 22.71
done 17 29888 37.58
done 18 1696 1.77
done 19 4096 5.49
done 20 4096 2.31
info prepare started for epoch edc4fa844da9dc98 day 69676e65756d2d6461792ffb50000000000000 from /root/x2/ctx/same-work-20261008-01/packs/hl-v6-all-d20731 (NVRTC sm_89 in the background)
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
prepared edc4fa844da9dc98d37e965176f6558a31560e40502ab3ae5491b21aaaabfb07 69676e65756d2d6461792ffb50000000000000 4084.0 nvrtc 2674 cache 3 dataset 28 hot 0 check 1173 race 0 ms variant base class v6 (state leaves 93, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 5aa17d87ef2c2856; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes) resident 2 programs 2 datasets
info ignored: noop
info switched to the prepared pair epoch edc4fa844da9dc98 day 69676e65756d2d6461792ffb50000000000000 (class v6) in 0.01 ms
done 21 4096 3.76
info dropped the previous pair (its program, cache and dataset)
done 22 4096 1.66
done 23 1920 0.65
done 24 21024 7.91
done 25 58176 85.12
done 26 16928 6.08
done 27 9216 10.24
done 28 61344 21.42
done 29 59360 91.22
done 30 33952 15.85
done 31 23968 41.30
done 32 29184 9.51
done 33 34944 13.40
done 34 2592 0.90
done 35 18816 15.78
done 36 55392 21.17
done 37 35904 12.97
done 38 6240 2.08
info transfers per chunk: down 5034 B (select, 534 full fallbacks); mean per chunk: kernel+sync 0.24 ms, select 0.038 ms, read-back 0.094 ms, scan 0.307 ms; 38 jobs 1039 chunks

19
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-01.fuzz.select.meta.json Normal file
View file

@ -0,0 +1,19 @@
{
"ctx": "same-work-20261008-01",
"object_id": "0x4de7b836cc40a4ea",
"shape": "fuzz",
"mode": "select",
"batch_log2": 10,
"jobs": 38,
"nonces": 1048576,
"worker": "/root/x2/bin/igneum-worker-cuda",
"worker_sha256": "4152c960d14210072a5c85ae675b56ae4dab12bedfb5f1af04ea2a2c6c025032",
"ready": "ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/edc4fa844da9dc98d37e965176f6558a31560e40502ab3ae5491b21aaaabfb07/day/69676e65756d2d6461792ffa50000000000000 dataset-log2 28 batch 1024 regs 93 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback select worker 1.0 (4 October 2026)",
"prepared": "prepared edc4fa844da9dc98d37e965176f6558a31560e40502ab3ae5491b21aaaabfb07 69676e65756d2d6461792ffb50000000000000 4084.0 nvrtc 2674 cache 3 dataset 28 hot 0 check 1173 race 0 ms variant base class v6 (state leaves 93, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 5aa17d87ef2c2856; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes) resident",
"transfers": "info transfers per chunk: down 5034 B (select, 534 full fallbacks); mean per chunk: kernel+sync 0.24 ms, select 0.038 ms, read-back 0.094 ms, scan 0.307 ms; 38 jobs 1039 chunks",
"errors": [],
"norm_sha256": "5f1e591db673f964ff53a60ebb9ea0e8f4cad974642a2354b3cc27571d935340",
"seconds": 9.6,
"exit": 0,
"at": "2026-10-09T08:47:14Z"
}

34
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-01.kf.compare.json Normal file
View file

@ -0,0 +1,34 @@
{
"ctx": "same-work-20261008-01",
"shape": "kf",
"expected_sha256": "e3e8913fb6788dfbbba4d43de14d674fa411de019980e2fc543be4ed7c17303c",
"modes": {
"fault": {
"sha256": "f7ad6ee87564f49abaee8e8ee64b62e41c6a45259c0cace233aecb898cbd78aa",
"equal_to_expected": false,
"first_difference": {
"line": 64,
"expected": "found 1 1053 0fa540bb2045b271",
"got": "found 1 1037 00cb3ac51a2ec48e"
},
"lines_missing": 0,
"lines_extra": 368
},
"full": {
"sha256": "e3e8913fb6788dfbbba4d43de14d674fa411de019980e2fc543be4ed7c17303c",
"equal_to_expected": true,
"first_difference": null,
"lines_missing": 0,
"lines_extra": 0
},
"select": {
"sha256": "e3e8913fb6788dfbbba4d43de14d674fa411de019980e2fc543be4ed7c17303c",
"equal_to_expected": true,
"first_difference": null,
"lines_missing": 0,
"lines_extra": 0
}
},
"known_failed_flagged": true,
"verdict": "EQUAL"
}

15
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-01.kf.expected.meta.json Normal file
View file

@ -0,0 +1,15 @@
{
"ctx": "same-work-20261008-01",
"shape": "kf",
"jobs": 4,
"nonces": 8192,
"found_lines": 533,
"labels": {
"kf": 4
},
"norm_sha256": "e3e8913fb6788dfbbba4d43de14d674fa411de019980e2fc543be4ed7c17303c",
"references": {
"D": "c49fed11d7279fee13c10a904ab21dd177c5ffad7974c049067ad72c23250b6f",
"D1": "5ae1220d24bed2677e0a9440f90393ae769da099d760dd90edbeef0bc864a914"
}
}

9
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-01.kf.fault.log Normal file
View file

@ -0,0 +1,9 @@
info igneum-worker-cuda 1.0 (4 October 2026): device 0 NVIDIA_GeForce_RTX_4090 (sm_89, 128 SMs), driver 13.2 from libcuda.so.1, NVRTC 12.8 from libnvrtc.so.12, target sm_89 (the device's architecture, listed by NVRTC)
ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/edc4fa844da9dc98d37e965176f6558a31560e40502ab3ae5491b21aaaabfb07/day/69676e65756d2d6461792ffa50000000000000 dataset-log2 28 batch 1024 regs 93 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback select worker 1.0 (4 October 2026)
info readback select (per dispatch of 1024 nonces: the hits and 34 sentinel words come back; the GPU scans)
info first pack /root/x2/ctx/same-work-20261008-01/packs/hl-v6-all: nvrtc 2586 cache 3 dataset 29 hot 0 check 1164 race 0 ms variant base class v6 (state leaves 93, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 448274a57f508cbc; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes)
done 1 2048 1.03
done 2 2048 0.92
done 3 2048 0.77
done 4 2048 0.74
info transfers per chunk: down 5470 B (select, 4 full fallbacks); mean per chunk: kernel+sync 0.24 ms, select 0.046 ms, read-back 0.111 ms, scan 0.028 ms; 4 jobs 8 chunks

19
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-01.kf.fault.meta.json Normal file
View file

@ -0,0 +1,19 @@
{
"ctx": "same-work-20261008-01",
"object_id": "0x4de7b836cc40a4ea",
"shape": "kf",
"mode": "fault",
"batch_log2": 10,
"jobs": 4,
"nonces": 8192,
"worker": "/root/x2/bin/igneum-worker-cuda",
"worker_sha256": "4152c960d14210072a5c85ae675b56ae4dab12bedfb5f1af04ea2a2c6c025032",
"ready": "ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/edc4fa844da9dc98d37e965176f6558a31560e40502ab3ae5491b21aaaabfb07/day/69676e65756d2d6461792ffa50000000000000 dataset-log2 28 batch 1024 regs 93 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback select worker 1.0 (4 October 2026)",
"prepared": null,
"transfers": "info transfers per chunk: down 5470 B (select, 4 full fallbacks); mean per chunk: kernel+sync 0.24 ms, select 0.046 ms, read-back 0.111 ms, scan 0.028 ms; 4 jobs 8 chunks",
"errors": [],
"norm_sha256": "f7ad6ee87564f49abaee8e8ee64b62e41c6a45259c0cace233aecb898cbd78aa",
"seconds": 4.2,
"exit": 0,
"at": "2026-10-09T08:45:59Z"
}

9
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-01.kf.full.log Normal file
View file

@ -0,0 +1,9 @@
info igneum-worker-cuda 1.0 (4 October 2026): device 0 NVIDIA_GeForce_RTX_4090 (sm_89, 128 SMs), driver 13.2 from libcuda.so.1, NVRTC 12.8 from libnvrtc.so.12, target sm_89 (the device's architecture, listed by NVRTC)
ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/edc4fa844da9dc98d37e965176f6558a31560e40502ab3ae5491b21aaaabfb07/day/69676e65756d2d6461792ffa50000000000000 dataset-log2 28 batch 1024 regs 93 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback full worker 1.0 (4 October 2026)
info readback full (per dispatch of 1024 nonces: 8 bytes per nonce come back and the host scans them)
info first pack /root/x2/ctx/same-work-20261008-01/packs/hl-v6-all: nvrtc 2724 cache 2 dataset 29 hot 0 check 1247 race 0 ms variant base class v6 (state leaves 93, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 448274a57f508cbc; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes)
done 1 2048 0.70
done 2 2048 0.56
done 3 2048 0.56
done 4 2048 0.56
info transfers per chunk: down 8192 B (full, 0 full fallbacks); mean per chunk: kernel+sync 0.23 ms, select 0.000 ms, read-back 0.044 ms, scan 0.017 ms; 4 jobs 8 chunks

19
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-01.kf.full.meta.json Normal file
View file

@ -0,0 +1,19 @@
{
"ctx": "same-work-20261008-01",
"object_id": "0x4de7b836cc40a4ea",
"shape": "kf",
"mode": "full",
"batch_log2": 10,
"jobs": 4,
"nonces": 8192,
"worker": "/root/x2/bin/igneum-worker-cuda",
"worker_sha256": "4152c960d14210072a5c85ae675b56ae4dab12bedfb5f1af04ea2a2c6c025032",
"ready": "ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/edc4fa844da9dc98d37e965176f6558a31560e40502ab3ae5491b21aaaabfb07/day/69676e65756d2d6461792ffa50000000000000 dataset-log2 28 batch 1024 regs 93 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback full worker 1.0 (4 October 2026)",
"prepared": null,
"transfers": "info transfers per chunk: down 8192 B (full, 0 full fallbacks); mean per chunk: kernel+sync 0.23 ms, select 0.000 ms, read-back 0.044 ms, scan 0.017 ms; 4 jobs 8 chunks",
"errors": [],
"norm_sha256": "e3e8913fb6788dfbbba4d43de14d674fa411de019980e2fc543be4ed7c17303c",
"seconds": 4.4,
"exit": 0,
"at": "2026-10-09T08:46:06Z"
}

9
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-01.kf.select.log Normal file
View file

@ -0,0 +1,9 @@
info igneum-worker-cuda 1.0 (4 October 2026): device 0 NVIDIA_GeForce_RTX_4090 (sm_89, 128 SMs), driver 13.2 from libcuda.so.1, NVRTC 12.8 from libnvrtc.so.12, target sm_89 (the device's architecture, listed by NVRTC)
ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/edc4fa844da9dc98d37e965176f6558a31560e40502ab3ae5491b21aaaabfb07/day/69676e65756d2d6461792ffa50000000000000 dataset-log2 28 batch 1024 regs 93 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback select worker 1.0 (4 October 2026)
info readback select (per dispatch of 1024 nonces: the hits and 34 sentinel words come back; the GPU scans)
info first pack /root/x2/ctx/same-work-20261008-01/packs/hl-v6-all: nvrtc 2630 cache 3 dataset 28 hot 0 check 1214 race 0 ms variant base class v6 (state leaves 93, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 448274a57f508cbc; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes)
done 1 2048 1.09
done 2 2048 1.73
done 3 2048 1.20
done 4 2048 0.75
info transfers per chunk: down 1342 B (select, 0 full fallbacks); mean per chunk: kernel+sync 0.26 ms, select 0.035 ms, read-back 0.223 ms, scan 0.020 ms; 4 jobs 8 chunks

19
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-01.kf.select.meta.json Normal file
View file

@ -0,0 +1,19 @@
{
"ctx": "same-work-20261008-01",
"object_id": "0x4de7b836cc40a4ea",
"shape": "kf",
"mode": "select",
"batch_log2": 10,
"jobs": 4,
"nonces": 8192,
"worker": "/root/x2/bin/igneum-worker-cuda",
"worker_sha256": "4152c960d14210072a5c85ae675b56ae4dab12bedfb5f1af04ea2a2c6c025032",
"ready": "ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/edc4fa844da9dc98d37e965176f6558a31560e40502ab3ae5491b21aaaabfb07/day/69676e65756d2d6461792ffa50000000000000 dataset-log2 28 batch 1024 regs 93 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback select worker 1.0 (4 October 2026)",
"prepared": null,
"transfers": "info transfers per chunk: down 1342 B (select, 0 full fallbacks); mean per chunk: kernel+sync 0.26 ms, select 0.035 ms, read-back 0.223 ms, scan 0.020 ms; 4 jobs 8 chunks",
"errors": [],
"norm_sha256": "e3e8913fb6788dfbbba4d43de14d674fa411de019980e2fc543be4ed7c17303c",
"seconds": 4.4,
"exit": 0,
"at": "2026-10-09T08:46:13Z"
}

24
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-01.prod.compare.json Normal file
View file

@ -0,0 +1,24 @@
{
"ctx": "same-work-20261008-01",
"shape": "prod",
"expected_sha256": "f278f0594dfd566d48fb59ee6b2f2889c60afb341856cf9efd53cce0d57ad60f",
"modes": {
"fault": "NOT RUN",
"full": {
"sha256": "f278f0594dfd566d48fb59ee6b2f2889c60afb341856cf9efd53cce0d57ad60f",
"equal_to_expected": true,
"first_difference": null,
"lines_missing": 0,
"lines_extra": 0
},
"select": {
"sha256": "f278f0594dfd566d48fb59ee6b2f2889c60afb341856cf9efd53cce0d57ad60f",
"equal_to_expected": true,
"first_difference": null,
"lines_missing": 0,
"lines_extra": 0
}
},
"known_failed_flagged": null,
"verdict": "EQUAL"
}

25
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-01.prod.expected.meta.json Normal file
View file

@ -0,0 +1,25 @@
{
"ctx": "same-work-20261008-01",
"shape": "prod",
"jobs": 16,
"nonces": 9437568,
"found_lines": 3146498,
"labels": {
"p1-all": 1,
"sep": 6,
"p1-t256": 1,
"p1-t257": 1,
"p1-zero": 1,
"p1-teq": 1,
"p1-teq-1": 1,
"p2a-all": 1,
"p2b-all": 1,
"p3-all": 1,
"p3-t256": 1
},
"norm_sha256": "f278f0594dfd566d48fb59ee6b2f2889c60afb341856cf9efd53cce0d57ad60f",
"references": {
"D": "c49fed11d7279fee13c10a904ab21dd177c5ffad7974c049067ad72c23250b6f",
"D1": "5ae1220d24bed2677e0a9440f90393ae769da099d760dd90edbeef0bc864a914"
}
}

39
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-01.prod.full.log Normal file
View file

@ -0,0 +1,39 @@
info igneum-worker-cuda 1.0 (4 October 2026): device 0 NVIDIA_GeForce_RTX_4090 (sm_89, 128 SMs), driver 13.2 from libcuda.so.1, NVRTC 12.8 from libnvrtc.so.12, target sm_89 (the device's architecture, listed by NVRTC)
ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/edc4fa844da9dc98d37e965176f6558a31560e40502ab3ae5491b21aaaabfb07/day/69676e65756d2d6461792ffa50000000000000 dataset-log2 28 batch 4194304 regs 93 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback full worker 1.0 (4 October 2026)
info readback full (per dispatch of 4194304 nonces: 8 bytes per nonce come back and the host scans them)
info first pack /root/x2/ctx/same-work-20261008-01/packs/hl-v6-all: nvrtc 2599 cache 2 dataset 29 hot 0 check 1173 race 0 ms variant base class v6 (state leaves 93, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 448274a57f508cbc; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes)
done 1 1048576 1591.17
done 2 64 0.44
done 3 1048576 32.92
done 4 64 0.31
done 5 1048576 32.30
done 6 64 0.25
done 7 1048576 31.86
done 8 64 0.25
done 9 1048576 32.85
done 10 64 0.29
done 11 1048576 32.81
done 12 524288 782.68
info prepare started for epoch edc4fa844da9dc98 day 69676e65756d2d6461792ffb50000000000000 from /root/x2/ctx/same-work-20261008-01/packs/hl-v6-all-d20731 (NVRTC sm_89 in the background)
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
prepared edc4fa844da9dc98d37e965176f6558a31560e40502ab3ae5491b21aaaabfb07 69676e65756d2d6461792ffb50000000000000 3969.6 nvrtc 2605 cache 2 dataset 29 hot 0 check 1168 race 0 ms variant base class v6 (state leaves 93, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 5aa17d87ef2c2856; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes) resident 2 programs 2 datasets
info ignored: noop
info switched to the prepared pair epoch edc4fa844da9dc98 day 69676e65756d2d6461792ffb50000000000000 (class v6) in 0.01 ms
done 13 524288 833.68
info dropped the previous pair (its program, cache and dataset)
done 14 1048576 1687.70
done 15 64 0.44
done 16 1048576 33.40
info transfers per chunk: down 4718784 B (full, 0 full fallbacks); mean per chunk: kernel+sync 17.25 ms, select 0.000 ms, read-back 0.987 ms, scan 300.041 ms; 16 jobs 16 chunks

19
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-01.prod.full.meta.json Normal file
View file

@ -0,0 +1,19 @@
{
"ctx": "same-work-20261008-01",
"object_id": "0x4de7b836cc40a4ea",
"shape": "prod",
"mode": "full",
"batch_log2": 22,
"jobs": 16,
"nonces": 9437568,
"worker": "/root/x2/bin/igneum-worker-cuda",
"worker_sha256": "4152c960d14210072a5c85ae675b56ae4dab12bedfb5f1af04ea2a2c6c025032",
"ready": "ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/edc4fa844da9dc98d37e965176f6558a31560e40502ab3ae5491b21aaaabfb07/day/69676e65756d2d6461792ffa50000000000000 dataset-log2 28 batch 4194304 regs 93 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback full worker 1.0 (4 October 2026)",
"prepared": "prepared edc4fa844da9dc98d37e965176f6558a31560e40502ab3ae5491b21aaaabfb07 69676e65756d2d6461792ffb50000000000000 3969.6 nvrtc 2605 cache 2 dataset 29 hot 0 check 1168 race 0 ms variant base class v6 (state leaves 93, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 5aa17d87ef2c2856; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes) resident",
"transfers": "info transfers per chunk: down 4718784 B (full, 0 full fallbacks); mean per chunk: kernel+sync 17.25 ms, select 0.000 ms, read-back 0.987 ms, scan 300.041 ms; 16 jobs 16 chunks",
"errors": [],
"norm_sha256": "f278f0594dfd566d48fb59ee6b2f2889c60afb341856cf9efd53cce0d57ad60f",
"seconds": 13.9,
"exit": 0,
"at": "2026-10-09T08:49:56Z"
}

40
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-01.prod.select.log Normal file
View file

@ -0,0 +1,40 @@
info igneum-worker-cuda 1.0 (4 October 2026): device 0 NVIDIA_GeForce_RTX_4090 (sm_89, 128 SMs), driver 13.2 from libcuda.so.1, NVRTC 12.8 from libnvrtc.so.12, target sm_89 (the device's architecture, listed by NVRTC)
ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/edc4fa844da9dc98d37e965176f6558a31560e40502ab3ae5491b21aaaabfb07/day/69676e65756d2d6461792ffa50000000000000 dataset-log2 28 batch 4194304 regs 93 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback select worker 1.0 (4 October 2026)
info readback select (per dispatch of 4194304 nonces: the hits and 34 sentinel words come back; the GPU scans)
info first pack /root/x2/ctx/same-work-20261008-01/packs/hl-v6-all: nvrtc 2602 cache 2 dataset 29 hot 0 check 1173 race 0 ms variant base class v6 (state leaves 93, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 448274a57f508cbc; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes)
done 1 1048576 1674.34
done 2 64 0.54
done 3 1048576 30.25
done 4 64 0.33
done 5 1048576 34.61
done 6 64 0.54
done 7 1048576 30.08
done 8 64 0.34
done 9 1048576 30.04
done 10 64 0.30
done 11 1048576 32.67
done 12 524288 804.42
info prepare started for epoch edc4fa844da9dc98 day 69676e65756d2d6461792ffb50000000000000 from /root/x2/ctx/same-work-20261008-01/packs/hl-v6-all-d20731 (NVRTC sm_89 in the background)
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
prepared edc4fa844da9dc98d37e965176f6558a31560e40502ab3ae5491b21aaaabfb07 69676e65756d2d6461792ffb50000000000000 4364.4 nvrtc 2656 cache 2 dataset 31 hot 0 check 1217 race 0 ms variant base class v6 (state leaves 93, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 5aa17d87ef2c2856; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes) resident 2 programs 2 datasets
info ignored: noop
info switched to the prepared pair epoch edc4fa844da9dc98 day 69676e65756d2d6461792ffb50000000000000 (class v6) in 0.01 ms
done 13 524288 843.60
info dropped the previous pair (its program, cache and dataset)
done 14 1048576 1705.49
done 15 64 0.54
done 16 1048576 30.24
info transfers per chunk: down 2097856 B (select, 5 full fallbacks); mean per chunk: kernel+sync 17.14 ms, select 0.245 ms, read-back 0.582 ms, scan 308.152 ms; 16 jobs 16 chunks

19
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-01.prod.select.meta.json Normal file
View file

@ -0,0 +1,19 @@
{
"ctx": "same-work-20261008-01",
"object_id": "0x4de7b836cc40a4ea",
"shape": "prod",
"mode": "select",
"batch_log2": 22,
"jobs": 16,
"nonces": 9437568,
"worker": "/root/x2/bin/igneum-worker-cuda",
"worker_sha256": "4152c960d14210072a5c85ae675b56ae4dab12bedfb5f1af04ea2a2c6c025032",
"ready": "ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/edc4fa844da9dc98d37e965176f6558a31560e40502ab3ae5491b21aaaabfb07/day/69676e65756d2d6461792ffa50000000000000 dataset-log2 28 batch 4194304 regs 93 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback select worker 1.0 (4 October 2026)",
"prepared": "prepared edc4fa844da9dc98d37e965176f6558a31560e40502ab3ae5491b21aaaabfb07 69676e65756d2d6461792ffb50000000000000 4364.4 nvrtc 2656 cache 2 dataset 31 hot 0 check 1217 race 0 ms variant base class v6 (state leaves 93, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 5aa17d87ef2c2856; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes) resident",
"transfers": "info transfers per chunk: down 2097856 B (select, 5 full fallbacks); mean per chunk: kernel+sync 17.14 ms, select 0.245 ms, read-back 0.582 ms, scan 308.152 ms; 16 jobs 16 chunks",
"errors": [],
"norm_sha256": "f278f0594dfd566d48fb59ee6b2f2889c60afb341856cf9efd53cce0d57ad60f",
"seconds": 14.5,
"exit": 0,
"at": "2026-10-09T08:50:16Z"
}

24
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-02.fuzz.compare.json Normal file
View file

@ -0,0 +1,24 @@
{
"ctx": "same-work-20261008-02",
"shape": "fuzz",
"expected_sha256": "8f39a1b15b16e86f09bedeaa9d2f811cd6329012a71139e93babd5199f468938",
"modes": {
"fault": "NOT RUN",
"full": {
"sha256": "8f39a1b15b16e86f09bedeaa9d2f811cd6329012a71139e93babd5199f468938",
"equal_to_expected": true,
"first_difference": null,
"lines_missing": 0,
"lines_extra": 0
},
"select": {
"sha256": "8f39a1b15b16e86f09bedeaa9d2f811cd6329012a71139e93babd5199f468938",
"equal_to_expected": true,
"first_difference": null,
"lines_missing": 0,
"lines_extra": 0
}
},
"known_failed_flagged": null,
"verdict": "EQUAL"
}

26
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-02.fuzz.expected.meta.json Normal file
View file

@ -0,0 +1,26 @@
{
"ctx": "same-work-20261008-02",
"shape": "fuzz",
"jobs": 35,
"nonces": 1048576,
"found_lines": 446422,
"labels": {
"fz5": 5,
"fz7": 4,
"fz2": 7,
"fz0": 3,
"fz3": 4,
"fz1": 3,
"fz6": 4,
"blk-d-all": 1,
"blk-d-q4": 1,
"blk-d1-all": 1,
"blk-d1-q4": 1,
"fz4": 1
},
"norm_sha256": "8f39a1b15b16e86f09bedeaa9d2f811cd6329012a71139e93babd5199f468938",
"references": {
"D": "7c7be9b549b191ffd628aa4722c6222e1752eac51d674bc0eefe518a67cedf48",
"D1": "cad7849f8df4e7d8faa6c0ffdb5561e90fbd1274c465cfd1efc61d8015122753"
}
}

54
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-02.fuzz.full.log Normal file
View file

@ -0,0 +1,54 @@
info igneum-worker-cuda 1.0 (4 October 2026): device 0 NVIDIA_GeForce_RTX_4090 (sm_89, 128 SMs), driver 13.2 from libcuda.so.1, NVRTC 12.8 from libnvrtc.so.12, target sm_89 (the device's architecture, listed by NVRTC)
ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/af89be5ddbadb6f6b4aee28ac8f249713be5d4c12621e3cea7f83ceada3c66b3/day/69676e65756d2d6461792ffa50000000000000 dataset-log2 28 batch 1024 regs 88 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback full worker 1.0 (4 October 2026)
info readback full (per dispatch of 1024 nonces: 8 bytes per nonce come back and the host scans them)
info first pack /root/x2/ctx/same-work-20261008-02/packs/chain-seed-d20730: nvrtc 2507 cache 3 dataset 28 hot 0 check 1092 race 0 ms variant base class v6 (state leaves 93, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 448274a57f508cbc; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes)
done 1 43968 79.40
done 2 9504 8.47
done 3 4992 2.14
done 4 51392 16.39
done 5 16768 6.58
done 6 40832 13.06
done 7 23744 9.21
done 8 53696 91.23
done 9 38144 12.11
done 10 416 0.35
done 11 36960 12.58
done 12 35072 51.13
done 13 3584 1.34
done 14 35136 28.32
done 15 47552 18.04
done 16 61856 36.36
done 17 4096 5.22
done 18 4096 1.87
info prepare started for epoch af89be5ddbadb6f6 day 69676e65756d2d6461792ffb50000000000000 from /root/x2/ctx/same-work-20261008-02/packs/chain-seed-d20731 (NVRTC sm_89 in the background)
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
prepared af89be5ddbadb6f6b4aee28ac8f249713be5d4c12621e3cea7f83ceada3c66b3 69676e65756d2d6461792ffb50000000000000 4106.3 nvrtc 2514 cache 3 dataset 28 hot 0 check 1104 race 0 ms variant base class v6 (state leaves 93, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 5aa17d87ef2c2856; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes) resident 2 programs 2 datasets
info ignored: noop
info switched to the prepared pair epoch af89be5ddbadb6f6 day 69676e65756d2d6461792ffb50000000000000 (class v6) in 0.01 ms
done 19 4096 7.39
info dropped the previous pair (its program, cache and dataset)
done 20 4096 1.61
done 21 4192 8.44
done 22 26848 12.92
done 23 33280 47.04
done 24 34112 31.90
done 25 48256 14.86
done 26 12928 4.03
done 27 59072 38.37
done 28 33984 50.65
done 29 28928 8.99
done 30 31808 44.69
done 31 37088 48.66
done 32 46976 52.81
done 33 40576 17.38
done 34 31360 11.96
done 35 59168 29.30
info transfers per chunk: down 8050 B (full, 0 full fallbacks); mean per chunk: kernel+sync 0.25 ms, select 0.000 ms, read-back 0.075 ms, scan 0.382 ms; 35 jobs 1042 chunks

19
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-02.fuzz.full.meta.json Normal file
View file

@ -0,0 +1,19 @@
{
"ctx": "same-work-20261008-02",
"object_id": "0x2a1d6caab4c24564",
"shape": "fuzz",
"mode": "full",
"batch_log2": 10,
"jobs": 35,
"nonces": 1048576,
"worker": "/root/x2/bin/igneum-worker-cuda",
"worker_sha256": "4152c960d14210072a5c85ae675b56ae4dab12bedfb5f1af04ea2a2c6c025032",
"ready": "ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/af89be5ddbadb6f6b4aee28ac8f249713be5d4c12621e3cea7f83ceada3c66b3/day/69676e65756d2d6461792ffa50000000000000 dataset-log2 28 batch 1024 regs 88 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback full worker 1.0 (4 October 2026)",
"prepared": "prepared af89be5ddbadb6f6b4aee28ac8f249713be5d4c12621e3cea7f83ceada3c66b3 69676e65756d2d6461792ffb50000000000000 4106.3 nvrtc 2514 cache 3 dataset 28 hot 0 check 1104 race 0 ms variant base class v6 (state leaves 93, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 5aa17d87ef2c2856; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes) resident",
"transfers": "info transfers per chunk: down 8050 B (full, 0 full fallbacks); mean per chunk: kernel+sync 0.25 ms, select 0.000 ms, read-back 0.075 ms, scan 0.382 ms; 35 jobs 1042 chunks",
"errors": [],
"norm_sha256": "8f39a1b15b16e86f09bedeaa9d2f811cd6329012a71139e93babd5199f468938",
"seconds": 9.5,
"exit": 0,
"at": "2026-10-09T08:48:03Z"
}

54
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-02.fuzz.select.log Normal file
View file

@ -0,0 +1,54 @@
info igneum-worker-cuda 1.0 (4 October 2026): device 0 NVIDIA_GeForce_RTX_4090 (sm_89, 128 SMs), driver 13.2 from libcuda.so.1, NVRTC 12.8 from libnvrtc.so.12, target sm_89 (the device's architecture, listed by NVRTC)
ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/af89be5ddbadb6f6b4aee28ac8f249713be5d4c12621e3cea7f83ceada3c66b3/day/69676e65756d2d6461792ffa50000000000000 dataset-log2 28 batch 1024 regs 88 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback select worker 1.0 (4 October 2026)
info readback select (per dispatch of 1024 nonces: the hits and 34 sentinel words come back; the GPU scans)
info first pack /root/x2/ctx/same-work-20261008-02/packs/chain-seed-d20730: nvrtc 2549 cache 3 dataset 29 hot 0 check 1124 race 0 ms variant base class v6 (state leaves 93, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 448274a57f508cbc; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes)
done 1 43968 41.51
done 2 9504 6.78
done 3 4992 2.68
done 4 51392 19.25
done 5 16768 8.28
done 6 40832 18.19
done 7 23744 14.37
done 8 53696 94.11
done 9 38144 12.32
done 10 416 0.36
done 11 36960 17.40
done 12 35072 51.94
done 13 3584 1.56
done 14 35136 28.13
done 15 47552 25.31
done 16 61856 36.94
done 17 4096 3.93
done 18 4096 2.86
info prepare started for epoch af89be5ddbadb6f6 day 69676e65756d2d6461792ffb50000000000000 from /root/x2/ctx/same-work-20261008-02/packs/chain-seed-d20731 (NVRTC sm_89 in the background)
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
prepared af89be5ddbadb6f6b4aee28ac8f249713be5d4c12621e3cea7f83ceada3c66b3 69676e65756d2d6461792ffb50000000000000 4114.7 nvrtc 2588 cache 3 dataset 28 hot 0 check 1164 race 0 ms variant base class v6 (state leaves 93, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 5aa17d87ef2c2856; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes) resident 2 programs 2 datasets
info ignored: noop
info switched to the prepared pair epoch af89be5ddbadb6f6 day 69676e65756d2d6461792ffb50000000000000 (class v6) in 0.01 ms
done 19 4096 7.67
info dropped the previous pair (its program, cache and dataset)
done 20 4096 3.06
done 21 4192 6.32
done 22 26848 17.62
done 23 33280 40.76
done 24 34112 36.23
done 25 48256 20.21
done 26 12928 5.68
done 27 59072 64.66
done 28 33984 47.38
done 29 28928 10.45
done 30 31808 42.06
done 31 37088 50.72
done 32 46976 56.77
done 33 40576 22.85
done 34 31360 13.20
done 35 59168 30.04
info transfers per chunk: down 5237 B (select, 599 full fallbacks); mean per chunk: kernel+sync 0.25 ms, select 0.053 ms, read-back 0.120 ms, scan 0.328 ms; 35 jobs 1042 chunks

19
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-02.fuzz.select.meta.json Normal file
View file

@ -0,0 +1,19 @@
{
"ctx": "same-work-20261008-02",
"object_id": "0x2a1d6caab4c24564",
"shape": "fuzz",
"mode": "select",
"batch_log2": 10,
"jobs": 35,
"nonces": 1048576,
"worker": "/root/x2/bin/igneum-worker-cuda",
"worker_sha256": "4152c960d14210072a5c85ae675b56ae4dab12bedfb5f1af04ea2a2c6c025032",
"ready": "ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/af89be5ddbadb6f6b4aee28ac8f249713be5d4c12621e3cea7f83ceada3c66b3/day/69676e65756d2d6461792ffa50000000000000 dataset-log2 28 batch 1024 regs 88 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback select worker 1.0 (4 October 2026)",
"prepared": "prepared af89be5ddbadb6f6b4aee28ac8f249713be5d4c12621e3cea7f83ceada3c66b3 69676e65756d2d6461792ffb50000000000000 4114.7 nvrtc 2588 cache 3 dataset 28 hot 0 check 1164 race 0 ms variant base class v6 (state leaves 93, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 5aa17d87ef2c2856; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes) resident",
"transfers": "info transfers per chunk: down 5237 B (select, 599 full fallbacks); mean per chunk: kernel+sync 0.25 ms, select 0.053 ms, read-back 0.120 ms, scan 0.328 ms; 35 jobs 1042 chunks",
"errors": [],
"norm_sha256": "8f39a1b15b16e86f09bedeaa9d2f811cd6329012a71139e93babd5199f468938",
"seconds": 9.8,
"exit": 0,
"at": "2026-10-09T08:48:15Z"
}

24
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-02.prod.compare.json Normal file
View file

@ -0,0 +1,24 @@
{
"ctx": "same-work-20261008-02",
"shape": "prod",
"expected_sha256": "ad091a0d49a9d3a2b6206873ab5c9639d62e76cb5e65006aa730e3444e0a24d7",
"modes": {
"fault": "NOT RUN",
"full": {
"sha256": "ad091a0d49a9d3a2b6206873ab5c9639d62e76cb5e65006aa730e3444e0a24d7",
"equal_to_expected": true,
"first_difference": null,
"lines_missing": 0,
"lines_extra": 0
},
"select": {
"sha256": "ad091a0d49a9d3a2b6206873ab5c9639d62e76cb5e65006aa730e3444e0a24d7",
"equal_to_expected": true,
"first_difference": null,
"lines_missing": 0,
"lines_extra": 0
}
},
"known_failed_flagged": null,
"verdict": "EQUAL"
}

25
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-02.prod.expected.meta.json Normal file
View file

@ -0,0 +1,25 @@
{
"ctx": "same-work-20261008-02",
"shape": "prod",
"jobs": 16,
"nonces": 9437568,
"found_lines": 3146498,
"labels": {
"p1-all": 1,
"sep": 6,
"p1-t256": 1,
"p1-t257": 1,
"p1-zero": 1,
"p1-teq": 1,
"p1-teq-1": 1,
"p2a-all": 1,
"p2b-all": 1,
"p3-all": 1,
"p3-t256": 1
},
"norm_sha256": "ad091a0d49a9d3a2b6206873ab5c9639d62e76cb5e65006aa730e3444e0a24d7",
"references": {
"D": "7c7be9b549b191ffd628aa4722c6222e1752eac51d674bc0eefe518a67cedf48",
"D1": "cad7849f8df4e7d8faa6c0ffdb5561e90fbd1274c465cfd1efc61d8015122753"
}
}

39
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-02.prod.full.log Normal file
View file

@ -0,0 +1,39 @@
info igneum-worker-cuda 1.0 (4 October 2026): device 0 NVIDIA_GeForce_RTX_4090 (sm_89, 128 SMs), driver 13.2 from libcuda.so.1, NVRTC 12.8 from libnvrtc.so.12, target sm_89 (the device's architecture, listed by NVRTC)
ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/af89be5ddbadb6f6b4aee28ac8f249713be5d4c12621e3cea7f83ceada3c66b3/day/69676e65756d2d6461792ffa50000000000000 dataset-log2 28 batch 4194304 regs 88 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback full worker 1.0 (4 October 2026)
info readback full (per dispatch of 4194304 nonces: 8 bytes per nonce come back and the host scans them)
info first pack /root/x2/ctx/same-work-20261008-02/packs/chain-seed-d20730: nvrtc 2574 cache 3 dataset 28 hot 0 check 1150 race 0 ms variant base class v6 (state leaves 93, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 448274a57f508cbc; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes)
done 1 1048576 1490.90
done 2 64 0.44
done 3 1048576 33.10
done 4 64 0.30
done 5 1048576 33.06
done 6 64 0.28
done 7 1048576 32.85
done 8 64 0.26
done 9 1048576 34.79
done 10 64 0.36
done 11 1048576 34.14
done 12 524288 772.89
info prepare started for epoch af89be5ddbadb6f6 day 69676e65756d2d6461792ffb50000000000000 from /root/x2/ctx/same-work-20261008-02/packs/chain-seed-d20731 (NVRTC sm_89 in the background)
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
prepared af89be5ddbadb6f6b4aee28ac8f249713be5d4c12621e3cea7f83ceada3c66b3 69676e65756d2d6461792ffb50000000000000 4073.6 nvrtc 2571 cache 3 dataset 28 hot 0 check 1160 race 0 ms variant base class v6 (state leaves 93, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 5aa17d87ef2c2856; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes) resident 2 programs 2 datasets
info ignored: noop
info switched to the prepared pair epoch af89be5ddbadb6f6 day 69676e65756d2d6461792ffb50000000000000 (class v6) in 0.00 ms
done 13 524288 774.37
info dropped the previous pair (its program, cache and dataset)
done 14 1048576 1450.11
done 15 64 0.43
done 16 1048576 34.46
info transfers per chunk: down 4718784 B (full, 0 full fallbacks); mean per chunk: kernel+sync 17.12 ms, select 0.000 ms, read-back 1.537 ms, scan 274.584 ms; 16 jobs 16 chunks

19
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-02.prod.full.meta.json Normal file
View file

@ -0,0 +1,19 @@
{
"ctx": "same-work-20261008-02",
"object_id": "0x2a1d6caab4c24564",
"shape": "prod",
"mode": "full",
"batch_log2": 22,
"jobs": 16,
"nonces": 9437568,
"worker": "/root/x2/bin/igneum-worker-cuda",
"worker_sha256": "4152c960d14210072a5c85ae675b56ae4dab12bedfb5f1af04ea2a2c6c025032",
"ready": "ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/af89be5ddbadb6f6b4aee28ac8f249713be5d4c12621e3cea7f83ceada3c66b3/day/69676e65756d2d6461792ffa50000000000000 dataset-log2 28 batch 4194304 regs 88 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback full worker 1.0 (4 October 2026)",
"prepared": "prepared af89be5ddbadb6f6b4aee28ac8f249713be5d4c12621e3cea7f83ceada3c66b3 69676e65756d2d6461792ffb50000000000000 4073.6 nvrtc 2571 cache 3 dataset 28 hot 0 check 1160 race 0 ms variant base class v6 (state leaves 93, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 5aa17d87ef2c2856; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes) resident",
"transfers": "info transfers per chunk: down 4718784 B (full, 0 full fallbacks); mean per chunk: kernel+sync 17.12 ms, select 0.000 ms, read-back 1.537 ms, scan 274.584 ms; 16 jobs 16 chunks",
"errors": [],
"norm_sha256": "ad091a0d49a9d3a2b6206873ab5c9639d62e76cb5e65006aa730e3444e0a24d7",
"seconds": 13.6,
"exit": 0,
"at": "2026-10-09T08:50:44Z"
}

39
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-02.prod.select.log Normal file
View file

@ -0,0 +1,39 @@
info igneum-worker-cuda 1.0 (4 October 2026): device 0 NVIDIA_GeForce_RTX_4090 (sm_89, 128 SMs), driver 13.2 from libcuda.so.1, NVRTC 12.8 from libnvrtc.so.12, target sm_89 (the device's architecture, listed by NVRTC)
ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/af89be5ddbadb6f6b4aee28ac8f249713be5d4c12621e3cea7f83ceada3c66b3/day/69676e65756d2d6461792ffa50000000000000 dataset-log2 28 batch 4194304 regs 88 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback select worker 1.0 (4 October 2026)
info readback select (per dispatch of 4194304 nonces: the hits and 34 sentinel words come back; the GPU scans)
info first pack /root/x2/ctx/same-work-20261008-02/packs/chain-seed-d20730: nvrtc 2580 cache 3 dataset 28 hot 0 check 1182 race 0 ms variant base class v6 (state leaves 93, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 448274a57f508cbc; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes)
done 1 1048576 1443.79
done 2 64 0.65
done 3 1048576 30.51
done 4 64 0.55
done 5 1048576 35.36
done 6 64 0.58
done 7 1048576 30.29
done 8 64 0.43
done 9 1048576 30.34
done 10 64 0.63
done 11 1048576 30.18
done 12 524288 765.61
info prepare started for epoch af89be5ddbadb6f6 day 69676e65756d2d6461792ffb50000000000000 from /root/x2/ctx/same-work-20261008-02/packs/chain-seed-d20731 (NVRTC sm_89 in the background)
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
prepared af89be5ddbadb6f6b4aee28ac8f249713be5d4c12621e3cea7f83ceada3c66b3 69676e65756d2d6461792ffb50000000000000 4133.5 nvrtc 2530 cache 3 dataset 28 hot 0 check 1103 race 0 ms variant base class v6 (state leaves 93, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 5aa17d87ef2c2856; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes) resident 2 programs 2 datasets
info ignored: noop
info switched to the prepared pair epoch af89be5ddbadb6f6 day 69676e65756d2d6461792ffb50000000000000 (class v6) in 0.02 ms
done 13 524288 865.89
info dropped the previous pair (its program, cache and dataset)
done 14 1048576 1426.68
done 15 64 0.57
done 16 1048576 30.36
info transfers per chunk: down 2097856 B (select, 5 full fallbacks); mean per chunk: kernel+sync 17.16 ms, select 0.094 ms, read-back 0.918 ms, scan 275.030 ms; 16 jobs 16 chunks

19
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-02.prod.select.meta.json Normal file
View file

@ -0,0 +1,19 @@
{
"ctx": "same-work-20261008-02",
"object_id": "0x2a1d6caab4c24564",
"shape": "prod",
"mode": "select",
"batch_log2": 22,
"jobs": 16,
"nonces": 9437568,
"worker": "/root/x2/bin/igneum-worker-cuda",
"worker_sha256": "4152c960d14210072a5c85ae675b56ae4dab12bedfb5f1af04ea2a2c6c025032",
"ready": "ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/af89be5ddbadb6f6b4aee28ac8f249713be5d4c12621e3cea7f83ceada3c66b3/day/69676e65756d2d6461792ffa50000000000000 dataset-log2 28 batch 4194304 regs 88 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback select worker 1.0 (4 October 2026)",
"prepared": "prepared af89be5ddbadb6f6b4aee28ac8f249713be5d4c12621e3cea7f83ceada3c66b3 69676e65756d2d6461792ffb50000000000000 4133.5 nvrtc 2530 cache 3 dataset 28 hot 0 check 1103 race 0 ms variant base class v6 (state leaves 93, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 5aa17d87ef2c2856; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes) resident",
"transfers": "info transfers per chunk: down 2097856 B (select, 5 full fallbacks); mean per chunk: kernel+sync 17.16 ms, select 0.094 ms, read-back 0.918 ms, scan 275.030 ms; 16 jobs 16 chunks",
"errors": [],
"norm_sha256": "ad091a0d49a9d3a2b6206873ab5c9639d62e76cb5e65006aa730e3444e0a24d7",
"seconds": 13.7,
"exit": 0,
"at": "2026-10-09T08:51:02Z"
}

24
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-03.fuzz.compare.json Normal file
View file

@ -0,0 +1,24 @@
{
"ctx": "same-work-20261008-03",
"shape": "fuzz",
"expected_sha256": "3a7a7cc495e2d4853f5c85b59309ccee637ae5f9d446b806c577ca904f54785f",
"modes": {
"fault": "NOT RUN",
"full": {
"sha256": "3a7a7cc495e2d4853f5c85b59309ccee637ae5f9d446b806c577ca904f54785f",
"equal_to_expected": true,
"first_difference": null,
"lines_missing": 0,
"lines_extra": 0
},
"select": {
"sha256": "3a7a7cc495e2d4853f5c85b59309ccee637ae5f9d446b806c577ca904f54785f",
"equal_to_expected": true,
"first_difference": null,
"lines_missing": 0,
"lines_extra": 0
}
},
"known_failed_flagged": null,
"verdict": "EQUAL"
}

26
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-03.fuzz.expected.meta.json Normal file
View file

@ -0,0 +1,26 @@
{
"ctx": "same-work-20261008-03",
"shape": "fuzz",
"jobs": 33,
"nonces": 1048576,
"found_lines": 342786,
"labels": {
"fz5": 2,
"fz7": 4,
"fz6": 4,
"fz1": 3,
"fz2": 6,
"fz4": 3,
"fz3": 6,
"blk-d-all": 1,
"blk-d-q4": 1,
"blk-d1-all": 1,
"blk-d1-q4": 1,
"fz0": 1
},
"norm_sha256": "3a7a7cc495e2d4853f5c85b59309ccee637ae5f9d446b806c577ca904f54785f",
"references": {
"D": "873528b18ad679ca9a748d52d45c4e1b6723ba703875742fb01e1cbfca169783",
"D1": "7f8e64ae899ed5452dab6a01a2b34d66f5a11ef1ca8c65a9b93ff0be2a88250c"
}
}

48
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-03.fuzz.full.log Normal file
View file

@ -0,0 +1,48 @@
info igneum-worker-cuda 1.0 (4 October 2026): device 0 NVIDIA_GeForce_RTX_4090 (sm_89, 128 SMs), driver 13.2 from libcuda.so.1, NVRTC 12.8 from libnvrtc.so.12, target sm_89 (the device's architecture, listed by NVRTC)
ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/3c3fab434f7fd894ce6db0b09f298c3865726253e4fc68fd77acd163ccea8221/day/69676e65756d2d6461792ffe50000000000000 dataset-log2 28 batch 1024 regs 31 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback full worker 1.0 (4 October 2026)
info readback full (per dispatch of 1024 nonces: 8 bytes per nonce come back and the host scans them)
info first pack /root/x2/ctx/same-work-20261008-03/packs/v5-epoch2-d20734: nvrtc 607 cache 2 dataset 28 hot 0 check 1183 race 0 ms variant base class v5 (state leaves 224, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 5cc75af0f884bd4c; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes)
done 1 60160 97.35
done 2 16160 14.42
done 3 51232 56.52
done 4 48800 63.28
done 5 49408 63.89
done 6 34624 6.78
done 7 13280 4.12
done 8 10208 2.32
done 9 22336 5.70
done 10 5440 1.70
done 11 18560 22.58
done 12 45024 14.56
done 13 52800 10.37
done 14 30432 6.28
done 15 38784 9.86
done 16 38784 42.85
done 17 50240 10.23
done 18 63392 16.53
done 19 4096 3.03
done 20 4096 1.83
info prepare started for epoch 3c3fab434f7fd894 day 69676e65756d2d6461792fff50000000000000 from /root/x2/ctx/same-work-20261008-03/packs/v5-epoch2-d20735 (NVRTC sm_89 in the background)
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
prepared 3c3fab434f7fd894ce6db0b09f298c3865726253e4fc68fd77acd163ccea8221 69676e65756d2d6461792fff50000000000000 2045.2 nvrtc 607 cache 3 dataset 28 hot 0 check 1111 race 0 ms variant base class v5 (state leaves 224, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 a51d063d7dfa091b; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes) resident 2 programs 2 datasets
info ignored: noop
info switched to the prepared pair epoch 3c3fab434f7fd894 day 69676e65756d2d6461792fff50000000000000 (class v5) in 0.01 ms
done 21 4096 6.87
info dropped the previous pair (its program, cache and dataset)
done 22 4096 1.69
done 23 1408 0.63
done 24 13440 8.12
done 25 34752 7.33
done 26 51936 15.93
done 27 36768 13.74
done 28 55040 10.13
done 29 41952 26.31
done 30 45184 9.64
done 31 448 0.37
done 32 44544 41.95
done 33 57056 12.77
info transfers per chunk: down 8105 B (full, 0 full fallbacks); mean per chunk: kernel+sync 0.15 ms, select 0.000 ms, read-back 0.047 ms, scan 0.325 ms; 33 jobs 1035 chunks

19
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-03.fuzz.full.meta.json Normal file
View file

@ -0,0 +1,19 @@
{
"ctx": "same-work-20261008-03",
"object_id": "0x81fbfa3aa413173f",
"shape": "fuzz",
"mode": "full",
"batch_log2": 10,
"jobs": 33,
"nonces": 1048576,
"worker": "/root/x2/bin/igneum-worker-cuda",
"worker_sha256": "4152c960d14210072a5c85ae675b56ae4dab12bedfb5f1af04ea2a2c6c025032",
"ready": "ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/3c3fab434f7fd894ce6db0b09f298c3865726253e4fc68fd77acd163ccea8221/day/69676e65756d2d6461792ffe50000000000000 dataset-log2 28 batch 1024 regs 31 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback full worker 1.0 (4 October 2026)",
"prepared": "prepared 3c3fab434f7fd894ce6db0b09f298c3865726253e4fc68fd77acd163ccea8221 69676e65756d2d6461792fff50000000000000 2045.2 nvrtc 607 cache 3 dataset 28 hot 0 check 1111 race 0 ms variant base class v5 (state leaves 224, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 a51d063d7dfa091b; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes) resident",
"transfers": "info transfers per chunk: down 8105 B (full, 0 full fallbacks); mean per chunk: kernel+sync 0.15 ms, select 0.000 ms, read-back 0.047 ms, scan 0.325 ms; 33 jobs 1035 chunks",
"errors": [],
"norm_sha256": "3a7a7cc495e2d4853f5c85b59309ccee637ae5f9d446b806c577ca904f54785f",
"seconds": 5.4,
"exit": 0,
"at": "2026-10-09T08:48:56Z"
}

48
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-03.fuzz.select.log Normal file
View file

@ -0,0 +1,48 @@
info igneum-worker-cuda 1.0 (4 October 2026): device 0 NVIDIA_GeForce_RTX_4090 (sm_89, 128 SMs), driver 13.2 from libcuda.so.1, NVRTC 12.8 from libnvrtc.so.12, target sm_89 (the device's architecture, listed by NVRTC)
ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/3c3fab434f7fd894ce6db0b09f298c3865726253e4fc68fd77acd163ccea8221/day/69676e65756d2d6461792ffe50000000000000 dataset-log2 28 batch 1024 regs 31 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback select worker 1.0 (4 October 2026)
info readback select (per dispatch of 1024 nonces: the hits and 34 sentinel words come back; the GPU scans)
info first pack /root/x2/ctx/same-work-20261008-03/packs/v5-epoch2-d20734: nvrtc 608 cache 2 dataset 29 hot 0 check 1204 race 0 ms variant base class v5 (state leaves 224, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 5cc75af0f884bd4c; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes)
done 1 60160 90.90
done 2 16160 14.55
done 3 51232 56.55
done 4 48800 62.53
done 5 49408 65.32
done 6 34624 13.99
done 7 13280 7.55
done 8 10208 4.88
done 9 22336 11.82
done 10 5440 3.52
done 11 18560 27.19
done 12 45024 25.93
done 13 52800 19.16
done 14 30432 14.12
done 15 38784 22.42
done 16 38784 66.31
done 17 50240 18.38
done 18 63392 33.77
done 19 4096 5.04
done 20 4096 3.07
info prepare started for epoch 3c3fab434f7fd894 day 69676e65756d2d6461792fff50000000000000 from /root/x2/ctx/same-work-20261008-03/packs/v5-epoch2-d20735 (NVRTC sm_89 in the background)
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
prepared 3c3fab434f7fd894ce6db0b09f298c3865726253e4fc68fd77acd163ccea8221 69676e65756d2d6461792fff50000000000000 1932.7 nvrtc 615 cache 2 dataset 28 hot 0 check 1116 race 0 ms variant base class v5 (state leaves 224, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 a51d063d7dfa091b; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes) resident 2 programs 2 datasets
info ignored: noop
info switched to the prepared pair epoch 3c3fab434f7fd894 day 69676e65756d2d6461792fff50000000000000 (class v5) in 0.01 ms
done 21 4096 7.86
info dropped the previous pair (its program, cache and dataset)
done 22 4096 2.95
done 23 1408 1.20
done 24 13440 8.04
done 25 34752 13.08
done 26 51936 26.96
done 27 36768 19.70
done 28 55040 22.25
done 29 41952 27.18
done 30 45184 22.00
done 31 448 0.56
done 32 44544 52.57
done 33 57056 29.01
info transfers per chunk: down 3909 B (select, 365 full fallbacks); mean per chunk: kernel+sync 0.17 ms, select 0.073 ms, read-back 0.182 ms, scan 0.295 ms; 33 jobs 1035 chunks

19
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-03.fuzz.select.meta.json Normal file
View file

@ -0,0 +1,19 @@
{
"ctx": "same-work-20261008-03",
"object_id": "0x81fbfa3aa413173f",
"shape": "fuzz",
"mode": "select",
"batch_log2": 10,
"jobs": 33,
"nonces": 1048576,
"worker": "/root/x2/bin/igneum-worker-cuda",
"worker_sha256": "4152c960d14210072a5c85ae675b56ae4dab12bedfb5f1af04ea2a2c6c025032",
"ready": "ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/3c3fab434f7fd894ce6db0b09f298c3865726253e4fc68fd77acd163ccea8221/day/69676e65756d2d6461792ffe50000000000000 dataset-log2 28 batch 1024 regs 31 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback select worker 1.0 (4 October 2026)",
"prepared": "prepared 3c3fab434f7fd894ce6db0b09f298c3865726253e4fc68fd77acd163ccea8221 69676e65756d2d6461792fff50000000000000 1932.7 nvrtc 615 cache 2 dataset 28 hot 0 check 1116 race 0 ms variant base class v5 (state leaves 224, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 a51d063d7dfa091b; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes) resident",
"transfers": "info transfers per chunk: down 3909 B (select, 365 full fallbacks); mean per chunk: kernel+sync 0.17 ms, select 0.073 ms, read-back 0.182 ms, scan 0.295 ms; 33 jobs 1035 chunks",
"errors": [],
"norm_sha256": "3a7a7cc495e2d4853f5c85b59309ccee637ae5f9d446b806c577ca904f54785f",
"seconds": 5.6,
"exit": 0,
"at": "2026-10-09T08:49:04Z"
}

24
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-03.prod.compare.json Normal file
View file

@ -0,0 +1,24 @@
{
"ctx": "same-work-20261008-03",
"shape": "prod",
"expected_sha256": "2aa97d5d63e5e01cb9fafc2d2508ad5df1c576467dfc46cb72b714fea35c93a5",
"modes": {
"fault": "NOT RUN",
"full": {
"sha256": "2aa97d5d63e5e01cb9fafc2d2508ad5df1c576467dfc46cb72b714fea35c93a5",
"equal_to_expected": true,
"first_difference": null,
"lines_missing": 0,
"lines_extra": 0
},
"select": {
"sha256": "2aa97d5d63e5e01cb9fafc2d2508ad5df1c576467dfc46cb72b714fea35c93a5",
"equal_to_expected": true,
"first_difference": null,
"lines_missing": 0,
"lines_extra": 0
}
},
"known_failed_flagged": null,
"verdict": "EQUAL"
}

25
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-03.prod.expected.meta.json Normal file
View file

@ -0,0 +1,25 @@
{
"ctx": "same-work-20261008-03",
"shape": "prod",
"jobs": 16,
"nonces": 9437568,
"found_lines": 3146498,
"labels": {
"p1-all": 1,
"sep": 6,
"p1-t256": 1,
"p1-t257": 1,
"p1-zero": 1,
"p1-teq": 1,
"p1-teq-1": 1,
"p2a-all": 1,
"p2b-all": 1,
"p3-all": 1,
"p3-t256": 1
},
"norm_sha256": "2aa97d5d63e5e01cb9fafc2d2508ad5df1c576467dfc46cb72b714fea35c93a5",
"references": {
"D": "873528b18ad679ca9a748d52d45c4e1b6723ba703875742fb01e1cbfca169783",
"D1": "7f8e64ae899ed5452dab6a01a2b34d66f5a11ef1ca8c65a9b93ff0be2a88250c"
}
}

34
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-03.prod.full.log Normal file
View file

@ -0,0 +1,34 @@
info igneum-worker-cuda 1.0 (4 October 2026): device 0 NVIDIA_GeForce_RTX_4090 (sm_89, 128 SMs), driver 13.2 from libcuda.so.1, NVRTC 12.8 from libnvrtc.so.12, target sm_89 (the device's architecture, listed by NVRTC)
ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/3c3fab434f7fd894ce6db0b09f298c3865726253e4fc68fd77acd163ccea8221/day/69676e65756d2d6461792ffe50000000000000 dataset-log2 28 batch 4194304 regs 31 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback full worker 1.0 (4 October 2026)
info readback full (per dispatch of 4194304 nonces: 8 bytes per nonce come back and the host scans them)
info first pack /root/x2/ctx/same-work-20261008-03/packs/v5-epoch2-d20734: nvrtc 618 cache 2 dataset 29 hot 0 check 1167 race 0 ms variant base class v5 (state leaves 224, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 5cc75af0f884bd4c; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes)
done 1 1048576 1412.89
done 2 64 0.37
done 3 1048576 21.37
done 4 64 0.26
done 5 1048576 19.13
done 6 64 0.24
done 7 1048576 18.81
done 8 64 0.24
done 9 1048576 18.89
done 10 64 0.24
done 11 1048576 18.89
done 12 524288 732.23
info prepare started for epoch 3c3fab434f7fd894 day 69676e65756d2d6461792fff50000000000000 from /root/x2/ctx/same-work-20261008-03/packs/v5-epoch2-d20735 (NVRTC sm_89 in the background)
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
prepared 3c3fab434f7fd894ce6db0b09f298c3865726253e4fc68fd77acd163ccea8221 69676e65756d2d6461792fff50000000000000 1759.3 nvrtc 615 cache 2 dataset 29 hot 0 check 1091 race 0 ms variant base class v5 (state leaves 224, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 a51d063d7dfa091b; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes) resident 2 programs 2 datasets
info ignored: noop
info switched to the prepared pair epoch 3c3fab434f7fd894 day 69676e65756d2d6461792fff50000000000000 (class v5) in 0.01 ms
done 13 524288 751.28
info dropped the previous pair (its program, cache and dataset)
done 14 1048576 1401.08
done 15 64 0.30
done 16 1048576 19.17
info transfers per chunk: down 4718784 B (full, 0 full fallbacks); mean per chunk: kernel+sync 8.69 ms, select 0.000 ms, read-back 1.705 ms, scan 265.515 ms; 16 jobs 16 chunks

19
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-03.prod.full.meta.json Normal file
View file

@ -0,0 +1,19 @@
{
"ctx": "same-work-20261008-03",
"object_id": "0x81fbfa3aa413173f",
"shape": "prod",
"mode": "full",
"batch_log2": 22,
"jobs": 16,
"nonces": 9437568,
"worker": "/root/x2/bin/igneum-worker-cuda",
"worker_sha256": "4152c960d14210072a5c85ae675b56ae4dab12bedfb5f1af04ea2a2c6c025032",
"ready": "ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/3c3fab434f7fd894ce6db0b09f298c3865726253e4fc68fd77acd163ccea8221/day/69676e65756d2d6461792ffe50000000000000 dataset-log2 28 batch 4194304 regs 31 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback full worker 1.0 (4 October 2026)",
"prepared": "prepared 3c3fab434f7fd894ce6db0b09f298c3865726253e4fc68fd77acd163ccea8221 69676e65756d2d6461792fff50000000000000 1759.3 nvrtc 615 cache 2 dataset 29 hot 0 check 1091 race 0 ms variant base class v5 (state leaves 224, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 a51d063d7dfa091b; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes) resident",
"transfers": "info transfers per chunk: down 4718784 B (full, 0 full fallbacks); mean per chunk: kernel+sync 8.69 ms, select 0.000 ms, read-back 1.705 ms, scan 265.515 ms; 16 jobs 16 chunks",
"errors": [],
"norm_sha256": "2aa97d5d63e5e01cb9fafc2d2508ad5df1c576467dfc46cb72b714fea35c93a5",
"seconds": 9.0,
"exit": 0,
"at": "2026-10-09T08:51:25Z"
}

35
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-03.prod.select.log Normal file
View file

@ -0,0 +1,35 @@
info igneum-worker-cuda 1.0 (4 October 2026): device 0 NVIDIA_GeForce_RTX_4090 (sm_89, 128 SMs), driver 13.2 from libcuda.so.1, NVRTC 12.8 from libnvrtc.so.12, target sm_89 (the device's architecture, listed by NVRTC)
ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/3c3fab434f7fd894ce6db0b09f298c3865726253e4fc68fd77acd163ccea8221/day/69676e65756d2d6461792ffe50000000000000 dataset-log2 28 batch 4194304 regs 31 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback select worker 1.0 (4 October 2026)
info readback select (per dispatch of 4194304 nonces: the hits and 34 sentinel words come back; the GPU scans)
info first pack /root/x2/ctx/same-work-20261008-03/packs/v5-epoch2-d20734: nvrtc 678 cache 2 dataset 29 hot 0 check 1415 race 0 ms variant base class v5 (state leaves 224, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 5cc75af0f884bd4c; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes)
done 1 1048576 1627.50
done 2 64 0.40
done 3 1048576 18.54
done 4 64 0.45
done 5 1048576 17.65
done 6 64 0.40
done 7 1048576 15.26
done 8 64 0.34
done 9 1048576 15.35
done 10 64 0.34
done 11 1048576 15.27
done 12 524288 817.65
info prepare started for epoch 3c3fab434f7fd894 day 69676e65756d2d6461792fff50000000000000 from /root/x2/ctx/same-work-20261008-03/packs/v5-epoch2-d20735 (NVRTC sm_89 in the background)
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
info ignored: noop
prepared 3c3fab434f7fd894ce6db0b09f298c3865726253e4fc68fd77acd163ccea8221 69676e65756d2d6461792fff50000000000000 1973.7 nvrtc 669 cache 2 dataset 28 hot 0 check 1176 race 0 ms variant base class v5 (state leaves 224, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 a51d063d7dfa091b; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes) resident 2 programs 2 datasets
info ignored: noop
info switched to the prepared pair epoch 3c3fab434f7fd894 day 69676e65756d2d6461792fff50000000000000 (class v5) in 0.00 ms
done 13 524288 781.04
info dropped the previous pair (its program, cache and dataset)
done 14 1048576 1554.47
done 15 64 0.56
done 16 1048576 15.32
info transfers per chunk: down 2097856 B (select, 5 full fallbacks); mean per chunk: kernel+sync 8.67 ms, select 0.074 ms, read-back 0.614 ms, scan 295.617 ms; 16 jobs 16 chunks

19
docs/analysis/class-v6/1p5x/x2/evidence/same-work-20261008-03.prod.select.meta.json Normal file
View file

@ -0,0 +1,19 @@
{
"ctx": "same-work-20261008-03",
"object_id": "0x81fbfa3aa413173f",
"shape": "prod",
"mode": "select",
"batch_log2": 22,
"jobs": 16,
"nonces": 9437568,
"worker": "/root/x2/bin/igneum-worker-cuda",
"worker_sha256": "4152c960d14210072a5c85ae675b56ae4dab12bedfb5f1af04ea2a2c6c025032",
"ready": "ready cuda NVIDIA_GeForce_RTX_4090 pack igneum-epoch/3c3fab434f7fd894ce6db0b09f298c3865726253e4fc68fd77acd163ccea8221/day/69676e65756d2d6461792ffe50000000000000 dataset-log2 28 batch 4194304 regs 31 prepare 1 path nvrtc 12.8 driver 13.2 arch sm_89 variant base race off readback select worker 1.0 (4 October 2026)",
"prepared": "prepared 3c3fab434f7fd894ce6db0b09f298c3865726253e4fc68fd77acd163ccea8221 69676e65756d2d6461792fff50000000000000 1973.7 nvrtc 669 cache 2 dataset 28 hot 0 check 1176 race 0 ms variant base class v5 (state leaves 224, uploaded for the build and freed); self-test PASS (cache head, last line and FNV-1a 64 a51d063d7dfa091b; dataset head, word [268435455] and 64 samples; 96 of 96 vector lanes) resident",
"transfers": "info transfers per chunk: down 2097856 B (select, 5 full fallbacks); mean per chunk: kernel+sync 8.67 ms, select 0.074 ms, read-back 0.614 ms, scan 295.617 ms; 16 jobs 16 chunks",
"errors": [],
"norm_sha256": "2aa97d5d63e5e01cb9fafc2d2508ad5df1c576467dfc46cb72b714fea35c93a5",
"seconds": 10.1,
"exit": 0,
"at": "2026-10-09T08:51:40Z"
}

40
docs/analysis/class-v6/1p5x/x2/evidence/self-test.log Normal file
View file

@ -0,0 +1,40 @@
{"ctx": "self", "shape": "kf", "jobs": 4, "nonces": 8192, "found_lines": 512, "labels": {"kf": 4}, "norm_sha256": "2d8beee1ed8174274445167274dc2a9d698d9604e8bf78e750433aacdefc8a16", "references": {"D": "43a9f33483a083801ffea2f933f8655ea539b0a3d315ba326f84a272776d364e", "D1": "ddde2638f1fb2479fc692d86cedf139734b2618554aceba8169c245a3d03bbf2"}}
{"ctx": "self", "object_id": null, "shape": "kf", "mode": "select", "batch_log2": 10, "jobs": 4, "nonces": 8192, "worker": "/tmp/x2-self-bvucnnwo/fakeworker.py", "worker_sha256": "d3b02b1122e7046e276ec1f8294ab0448e285e0048d3ad1756f879f03a69ac5e", "ready": "ready fake-gpu batch 1024 readback select", "prepared": null, "transfers": null, "errors": [], "norm_sha256": "2d8beee1ed8174274445167274dc2a9d698d9604e8bf78e750433aacdefc8a16", "seconds": 0.0, "exit": 0, "at": "2026-10-09T08:45:58Z"}
self-test: shape kf fault "none": EQUAL at None
{"ctx": "self", "object_id": null, "shape": "kf", "mode": "select", "batch_log2": 10, "jobs": 4, "nonces": 8192, "worker": "/tmp/x2-self-bvucnnwo/fakeworker.py", "worker_sha256": "d3b02b1122e7046e276ec1f8294ab0448e285e0048d3ad1756f879f03a69ac5e", "ready": "ready fake-gpu batch 1024 readback select", "prepared": null, "transfers": null, "errors": [], "norm_sha256": "2d8beee1ed8174274445167274dc2a9d698d9604e8bf78e750433aacdefc8a16", "seconds": 0.0, "exit": 0, "at": "2026-10-09T08:46:00Z"}
self-test: shape kf fault "drop257": EQUAL at None
{"ctx": "self", "object_id": null, "shape": "kf", "mode": "select", "batch_log2": 10, "jobs": 4, "nonces": 8192, "worker": "/tmp/x2-self-bvucnnwo/fakeworker.py", "worker_sha256": "d3b02b1122e7046e276ec1f8294ab0448e285e0048d3ad1756f879f03a69ac5e", "ready": "ready fake-gpu batch 1024 readback select", "prepared": null, "transfers": null, "errors": [], "norm_sha256": "f0739b6c03045efed05406d7b5d2d85d9836664aae383b093adf4b499456118c", "seconds": 0.0, "exit": 0, "at": "2026-10-09T08:46:03Z"}
self-test: shape kf fault "digest": DIFFERENT at {'line': 1, 'expected': 'found 1 0 0000000000000000', 'got': 'found 1 0 0000000000000001'}
{"ctx": "self", "object_id": null, "shape": "kf", "mode": "select", "batch_log2": 10, "jobs": 4, "nonces": 8192, "worker": "/tmp/x2-self-bvucnnwo/fakeworker.py", "worker_sha256": "d3b02b1122e7046e276ec1f8294ab0448e285e0048d3ad1756f879f03a69ac5e", "ready": "ready fake-gpu batch 1024 readback select", "prepared": null, "transfers": null, "errors": [], "norm_sha256": "dc634f68214398676a9801b25e6c55d66a8c7a02bf3c6d4547c094b1fb054e5d", "seconds": 0.0, "exit": 0, "at": "2026-10-09T08:46:05Z"}
self-test: shape kf fault "stale": DIFFERENT at {'line': 65, 'expected': 'found 1 1034 0c11a73c22d92cd2', 'got': 'found 1 1024 0000000000000000'}
{"ctx": "self", "object_id": null, "shape": "kf", "mode": "select", "batch_log2": 10, "jobs": 4, "nonces": 8192, "worker": "/tmp/x2-self-bvucnnwo/fakeworker.py", "worker_sha256": "d3b02b1122e7046e276ec1f8294ab0448e285e0048d3ad1756f879f03a69ac5e", "ready": "ready fake-gpu batch 1024 readback select", "prepared": null, "transfers": null, "errors": [], "norm_sha256": "2d8beee1ed8174274445167274dc2a9d698d9604e8bf78e750433aacdefc8a16", "seconds": 0.0, "exit": 0, "at": "2026-10-09T08:46:08Z"}
self-test: shape kf fault "noswitch": EQUAL at None
{"ctx": "self", "object_id": null, "shape": "kf", "mode": "select", "batch_log2": 10, "jobs": 4, "nonces": 8192, "worker": "/tmp/x2-self-bvucnnwo/fakeworker.py", "worker_sha256": "d3b02b1122e7046e276ec1f8294ab0448e285e0048d3ad1756f879f03a69ac5e", "ready": "ready fake-gpu batch 1024 readback select", "prepared": null, "transfers": null, "errors": [], "norm_sha256": "2d8beee1ed8174274445167274dc2a9d698d9604e8bf78e750433aacdefc8a16", "seconds": 0.0, "exit": 0, "at": "2026-10-09T08:46:11Z"}
self-test: shape kf fault "sentinel": EQUAL at None
{"ctx": "self", "shape": "prod", "jobs": 10, "nonces": 9437184, "found_lines": 3146500, "labels": {"p1-all": 1, "p1-t256": 1, "p1-t257": 1, "p1-zero": 1, "p1-teq": 1, "p1-teq-1": 1, "p2a-all": 1, "p2b-all": 1, "p3-all": 1, "p3-t256": 1}, "norm_sha256": "f3a4923f08ed8152e4882856873c09d3bb56ee43b2165a6c19f0e0c9b7413984", "references": {"D": "43a9f33483a083801ffea2f933f8655ea539b0a3d315ba326f84a272776d364e", "D1": "ddde2638f1fb2479fc692d86cedf139734b2618554aceba8169c245a3d03bbf2"}}
{"ctx": "self", "object_id": null, "shape": "prod", "mode": "select", "batch_log2": 22, "jobs": 10, "nonces": 9437184, "worker": "/tmp/x2-self-bvucnnwo/fakeworker.py", "worker_sha256": "d3b02b1122e7046e276ec1f8294ab0448e285e0048d3ad1756f879f03a69ac5e", "ready": "ready fake-gpu batch 4194304 readback select", "prepared": "prepared fake", "transfers": null, "errors": [], "norm_sha256": "f3a4923f08ed8152e4882856873c09d3bb56ee43b2165a6c19f0e0c9b7413984", "seconds": 9.4, "exit": 0, "at": "2026-10-09T08:46:32Z"}
self-test: shape prod fault "none": EQUAL at None
{"ctx": "self", "object_id": null, "shape": "prod", "mode": "select", "batch_log2": 22, "jobs": 10, "nonces": 9437184, "worker": "/tmp/x2-self-bvucnnwo/fakeworker.py", "worker_sha256": "d3b02b1122e7046e276ec1f8294ab0448e285e0048d3ad1756f879f03a69ac5e", "ready": "ready fake-gpu batch 4194304 readback select", "prepared": "prepared fake", "transfers": null, "errors": [], "norm_sha256": "52e0116ae1702512223c3cff031d36c2dbac6fc8427f514ec7143a1ae4b453a2", "seconds": 10.0, "exit": 0, "at": "2026-10-09T08:46:47Z"}
self-test: shape prod fault "drop257": DIFFERENT at {'line': 1049091, 'expected': 'found 3 1043585 0009cd2f1426e295', 'got': 'done 3 1048576'}
{"ctx": "self", "object_id": null, "shape": "prod", "mode": "select", "batch_log2": 22, "jobs": 10, "nonces": 9437184, "worker": "/tmp/x2-self-bvucnnwo/fakeworker.py", "worker_sha256": "d3b02b1122e7046e276ec1f8294ab0448e285e0048d3ad1756f879f03a69ac5e", "ready": "ready fake-gpu batch 4194304 readback select", "prepared": "prepared fake", "transfers": null, "errors": [], "norm_sha256": "eb9a7c451352e0925b8c77a729eff53276262599621fc211ff37c36d28de05cc", "seconds": 10.4, "exit": 0, "at": "2026-10-09T08:47:02Z"}
self-test: shape prod fault "digest": DIFFERENT at {'line': 1, 'expected': 'found 1 0 0000000000000000', 'got': 'found 1 0 0000000000000001'}
{"ctx": "self", "object_id": null, "shape": "prod", "mode": "select", "batch_log2": 22, "jobs": 10, "nonces": 9437184, "worker": "/tmp/x2-self-bvucnnwo/fakeworker.py", "worker_sha256": "d3b02b1122e7046e276ec1f8294ab0448e285e0048d3ad1756f879f03a69ac5e", "ready": "ready fake-gpu batch 4194304 readback select", "prepared": "prepared fake", "transfers": null, "errors": [], "norm_sha256": "f3a4923f08ed8152e4882856873c09d3bb56ee43b2165a6c19f0e0c9b7413984", "seconds": 11.5, "exit": 0, "at": "2026-10-09T08:47:18Z"}
self-test: shape prod fault "stale": EQUAL at None
{"ctx": "self", "object_id": null, "shape": "prod", "mode": "select", "batch_log2": 22, "jobs": 10, "nonces": 9437184, "worker": "/tmp/x2-self-bvucnnwo/fakeworker.py", "worker_sha256": "d3b02b1122e7046e276ec1f8294ab0448e285e0048d3ad1756f879f03a69ac5e", "ready": "ready fake-gpu batch 4194304 readback select", "prepared": "prepared fake", "transfers": null, "errors": [], "norm_sha256": "bd670702dc11916a3e4b352e1849bf6a2cd49f1aee7530de1d665445a0dfb68e", "seconds": 9.5, "exit": 0, "at": "2026-10-09T08:47:32Z"}
self-test: shape prod fault "noswitch": DIFFERENT at {'line': 1573388, 'expected': 'found 8 1572864 6963eefba1f85555', 'got': 'found 8 1572864 6963eefba1f80000'}
{"ctx": "self", "object_id": null, "shape": "prod", "mode": "select", "batch_log2": 22, "jobs": 10, "nonces": 9437184, "worker": "/tmp/x2-self-bvucnnwo/fakeworker.py", "worker_sha256": "d3b02b1122e7046e276ec1f8294ab0448e285e0048d3ad1756f879f03a69ac5e", "ready": "ready fake-gpu batch 4194304 readback select", "prepared": "prepared fake", "transfers": null, "errors": [], "norm_sha256": "5fc50318fe56fa9727c9cd904f2a715783fe7d29070b7b1f1173c89ec43fa5b2", "seconds": 9.9, "exit": 0, "at": "2026-10-09T08:47:47Z"}
self-test: shape prod fault "sentinel": DIFFERENT at {'line': 1048671, 'expected': 'found 2 383487 000ff9205ac865eb', 'got': 'found 2 386071 0004a17d329cd9e3'}
{"ctx": "self", "shape": "fuzz", "jobs": 37, "nonces": 1048576, "found_lines": 237746, "labels": {"fz2": 2, "fz3": 4, "fz1": 2, "fz4": 6, "fz0": 6, "fz7": 4, "fz6": 4, "fz5": 5, "blk-d-all": 1, "blk-d-q4": 1, "blk-d1-all": 1, "blk-d1-q4": 1}, "norm_sha256": "92dbf4accd7040a390cb9df8e1f2a01723624708050c5d0ec9daba7c50c56624", "references": {"D": "43a9f33483a083801ffea2f933f8655ea539b0a3d315ba326f84a272776d364e", "D1": "ddde2638f1fb2479fc692d86cedf139734b2618554aceba8169c245a3d03bbf2"}}
{"ctx": "self", "object_id": null, "shape": "fuzz", "mode": "select", "batch_log2": 10, "jobs": 37, "nonces": 1048576, "worker": "/tmp/x2-self-bvucnnwo/fakeworker.py", "worker_sha256": "d3b02b1122e7046e276ec1f8294ab0448e285e0048d3ad1756f879f03a69ac5e", "ready": "ready fake-gpu batch 1024 readback select", "prepared": "prepared fake", "transfers": null, "errors": [], "norm_sha256": "92dbf4accd7040a390cb9df8e1f2a01723624708050c5d0ec9daba7c50c56624", "seconds": 1.1, "exit": 0, "at": "2026-10-09T08:47:53Z"}
self-test: shape fuzz fault "none": EQUAL at None
{"ctx": "self", "object_id": null, "shape": "fuzz", "mode": "select", "batch_log2": 10, "jobs": 37, "nonces": 1048576, "worker": "/tmp/x2-self-bvucnnwo/fakeworker.py", "worker_sha256": "d3b02b1122e7046e276ec1f8294ab0448e285e0048d3ad1756f879f03a69ac5e", "ready": "ready fake-gpu batch 1024 readback select", "prepared": "prepared fake", "transfers": null, "errors": [], "norm_sha256": "979f1b0868425b18a797a58827f0dbeaff4b4e93358fed3ab7b97b9e7971fd99", "seconds": 1.1, "exit": 0, "at": "2026-10-09T08:47:56Z"}
self-test: shape fuzz fault "drop257": DIFFERENT at {'line': 26423, 'expected': 'found 4 490237 20e82e05b051a2c1', 'got': 'found 4 490242 37fd8ea52cc60f2a'}
{"ctx": "self", "object_id": null, "shape": "fuzz", "mode": "select", "batch_log2": 10, "jobs": 37, "nonces": 1048576, "worker": "/tmp/x2-self-bvucnnwo/fakeworker.py", "worker_sha256": "d3b02b1122e7046e276ec1f8294ab0448e285e0048d3ad1756f879f03a69ac5e", "ready": "ready fake-gpu batch 1024 readback select", "prepared": "prepared fake", "transfers": null, "errors": [], "norm_sha256": "8402b980887ec36a856842e131b57aa4ad7ec1e147f03137544ac06325b1a24a", "seconds": 1.1, "exit": 0, "at": "2026-10-09T08:48:00Z"}
self-test: shape fuzz fault "digest": DIFFERENT at {'line': 1, 'expected': 'found 1 150050 0000c80367ec8cca', 'got': 'found 1 150050 0000c80367ec8ccb'}
{"ctx": "self", "object_id": null, "shape": "fuzz", "mode": "select", "batch_log2": 10, "jobs": 37, "nonces": 1048576, "worker": "/tmp/x2-self-bvucnnwo/fakeworker.py", "worker_sha256": "d3b02b1122e7046e276ec1f8294ab0448e285e0048d3ad1756f879f03a69ac5e", "ready": "ready fake-gpu batch 1024 readback select", "prepared": "prepared fake", "transfers": null, "errors": [], "norm_sha256": "b5b5be9861bde2d15d0397a484c74bf436a93553be33def842bf281b7e023fdc", "seconds": 2.3, "exit": 0, "at": "2026-10-09T08:48:04Z"}
self-test: shape fuzz fault "stale": DIFFERENT at {'line': 2, 'expected': 'found 1 154231 0007ca8f53690fc3', 'got': 'found 1 151074 0000c80367ec8cca'}
{"ctx": "self", "object_id": null, "shape": "fuzz", "mode": "select", "batch_log2": 10, "jobs": 37, "nonces": 1048576, "worker": "/tmp/x2-self-bvucnnwo/fakeworker.py", "worker_sha256": "d3b02b1122e7046e276ec1f8294ab0448e285e0048d3ad1756f879f03a69ac5e", "ready": "ready fake-gpu batch 1024 readback select", "prepared": "prepared fake", "transfers": null, "errors": [], "norm_sha256": "af74de6e90d83442e7749d2ac536c87be0342c75d89a53562d63c1be266b5a28", "seconds": 1.1, "exit": 0, "at": "2026-10-09T08:48:08Z"}
self-test: shape fuzz fault "noswitch": DIFFERENT at {'line': 76658, 'expected': 'found 17 1572864 6963eefba1f85555', 'got': 'found 17 1572864 6963eefba1f80000'}
{"ctx": "self", "object_id": null, "shape": "fuzz", "mode": "select", "batch_log2": 10, "jobs": 37, "nonces": 1048576, "worker": "/tmp/x2-self-bvucnnwo/fakeworker.py", "worker_sha256": "d3b02b1122e7046e276ec1f8294ab0448e285e0048d3ad1756f879f03a69ac5e", "ready": "ready fake-gpu batch 1024 readback select", "prepared": "prepared fake", "transfers": null, "errors": [], "norm_sha256": "8f27c0866bf219b18b1aab5c20e47bdefe4986937de41da1a132d57997f2e1d4", "seconds": 1.1, "exit": 0, "at": "2026-10-09T08:48:11Z"}
self-test: shape fuzz fault "sentinel": DIFFERENT at {'line': 33503, 'expected': 'found 7 624490 0baeb28d590373b2', 'got': 'found 7 624511 063baec4ca1fa16b'}
SELF-TEST PASS: the clean fake is EQUAL on every shape; drop-257, a wrong digest, stale hits, a missed day switch and a < for <= are each DIFFERENT

20
tools/1p5x/x2/registry-batch-x2-1p5x-20261009.json Normal file
View file

@ -0,0 +1,20 @@
{
"run_id": "x2-1p5x-20261009",
"manifest_sha": "1a938abe4",
"evidence_dir": "docs/analysis/class-v6/1p5x/x2",
"cells": [
{
"cell": "harness:cuda-select-check",
"status": "NOT RUN",
"evidence": "docs/analysis/class-v6/1p5x/x2/README.md;docs/analysis/class-v6/1p5x/x2/evidence/run-equiv.log;docs/analysis/class-v6/1p5x/x2/evidence/run-prod-then-pairs.log;docs/analysis/class-v6/1p5x/x2/evidence/self-test.log;docs/analysis/class-v6/1p5x/x2/evidence/norm.sha256",
"method": "GPU"
},
{
"cell": "bench:fleet-pods",
"status": "NOT RUN",
"evidence": "docs/analysis/class-v6/1p5x/x2/README.md;docs/analysis/class-v6/1p5x/x2/evidence/pairs/summary.txt;docs/analysis/class-v6/1p5x/x2/evidence/pairs/kernel-bench.row",
"method": "GPU"
}
],
"note": "X2 rows for the steward to record (NOT RUN, evidence named by file). Not under tools/ci/batches/ on purpose: merge-to-master replays every batch a branch adds, and a NOT RUN write over these cells could overwrite another lane's rows for the mapped cases; the steward records it with the cases it chooses."
}

41
tools/1p5x/x2/run-equiv.sh Executable file
View file

@ -0,0 +1,41 @@
#!/usr/bin/env bash
# X2 equivalence campaign on one GPU host (9 October 2026). Known-failed first: the fault run (select with
# IGNEUM_READBACK_FAULT_TEST=1) on the kf plan must compare DIFFERENT before any select row counts; then per same-work
# context the expected file from the CPU references, the full run and the select run on the prod and fuzz plans, and the
# compare. Every process runs under a pid file in $PIDS (the fleet rule: stop by pid file, never by name).
# run-equiv.sh <worker> <contexts root> <out dir> <pid dir> [context ids...]
set -uo pipefail
W="$1"; ROOT="$2"; OUT="$3"; PIDS="$4"; shift 4
CTXS=("$@"); [ ${#CTXS[@]} -gt 0 ] || CTXS=(same-work-20261008-01 same-work-20261008-02 same-work-20261008-03)
HERE="$(cd "$(dirname "$0")" && pwd)"; D="$HERE/x2-equiv.py"
mkdir -p "$OUT" "$PIDS"
echo "$$ $(date -u +%FT%TZ) x2 run-equiv" > "$PIDS/x2-run-equiv.pid"
step() { echo "== $(date -u +%T) $*"; }
fail=0
c0="${CTXS[0]}"
step "known-failed first: kf plan on $c0, fault then select"
python3 "$D" expect --ctx "$ROOT/$c0" --shape kf --out "$OUT" > /dev/null
python3 "$D" run --worker "$W" --ctx "$ROOT/$c0" --mode fault --shape kf --out "$OUT" --pid-dir "$PIDS" > /dev/null
python3 "$D" run --worker "$W" --ctx "$ROOT/$c0" --mode full --shape kf --out "$OUT" --pid-dir "$PIDS" > /dev/null
python3 "$D" run --worker "$W" --ctx "$ROOT/$c0" --mode select --shape kf --out "$OUT" --pid-dir "$PIDS" > /dev/null
python3 "$D" compare --out "$OUT" --ctx-id "$c0" --shape kf | tail -1
python3 - "$OUT/$c0.kf.compare.json" <<'EOF' || { echo "STOP: the known-failed run was not flagged; no select row counts"; rm -f "$PIDS/x2-run-equiv.pid"; exit 1; }
import json, sys
r = json.load(open(sys.argv[1]))
ok = r['known_failed_flagged'] is True and r['verdict'] == 'EQUAL'
print('known-failed:', 'flagged' if r['known_failed_flagged'] else 'NOT FLAGGED', r['modes']['fault'].get('first_difference'), '| kf full/select:', r['verdict'])
sys.exit(0 if ok else 1)
EOF
for c in "${CTXS[@]}"; do
for shape in prod fuzz; do
step "$c $shape: expected, full, select"
python3 "$D" expect --ctx "$ROOT/$c" --shape "$shape" --out "$OUT" > /dev/null
python3 "$D" run --worker "$W" --ctx "$ROOT/$c" --mode full --shape "$shape" --out "$OUT" --pid-dir "$PIDS" > /dev/null
python3 "$D" run --worker "$W" --ctx "$ROOT/$c" --mode select --shape "$shape" --out "$OUT" --pid-dir "$PIDS" > /dev/null
python3 "$D" compare --out "$OUT" --ctx-id "$c" --shape "$shape" > /dev/null || fail=1
python3 -c "import json,sys; r=json.load(open(sys.argv[1])); print(r['ctx'], r['shape'], r['verdict'], {m: (v if isinstance(v, str) else (v['equal_to_expected'], v['sha256'][:16])) for m, v in r['modes'].items()}, 'expected', r['expected_sha256'][:16])" "$OUT/$c.$shape.compare.json"
done
done
step "end, fail=$fail"
rm -f "$PIDS/x2-run-equiv.pid"
exit $fail

62
tools/1p5x/x2/run-pairs.sh Executable file
View file

@ -0,0 +1,62 @@
#!/usr/bin/env bash
# X2 measured rows on one GPU host (9 October 2026): same-session pairs of the serving mode, compacted (--readback select)
# against uncompacted (--readback full), in ABBA order so a thermal drift falls on both sides, plus the kernel timer (the
# instrument the record's E_GPU rows use: --bench --batch-log2 24 --batches 250) with nvidia-smi at 1 Hz around it, all on
# one pack: the frozen object's own (same-work context 02, chain-seed-d20730, program id 0x2a1d6caab4c24564, day 20730).
# run-pairs.sh <worker> <pack dir> <out dir> <pid dir> [pairs 3] [jobs 192] [batch-log2 22]
# Writes <out>/<run>.row per serving run (serve-pair.sh) and <out>/kernel-bench.{log,smi,row}.
set -uo pipefail
W="$1"; PACK0="$2"; OUT="$3"; PIDS="$4"; PAIRS="${5:-3}"; JOBS="${6:-192}"; B="${7:-22}"
HERE="$(cd "$(dirname "$0")" && pwd)"
mkdir -p "$OUT" "$PIDS"
echo "$$ $(date -u +%FT%TZ) x2 run-pairs" > "$PIDS/x2-run-pairs.pid"
# serve-pair.sh reads the seeds from <pack>/seeds.txt; the same-work packs carry them in program.json only
PACK="$OUT/pack-$(basename "$PACK0")"
mkdir -p "$PACK"
cp -p "$PACK0"/* "$PACK"/
python3 - "$PACK" <<'EOF'
import json, sys
j = json.load(open(sys.argv[1] + '/program.json'))
open(sys.argv[1] + '/seeds.txt', 'w').write('epoch_seed_hex %s\nday_seed_hex %s\n' % (j['seed_bytes'], j['dataset']['day_bytes']))
EOF
export IGNEUM_PID_DIR="$PIDS"
echo "== $(date -u +%T) idle card before the rows"
nvidia-smi --query-gpu=name,driver_version,power.limit,power.draw,clocks.sm,clocks.mem,temperature.gpu,utilization.gpu,pcie.link.gen.current,pcie.link.width.current --format=csv
for i in $(seq 1 "$PAIRS"); do
if [ $((i % 2)) = 1 ]; then order="full select"; else order="select full"; fi
for m in $order; do
echo "== $(date -u +%T) pair $i $m"
"$HERE/serve-pair.sh" "$W" "$PACK" "$OUT" "x2-$m-r$i" "$m" "$JOBS" "$B" 0 | head -1
sleep 5
done
done
echo "== $(date -u +%T) kernel timer (--bench 250 x 2^24, nvidia-smi 1 Hz)"
nvidia-smi -i 0 --query-gpu=timestamp,power.draw,clocks.sm,temperature.gpu --format=csv,noheader,nounits -lms 1000 > "$OUT/kernel-bench.smi" 2>/dev/null &
echo $! > "$PIDS/x2-kernel-bench.smi.pid"
sleep 2
t0=$(date +%s.%N)
"$W" --bench --pack "$PACK" --batch-log2 24 --batches 250 --race off > "$OUT/kernel-bench.log" 2>&1 &
echo $! > "$PIDS/x2-kernel-bench.pid"; wait "$(cat "$PIDS/x2-kernel-bench.pid")"
t1=$(date +%s.%N)
sleep 2
kill "$(cat "$PIDS/x2-kernel-bench.smi.pid")" 2>/dev/null
rm -f "$PIDS/x2-kernel-bench.pid" "$PIDS/x2-kernel-bench.smi.pid"
grep '^RESULT' "$OUT/kernel-bench.log" | tail -1
# the bench's own warm-up and build precede the timed loop: the mean watts of the samples above 100 W inside the run stand
# for the loaded card (the card idles near 30 W on this host)
python3 - "$OUT/kernel-bench.smi" "$OUT/kernel-bench.log" "$t0" "$t1" > "$OUT/kernel-bench.row" <<'EOF'
import sys, re, datetime
smi, log, t0, t1 = sys.argv[1], sys.argv[2], float(sys.argv[3]), float(sys.argv[4])
w = []
for line in open(smi):
p = [x.strip() for x in line.split(',')]
ts = datetime.datetime.strptime(p[0], '%Y/%m/%d %H:%M:%S.%f').timestamp()
if t0 <= ts <= t1 and float(p[1]) > 100: w.append(float(p[1]))
r = [l for l in open(log) if l.startswith('RESULT')][-1]
mhs = float(re.search(r'mhs=([0-9.]+)', r).group(1)); fp = re.search(r'fingerprint=([0-9a-f]+)', r).group(1)
mw = sum(w) / len(w)
print(f'kernel_mhs={mhs:.3f} loaded_mean_w={mw:.1f} samples={len(w)} nj_per_hash={mw / mhs * 1000:.1f} fingerprint={fp}')
EOF
cat "$OUT/kernel-bench.row"
echo "== $(date -u +%T) end"
rm -f "$PIDS/x2-run-pairs.pid"

97
tools/1p5x/x2/serve-pair.sh Executable file
View file

@ -0,0 +1,97 @@
#!/usr/bin/env bash
# X2 (the 1.5x programme, 9 October 2026): proto-cuda/nvrtc/serve-bench.sh at 1a938abe4 with three changes for the X2 rows:
# nvidia-smi at 1 Hz (SMI_MS, default 1000; the brief's rate), the batch default 22 (the worker's and the miner's), and the
# accepted-share fields (every found line re-checked against the job's share target: hash <= target; shares and nJ per
# share and per hash in the row). Everything else is the F10 instrument unchanged.
# Serving-mode measurement of igneum-worker-cuda (8 October 2026, review B finding F10): the production number beside the
# kernel timer, never interchangeable with it. The worker runs --serve exactly as the miner drives it: a stub pool here
# feeds jobs of 2^B nonces (one queued behind the one in flight, as the miner pipelines them), reads every found line
# (the miner's submit point) and every done line, and quits after N jobs. The clock starts before the worker process
# starts, so the setup (NVRTC compile, cache, dataset, self-test) is in the production number; the steady number runs
# from the ready line. Power is nvidia-smi's board reading at 4 Hz over the same windows (the pod exposes no host
# rail); the worker's own CPU seconds (utime + stime from /proc) stand in for the host-side cost of the read-back and
# the scan. The kernel timer is --bench on the same pack and batch, printed beside the rows.
# serve-bench.sh <worker> <pack dir> <out dir> <run name> <readback: full|select> [jobs 32] [batch-log2 24] [device 0]
# Pid files: every process this script starts writes <out dir>/pids/<run name>.<what>.pid (IGNEUM_PID_DIR overrides the directory) (the fleet's rule: a job is
# stopped through its pid file, never by a name pattern). Output: <out dir>/<run name>.row (one line, key=value) plus the
# worker log, the stub's timeline and the power samples.
set -euo pipefail
W="$1"; PACK="$2"; OUT="$3"; RUN="$4"; MODE="$5"; JOBS="${6:-32}"; B="${7:-22}"; DEV="${8:-0}"; SMI_MS="${SMI_MS:-1000}"
[ "$MODE" = full ] || [ "$MODE" = select ] || { echo "readback must be full or select" >&2; exit 2; }
PIDDIR="${IGNEUM_PID_DIR:-$OUT/pids}" # the fleet keeps pid files under one directory per pod
mkdir -p "$PIDDIR"
LOG="$OUT/$RUN.worker.log"; TL="$OUT/$RUN.timeline"; SMI="$OUT/$RUN.smi"; ROW="$OUT/$RUN.row"
: > "$TL"
now() { date +%s.%N; }
count=$((1 << B))
target="00000fffffffffff" # one hit per 2^20 nonces: 16 found lines per 2^24-nonce job, the serving shape (a block target is far tighter)
seedline() { sed -n "s/^$2 //p" "$1/seeds.txt"; }
EPOCH="$(seedline "$PACK" epoch_seed_hex)"; DAY="$(seedline "$PACK" day_seed_hex)"
[ -n "$EPOCH" ] && [ -n "$DAY" ] || { echo "no seeds.txt in $PACK" >&2; exit 2; }
jobline() { printf 'job %d %064x %s %d %d %s %s\n' "$1" "$1" "$target" $(( $1 * count )) "$count" "$EPOCH" "$DAY"; }
# power sampler first (its own pid file)
nvidia-smi -i "$DEV" --query-gpu=timestamp,power.draw,clocks.sm,temperature.gpu --format=csv,noheader,nounits -lms "$SMI_MS" > "$SMI" 2>/dev/null &
echo $! > "$PIDDIR/$RUN.smi.pid"
sleep 1
t0=$(now)
coproc WK { IGNEUM_READBACK="$MODE" "$W" --serve --pack "$PACK" --device "$DEV" --batch-log2 "$B" --race off --readback "$MODE" 2>&1; }
wkpid=$WK_PID # bash clears WK_PID when the coproc ends; the copy outlives it
echo "$wkpid" > "$PIDDIR/$RUN.worker.pid"
sent=0; done_n=0; found=0; bad=0; tready=""; tlast=""; cpu=""
: > "$LOG"
while IFS= read -r line <&"${WK[0]}"; do
t=$(now)
echo "$line" >> "$LOG"
case "$line" in
ready*) tready=$t; echo "$t ready" >> "$TL"; jobline $sent >&"${WK[1]}"; sent=$((sent + 1)); jobline $sent >&"${WK[1]}"; sent=$((sent + 1)) ;;
found*) found=$((found + 1)); echo "$t submit $line" >> "$TL"
read -r _ _ _ h _ <<< "$line"; if [[ "$h" > "$target" ]]; then bad=$((bad + 1)); fi ;;
done*) done_n=$((done_n + 1)); tlast=$t; echo "$t $line" >> "$TL"
if [ "$sent" -lt "$JOBS" ]; then jobline $sent >&"${WK[1]}"; sent=$((sent + 1)); fi
if [ "$done_n" -ge "$JOBS" ]; then
cpu=$(awk '{ printf "%.2f", ($14 + $15) / 100.0 }' "/proc/$wkpid/stat" 2>/dev/null || echo "")
echo "quit" >&"${WK[1]}"
fi ;;
error*) echo "$t $line" >> "$TL" ;;
esac
done
wait "$wkpid" 2>/dev/null || true
t1=$(now)
sleep 1
kill "$(cat "$PIDDIR/$RUN.smi.pid")" 2>/dev/null || true
[ -n "$tready" ] && [ -n "$tlast" ] || { echo "FAIL: no ready or no done line; see $LOG"; tail -5 "$LOG"; exit 1; }
[ "$done_n" = "$JOBS" ] || { echo "FAIL: $done_n of $JOBS jobs done; see $LOG"; exit 1; }
# the power samples over the two windows: nvidia-smi stamps local time "YYYY/MM/DD HH:MM:SS.mmm"
meanw() { # <from epoch s> <to epoch s>: mean W of the samples inside the window, and their count
awk -F', *' -v a="$1" -v b="$2" 'BEGIN { n = 0; s = 0 }
{ gsub("/", "-", $1); split($1, d, " "); cmd = "date -d \"" d[1] " " d[2] "\" +%s.%N"; cmd | getline ts; close(cmd); if (ts >= a && ts <= b) { s += $2; n++ } }
END { if (n) printf "%.1f %d", s / n, n; else printf "0 0" }' "$SMI"
}
read -r wProd nProd <<< "$(meanw "$t0" "$tlast")"
read -r wSteady nSteady <<< "$(meanw "$tready" "$tlast")"
hashes=$(( JOBS * count ))
wallProd=$(awk -v a="$t0" -v b="$tlast" 'BEGIN { printf "%.3f", b - a }')
wallSteady=$(awk -v a="$tready" -v b="$tlast" 'BEGIN { printf "%.3f", b - a }')
setup=$(awk -v a="$t0" -v b="$tready" 'BEGIN { printf "%.3f", b - a }')
mhsProd=$(awk -v h="$hashes" -v w="$wallProd" 'BEGIN { printf "%.2f", h / w / 1e6 }')
mhsSteady=$(awk -v h="$hashes" -v w="$wallSteady" 'BEGIN { printf "%.2f", h / w / 1e6 }')
jProd=$(awk -v w="$wProd" -v s="$wallProd" -v n="$JOBS" 'BEGIN { printf "%.1f", w * s / n }')
jSteady=$(awk -v w="$wSteady" -v s="$wallSteady" -v n="$JOBS" 'BEGIN { printf "%.1f", w * s / n }')
mhwProd=$(awk -v m="$mhsProd" -v w="$wProd" 'BEGIN { if (w > 0) printf "%.4f", m / w; else printf "n/a" }')
mhwSteady=$(awk -v m="$mhsSteady" -v w="$wSteady" 'BEGIN { if (w > 0) printf "%.4f", m / w; else printf "n/a" }')
card=$(nvidia-smi -i "$DEV" --query-gpu=name --format=csv,noheader | tr ' ' '_')
transfers=$(grep 'transfers per chunk' "$LOG" | tail -1 | sed 's/^info //')
ready=$(grep '^ready ' "$LOG" | head -1)
{
printf 'card=%s mode=%s jobs=%d batch_log2=%d hashes=%d found=%d setup_s=%s ' "$card" "$MODE" "$JOBS" "$B" "$hashes" "$found" "$setup"
printf 'prod_wall_s=%s prod_mhs=%s prod_mean_w=%s prod_j_per_batch=%s prod_mh_per_w=%s samples=%s ' "$wallProd" "$mhsProd" "$wProd" "$jProd" "$mhwProd" "$nProd"
printf 'steady_wall_s=%s steady_mhs=%s steady_mean_w=%s steady_j_per_batch=%s steady_mh_per_w=%s samples=%s ' "$wallSteady" "$mhsSteady" "$wSteady" "$jSteady" "$mhwSteady" "$nSteady"
accepted=$((found - bad))
printf 'accepted=%d over_target=%d nj_per_hash_steady=%s j_per_share_steady=%s ' "$accepted" "$bad" "$(awk -v w="$wSteady" -v m="$mhsSteady" 'BEGIN { if (m > 0) printf "%.1f", w / m * 1000; else printf "n/a" }')" "$(awk -v w="$wSteady" -v s="$wallSteady" -v a="$accepted" 'BEGIN { if (a > 0) printf "%.3f", w * s / a; else printf "n/a" }')"
printf 'smi_ms=%s worker_cpu_s=%s worker_sha=%s\n' "$SMI_MS" "${cpu:-n/a}" "$(sha256sum "$W" | cut -c1-16)"
printf 'ready: %s\n' "$ready"
printf 'transfers: %s\n' "$transfers"
} > "$ROW"
cat "$ROW"

70
tools/1p5x/x2/summarize.py Executable file
View file

@ -0,0 +1,70 @@
#!/usr/bin/env python3
"""X2: the measured rows and the arithmetic, generated from the row files (9 October 2026).
summarize.py <pairs dir> reads x2-{full,select}-r<i>.row (serve-pair.sh) and kernel-bench.row (run-pairs.sh), prints
the per-run table, the paired means, g against the full path and against the kernel timer,
and r_new = r0 * g / a for r0 in (1.6, 2.0, 2.4) with a = 1 (the brief's identity).
Every figure in docs/analysis/class-v6/1p5x/x2/README.md's measured section is this script's output.
"""
import glob, os, re, statistics, sys
def row(path):
first = open(path).readline()
return {k: v for k, v in re.findall(r'(\w+)=(\S+)', first)}
def main(d):
runs = []
for p in sorted(glob.glob(os.path.join(d, 'x2-*-r*.row'))):
r = row(p)
r['run'] = os.path.basename(p)[:-4]
runs.append(r)
keys = [('mode', 's'), ('prod_mhs', 'f'), ('steady_mhs', 'f'), ('steady_mean_w', 'f'), ('nj_per_hash_steady', 'f'), ('accepted', 'i'),
('over_target', 'i'), ('j_per_share_steady', 'f'), ('steady_j_per_batch', 'f'), ('worker_cpu_s', 's'), ('samples', 's')]
print('| run | ' + ' | '.join(k for k, _ in keys) + ' |')
print('|' + '---|' * (len(keys) + 1))
for r in runs:
print(f"| {r['run']} | " + ' | '.join(str(r.get(k, 'n/a')) for k, _ in keys) + ' |')
by = {m: [r for r in runs if r['mode'] == m] for m in ('full', 'select')}
mean = lambda m, k: statistics.mean(float(r[k]) for r in by[m])
sd = lambda m, k: statistics.stdev(float(r[k]) for r in by[m]) if len(by[m]) > 1 else 0.0
print()
print('| mode | runs | steady MH/s (sd) | production MH/s | steady W | nJ per hash (sd) | J per accepted share | accepted shares per run |')
print('|---|---|---|---|---|---|---|---|')
for m in ('full', 'select'):
if not by[m]:
continue
print(f"| {m} | {len(by[m])} | {mean(m, 'steady_mhs'):.2f} ({sd(m, 'steady_mhs'):.2f}) | {mean(m, 'prod_mhs'):.2f} | {mean(m, 'steady_mean_w'):.1f} | "
f"{mean(m, 'nj_per_hash_steady'):.1f} ({sd(m, 'nj_per_hash_steady'):.1f}) | {mean(m, 'j_per_share_steady'):.3f} | {mean(m, 'accepted'):.0f} |")
if not (by['full'] and by['select']):
return
# paired: run i of each mode was taken back to back (ABBA order)
pairs = []
for f in by['full']:
i = f['run'].split('-r')[-1]
s = [x for x in by['select'] if x['run'].split('-r')[-1] == i]
if s:
pairs.append((float(s[0]['nj_per_hash_steady']) / float(f['nj_per_hash_steady']), float(s[0]['steady_mhs']) / float(f['steady_mhs'])))
g_serve = mean('select', 'nj_per_hash_steady') / mean('full', 'nj_per_hash_steady')
rate = mean('select', 'steady_mhs') / mean('full', 'steady_mhs')
print()
print(f"paired g (select over full, nJ per hash) per pair: {', '.join(f'{g:.4f}' for g, _ in pairs)}; mean of means {g_serve:.4f}")
print(f"paired rate ratio (select over full) per pair: {', '.join(f'{x:.4f}' for _, x in pairs)}; mean of means {rate:.4f}")
kb = os.path.join(d, 'kernel-bench.row')
g_yard = None
if os.path.exists(kb):
k = {a: b for a, b in re.findall(r'(\w+)=(\S+)', open(kb).read())}
g_yard = mean('select', 'nj_per_hash_steady') / float(k['nj_per_hash'])
print(f"kernel timer: {k['kernel_mhs']} MH/s, {k['loaded_mean_w']} W loaded, {k['nj_per_hash']} nJ per hash, fingerprint {k['fingerprint']}")
print(f"g against the kernel timer (select serving over --bench): {g_yard:.4f}; full serving over --bench: {mean('full', 'nj_per_hash_steady') / float(k['nj_per_hash']):.4f}")
print()
print('| r0 (the brief\'s current ratio) | r_new with g = select/full | r_new with g = select/kernel timer |')
print('|---|---|---|')
for r0 in (1.6, 2.0, 2.4):
print(f"| {r0} | {r0 * g_serve:.3f} | {r0 * g_yard:.3f} |" if g_yard else f"| {r0} | {r0 * g_serve:.3f} | n/a |")
print('(a = 1: the specialist side is not touched by a host-side result path; the identity r_new = r0 * g / a, the brief.)')
if __name__ == '__main__':
main(sys.argv[1])

435
tools/1p5x/x2/x2-equiv.py Executable file
View file

@ -0,0 +1,435 @@
#!/usr/bin/env python3
"""X2 (the 1.5x programme): serving-mode result compaction, the equivalence driver (9 October 2026).
The compaction under test is the CUDA kit worker's GPU select pass (proto-cuda/nvrtc/worker.cpp, igneum_select, review B
F10, an ancestor of the frozen control 1a938abe4): the hits come back behind an atomic counter with 34 sentinel words, a
chunk with more than IG_MAX_HITS (256) hits falls back to the full read. The uncompacted path is `--readback full` (8 bytes
per nonce over the bus, the host scans). This driver runs ONE worker binary in serve mode, exactly as the miner drives it
(job lines on stdin, a `prepare` for the next day's pack before the day switch), over a deterministic job plan per same-work
context, and writes the normalized protocol output (found lines as `found <job> <nonce> <hash16>`, done lines without their
wall ms, error lines). The EXPECTED output is computed from the CPU references alone (igneum-pow hash-bound, the context's
references/ref-D.txt and ref-D1.txt): for each job, every nonce n in [start, start + count) with ref[n] <= target, in nonce
order, then `done <job> <count>`. Equivalence = full output == select output == expected, byte for byte, which is every
nonce's verdict (found or not) at every target and every found nonce's digest.
The plan per context (seeded by the context id, so every mode and every run sees the same jobs):
shape "prod" (--batch-log2 22, the worker's and the miner's default): phase 1 [0, 2^20) at the sentinel targets
all-ones (every nonce a hit: the overflow fallback, every digest), T256 (exactly 256 hits in the job's one chunk: the
hit table full), T257 (257 hits: overflow by one), 0, Teq (the job's smallest hash: the <= edge, one hit) and Teq - 1;
phase 2 across the day switch: [2^20, 1572864) on pack D at all-ones, `prepare` pack D1, [1572864, 2^21) on D1 at
all-ones; phase 3 [2^21, 3 * 2^20) on D1 at all-ones and at its own T256.
shape "fuzz" (--batch-log2 10: 1,024-nonce chunks, so a q/4 target mixes overflow and table chunks inside one job):
1,048,576 nonces of seeded random jobs (32-aligned starts, counts 32 to 65,536 in steps of 32, never across the day
boundary), targets drawn from {0, all-ones, 2^k - 1 for k in 50..63, q/8, q/4, an in-job hash (the <= edge), that hash
minus 1}, the day-D jobs in nonce order, then the 8,192-nonce boundary block ([1568768, 1572864) on D, `prepare` D1,
[1572864, 1576960) on D1, each at all-ones and at q/4), then the day-D1 jobs.
Usage (on the GPU host, under a pid file; never on the Mac):
x2-equiv.py run --worker <bin> --ctx <same-work context dir> --mode full|select|fault --shape prod|fuzz --out <dir>
x2-equiv.py expect --ctx <dir> --shape prod|fuzz --out <dir> (the expected file from the references alone)
x2-equiv.py compare --out <dir> --ctx-id <id> --shape prod|fuzz (verdict JSON over the three files)
x2-equiv.py --self-test (the driver's own known-failed cases, a fake worker)
`fault` = select with IGNEUM_READBACK_FAULT_TEST=1 (the worker skips the hit counter reset from the second chunk on): the
known-failed run; compare must FAIL it before a select run counts.
"""
import hashlib, json, os, random, subprocess, sys, tempfile, threading, time
ALL = (1 << 64) - 1
BOUNDARY = 1572864
BLOCK_LO, BLOCK_HI = BOUNDARY - 4096, BOUNDARY + 4096
def read_refs(ctx):
j = json.load(open(os.path.join(ctx, 'job-context.json')))
ref = []
for k in ('D', 'D1'):
with open(os.path.join(ctx, j['references'][k])) as f:
for line in f:
p = line.split()
if len(p) >= 2:
ref.append((int(p[0]), int(p[1], 16)))
ref.sort()
assert [n for n, _ in ref[:3]] == [0, 1, 2] and len(ref) == 3 * (1 << 20), 'references must cover [0, 3 * 2^20)'
return [h for _, h in ref]
def ctx_fields(ctx):
j = json.load(open(os.path.join(ctx, 'job-context.json')))
toks = j['job_line'].split()
extra = ' '.join(t for t in toks if t.startswith('class=') or t.startswith('era='))
packs = {}
for k in ('D', 'D1'):
pj = json.load(open(os.path.join(ctx, j['packs'][k], 'program.json')))
packs[k] = {'dir': os.path.abspath(os.path.join(ctx, j['packs'][k])), 'epoch': pj['seed_bytes'], 'day': pj['dataset']['day_bytes'],
'program_id': pj.get('program_id')}
assert j['boundary_nonce'] == BOUNDARY
return {'id': j['run_id'], 'prehash': j['prehash'], 'extra': extra, 'packs': packs, 'object_id': j.get('object_id')}
def kth(ref, lo, hi, k): # the k-th smallest hash in [lo, hi), 1-based
return sorted(ref[lo:hi])[k - 1]
def plan(ctx_id, ref, shape):
"""[(pair, start, count, target, label)] with pair 'D' or 'D1'; a 'prepare' entry switches to D1."""
jobs = []
if shape == 'prod':
# The same range at several targets is the same work twice in a row, which the worker's stale-output detector
# reads as a kernel that did not run (worker.cpp: equal first/middle/last signature of consecutive chunks, error
# and exit 3; finding X2-F1, the same on both paths). A 64-nonce separator job of other nonces at target 0 sits
# between repeats, so every repeat is a real one and the detector stays armed.
sep = [0]
def separator(pair, base):
jobs.append((pair, base + 64 * sep[0], 64, 0, 'sep')); sep[0] += 1
lo, hi = 0, 1 << 20
t256, t257, teq = kth(ref, lo, hi, 256), kth(ref, lo, hi, 257), kth(ref, lo, hi, 1)
for i, (t, lab) in enumerate(((ALL, 'p1-all'), (t256, 'p1-t256'), (t257, 'p1-t257'), (0, 'p1-zero'), (teq, 'p1-teq'), (max(teq - 1, 0), 'p1-teq-1'))):
if i:
separator('D', 1 << 20)
jobs.append(('D', lo, hi - lo, t, lab))
jobs.append(('D', 1 << 20, BOUNDARY - (1 << 20), ALL, 'p2a-all'))
jobs.append(('prepare',))
jobs.append(('D1', BOUNDARY, (1 << 21) - BOUNDARY, ALL, 'p2b-all'))
lo, hi = 1 << 21, 3 << 20
jobs.append(('D1', lo, hi - lo, ALL, 'p3-all'))
separator('D1', BOUNDARY)
jobs.append(('D1', lo, hi - lo, kth(ref, lo, hi, 256), 'p3-t256'))
return jobs
if shape == 'kf':
# the known-failed plan: 2,048-nonce jobs at a 1/16 target over 1,024-nonce chunks, so the counter-reset fault
# (IGNEUM_READBACK_FAULT_TEST) leaves the first chunk's hits in the table under 256 and they print again
return [('D', 2048 * i, 2048, (1 << 60) - 1, 'kf') for i in range(4)]
assert shape == 'fuzz'
rng = random.Random('x2-fuzz/' + ctx_id)
q8, q4 = (1 << 61) - 1, (1 << 62) - 1
total, want = 0, 1 << 20
want -= 4 * 4096 # the boundary block's four jobs of 4,096 below count toward the 1,048,576
d, d1 = [], []
while total < want:
cnt = min(32 * rng.randint(1, 2048), want - total)
s = 32 * rng.randrange(0, (3 << 20) // 32)
if s < BOUNDARY < s + cnt:
cnt = BOUNDARY - s
if s + cnt > (3 << 20):
cnt = (3 << 20) - s
if cnt <= 0:
continue
kind = rng.randrange(8)
if kind == 0: t = 0
elif kind == 1: t = ALL
elif kind == 2: t = (1 << rng.randint(50, 63)) - 1
elif kind == 3: t = q8
elif kind == 4: t = q4
else:
h = ref[s + rng.randrange(cnt)]
t = h if kind in (5, 6) else max(h - 1, 0)
(d if s < BOUNDARY else d1).append((s, cnt, t, f'fz{kind}'))
total += cnt
for s, c, t, lab in sorted(d):
jobs.append(('D', s, c, t, lab))
jobs.append(('D', BLOCK_LO, 4096, ALL, 'blk-d-all'))
jobs.append(('D', BLOCK_LO, 4096, q4, 'blk-d-q4'))
jobs.append(('prepare',))
jobs.append(('D1', BOUNDARY, 4096, ALL, 'blk-d1-all'))
jobs.append(('D1', BOUNDARY, 4096, q4, 'blk-d1-q4'))
for s, c, t, lab in sorted(d1):
jobs.append(('D1', s, c, t, lab))
n = sum(j[2] for j in jobs if j[0] != 'prepare')
assert n == 1 << 20, n
return jobs
def expected_lines(ref, jobs):
seq = 0
for j in jobs:
if j[0] == 'prepare':
continue
seq += 1
_, s, c, t, _ = j
for n in range(s, s + c):
if ref[n] <= t:
yield f'found {seq} {n} {ref[n]:016x}\n'
yield f'done {seq} {c}\n'
def norm(line):
p = line.split()
if not p:
return None
if p[0] == 'found' and len(p) >= 4:
return f'found {p[1]} {int(p[2])} {p[3].lower().zfill(16)}\n'
if p[0] == 'done' and len(p) >= 3:
return f'done {p[1]} {p[2]}\n'
if p[0] == 'error':
return ' '.join(p[:2]) + ' error\n'
return None
def run(a):
ctx = ctx_fields(a.ctx)
ref = read_refs(a.ctx)
jobs = plan(ctx['id'], ref, a.shape)
batch = {'prod': 22, 'fuzz': 10, 'kf': 10}[a.shape]
mode = 'select' if a.mode == 'fault' else a.mode
env = dict(os.environ)
env['IGNEUM_READBACK'] = mode
env.pop('IGNEUM_READBACK_FAULT_TEST', None)
if a.mode == 'fault':
env['IGNEUM_READBACK_FAULT_TEST'] = '1'
cmd = ([a.worker] if not a.worker.endswith('.py') else [sys.executable, a.worker]) + \
['--serve', '--pack', ctx['packs']['D']['dir'], '--device', '0', '--batch-log2', str(batch), '--race', 'off', '--readback', mode]
os.makedirs(a.out, exist_ok=True)
tag = f"{ctx['id']}.{a.shape}.{a.mode}"
outp = os.path.join(a.out, tag + '.norm')
logp = os.path.join(a.out, tag + '.log')
t0 = time.time()
p = subprocess.Popen(cmd, stdin=subprocess.PIPE, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, text=True, bufsize=1, env=env)
if a.pid_dir:
open(os.path.join(a.pid_dir, f'x2-{tag}.worker.pid'), 'w').write(f'{p.pid} {time.strftime("%Y-%m-%dT%H:%M:%SZ", time.gmtime())} x2 {tag}\n')
prepared = threading.Event()
ready = threading.Event()
state = {'ready': None, 'prepared': None, 'transfers': None, 'errors': [], 'need': 0}
fo = open(outp, 'w'); lo = open(logp, 'w')
def reader():
for line in p.stdout:
if not line.startswith('found '):
lo.write(line)
if line.startswith('ready'):
state['ready'] = line.strip(); ready.set()
elif line.startswith('prepared'):
state['prepared'] = line.strip()[:400]; prepared.set()
elif line.startswith('prepare-failed'):
state['errors'].append(line.strip()[:300]); prepared.set()
elif 'transfers per chunk' in line:
state['transfers'] = line.strip()
n = norm(line)
if n:
fo.write(n)
if n.startswith('error'):
state['errors'].append(line.strip()[:300])
ready.set(); prepared.set()
th = threading.Thread(target=reader, daemon=True); th.start()
ready.wait(600)
seq = 0
pair = 'D'
for j in jobs:
if j[0] == 'prepare':
pk = ctx['packs']['D1']
p.stdin.write(f"prepare {pk['epoch']} {pk['day']} {pk['dir']} {ctx['extra']}\n"); p.stdin.flush()
# the worker reports a finished prepare only when it reads its next line (worker.cpp, "between lines")
for _ in range(1200):
if prepared.is_set():
break
p.stdin.write('noop\n'); p.stdin.flush(); time.sleep(0.5)
pair = 'D1'
continue
seq += 1
assert j[0] == pair
pk = ctx['packs'][pair]
_, s, c, t, _ = j
try:
p.stdin.write(f"job {seq} {ctx['prehash']} {t:016x} {s} {c} {pk['epoch']} {pk['day']} {ctx['extra']}\n")
except BrokenPipeError:
state['errors'].append(f'the worker closed its stdin before job {seq} (exit {p.poll()})'); break
try:
p.stdin.write('quit\n'); p.stdin.flush()
except BrokenPipeError:
pass
try:
p.stdin.close()
except Exception:
pass
p.wait(timeout=3600); th.join(60); fo.close(); lo.close()
if a.pid_dir:
try: os.remove(os.path.join(a.pid_dir, f'x2-{tag}.worker.pid'))
except OSError: pass
meta = {'ctx': ctx['id'], 'object_id': ctx['object_id'], 'shape': a.shape, 'mode': a.mode, 'batch_log2': batch, 'jobs': seq,
'nonces': sum(j[2] for j in jobs if j[0] != 'prepare'), 'worker': a.worker, 'worker_sha256': sha_file(a.worker),
'ready': state['ready'], 'prepared': state['prepared'], 'transfers': state['transfers'], 'errors': state['errors'][:10],
'norm_sha256': sha_file(outp), 'seconds': round(time.time() - t0, 1), 'exit': p.returncode,
'at': time.strftime('%Y-%m-%dT%H:%M:%SZ', time.gmtime())}
json.dump(meta, open(os.path.join(a.out, tag + '.meta.json'), 'w'), indent=1)
print(json.dumps(meta))
return 0
def expect(a):
ctx = ctx_fields(a.ctx)
ref = read_refs(a.ctx)
jobs = plan(ctx['id'], ref, a.shape)
os.makedirs(a.out, exist_ok=True)
outp = os.path.join(a.out, f"{ctx['id']}.{a.shape}.expected.norm")
with open(outp, 'w') as f:
for line in expected_lines(ref, jobs):
f.write(line)
labels = {}
for j in jobs:
if j[0] != 'prepare':
labels[j[4]] = labels.get(j[4], 0) + 1
hits = sum(1 for line in open(outp) if line.startswith('found'))
meta = {'ctx': ctx['id'], 'shape': a.shape, 'jobs': sum(1 for j in jobs if j[0] != 'prepare'), 'nonces': sum(j[2] for j in jobs if j[0] != 'prepare'),
'found_lines': hits, 'labels': labels, 'norm_sha256': sha_file(outp),
'references': {k: sha_file(os.path.join(a.ctx, v)) for k, v in json.load(open(os.path.join(a.ctx, 'job-context.json')))['references'].items()}}
json.dump(meta, open(os.path.join(a.out, f"{ctx['id']}.{a.shape}.expected.meta.json"), 'w'), indent=1)
print(json.dumps(meta))
return 0
def first_diff(pa, pb):
with open(pa) as fa, open(pb) as fb:
i = 0
while True:
la, lb = fa.readline(), fb.readline()
i += 1
if la != lb:
return {'line': i, 'expected': la.strip(), 'got': lb.strip()}
if not la:
return None
def count_diff(pa, pb):
a = set(open(pa)); b = set(open(pb))
return len(a - b), len(b - a)
def compare(a):
base = os.path.join(a.out, f'{a.ctx_id}.{a.shape}')
exp = base + '.expected.norm'
res = {'ctx': a.ctx_id, 'shape': a.shape, 'expected_sha256': sha_file(exp), 'modes': {}}
for mode in ('fault', 'full', 'select'):
pth = f'{base}.{mode}.norm'
if not os.path.exists(pth):
res['modes'][mode] = 'NOT RUN'
continue
d = first_diff(exp, pth)
miss, extra = count_diff(exp, pth) if d else (0, 0)
res['modes'][mode] = {'sha256': sha_file(pth), 'equal_to_expected': d is None, 'first_difference': d, 'lines_missing': miss, 'lines_extra': extra}
ok_fault = isinstance(res['modes'].get('fault'), dict) and not res['modes']['fault']['equal_to_expected']
ok_full = isinstance(res['modes'].get('full'), dict) and res['modes']['full']['equal_to_expected']
ok_sel = isinstance(res['modes'].get('select'), dict) and res['modes']['select']['equal_to_expected']
if res['modes'].get('fault') == 'NOT RUN':
ok_fault = None
res['known_failed_flagged'] = ok_fault
res['verdict'] = 'EQUAL' if (ok_full and ok_sel and ok_fault is not False) else 'DIFFERENT'
json.dump(res, open(base + '.compare.json', 'w'), indent=1)
print(json.dumps(res))
return 0 if res['verdict'] == 'EQUAL' else 1
def sha_file(p):
h = hashlib.sha256()
with open(p, 'rb') as f:
for b in iter(lambda: f.read(1 << 20), b''):
h.update(b)
return h.hexdigest()
FAKE = r'''import os, sys, json
# A fake serve worker over a fake hash (n * 0x9e3779b97f4a7c15 mod 2^64). X2_FAULT selects one injected fault:
# drop257 the 257th hit of a chunk is lost (an overflow handled as a full table)
# digest one found hash word is wrong (the run's first hit)
# stale the previous chunk's hits come back again (the counter-reset fault)
# noswitch the D1 jobs are hashed with D's words (the day switch missed)
# sentinel a target equal to a hash is treated as < (the <= edge)
F = os.environ.get("X2_FAULT", "")
B = 1 << int(sys.argv[sys.argv.index("--batch-log2") + 1])
def h(n, day):
v = (n * 0x9e3779b97f4a7c15) % (1 << 64)
return v ^ (0x5555 if day == "D1" else 0)
print("ready fake-gpu batch %d readback %s" % (B, os.environ.get("IGNEUM_READBACK")), flush=True)
cur, prep, table, nchunks, flipped = "D", None, [], 0, False
for line in sys.stdin:
p = line.split()
if not p: continue
if p[0] == "quit": break
if p[0] == "prepare": prep = "D1"; continue
if p[0] == "noop":
if prep: print("prepared fake", flush=True); prep = "ready"
continue
if p[0] != "job": continue
seq, t, s, c, day = p[1], int(p[3], 16), int(p[4]), int(p[5]), p[7]
pair = "D1" if day.endswith("fb50000000000000") or day.endswith("fff50000000000000") or day == "d1" else "D"
hp = "D" if F == "noswitch" else pair
off = 0
while off < c:
n0, m = s + off, min(B, c - off); hits = []
for n in range(n0, n0 + m):
v = h(n, hp)
ok = v < t if F == "sentinel" else v <= t
if ok:
if F == "digest" and not flipped: v ^= 1; flipped = True
hits.append((n, v))
if F == "drop257" and len(hits) == 257: hits = hits[:256]
if F == "stale":
# worker.cpp under IGNEUM_READBACK_FAULT_TEST: the counter is reset only before the first chunk; the table keeps
# (chunk index, hash) in arrival order up to 256; over 256 the chunk falls back to the full read (correct output)
if nchunks == 0: table = []
table = table + [(n - n0, v) for n, v in hits]
out = hits if len(table) > 256 else [(n0 + i, v) for i, v in table[:256]]
else:
out = hits
for n, v in sorted(out): print("found %s %d %016x" % (seq, n, v))
nchunks += 1; off += m
print("done %s %d 1.0" % (seq, c), flush=True)
'''
def self_test():
"""The driver's own known-failed cases: a fake context and a fake worker; every injected fault must be DIFFERENT and the
clean worker EQUAL. Runs anywhere with Python (no GPU); on the GPU host before the real runs."""
d = tempfile.mkdtemp(prefix='x2-self-')
ctx = os.path.join(d, 'ctx'); os.makedirs(os.path.join(ctx, 'references'))
for k, day in (('D', '69676e65756d2d6461792ffa50000000000000'), ('D1', '69676e65756d2d6461792ffb50000000000000')):
pd = os.path.join(ctx, 'packs', k); os.makedirs(pd)
json.dump({'seed_bytes': 'aa' * 32, 'program_id': '0xfake', 'dataset': {'day_bytes': day}}, open(os.path.join(pd, 'program.json'), 'w'))
json.dump({'run_id': 'self', 'prehash': '00' * 31 + '01', 'job_line': 'job <seq> x class=v6 era=bb', 'boundary_nonce': BOUNDARY,
'packs': {'D': 'packs/D', 'D1': 'packs/D1'}, 'references': {'D': 'references/ref-D.txt', 'D1': 'references/ref-D1.txt'}},
open(os.path.join(ctx, 'job-context.json'), 'w'))
with open(os.path.join(ctx, 'references', 'ref-D.txt'), 'w') as f:
for n in range(BOUNDARY):
f.write(f'{n} {(n * 0x9e3779b97f4a7c15) % (1 << 64):016x}\n')
with open(os.path.join(ctx, 'references', 'ref-D1.txt'), 'w') as f:
for n in range(BOUNDARY, 3 << 20):
f.write(f'{n} {((n * 0x9e3779b97f4a7c15) % (1 << 64)) ^ 0x5555:016x}\n')
w = os.path.join(d, 'fakeworker.py'); open(w, 'w').write(FAKE)
out = os.path.join(d, 'out')
class A: pass
fails = 0
for shape in ('kf', 'prod', 'fuzz'):
a = A(); a.ctx = ctx; a.shape = shape; a.out = out
expect(a)
cases = [('', 'select', True)] + [(f, 'select', False) for f in ('drop257', 'digest', 'stale', 'noswitch', 'sentinel')]
for fault, mode, want_equal in cases:
os.environ['X2_FAULT'] = fault
a = A(); a.worker = w; a.ctx = ctx; a.mode = mode; a.shape = shape; a.out = out; a.pid_dir = None
run(a)
c = A(); c.out = out; c.ctx_id = 'self'; c.shape = shape
base = os.path.join(out, f'self.{shape}')
d0 = first_diff(base + '.expected.norm', base + '.select.norm')
got_equal = d0 is None
# a fault the shape cannot reach (drop257 needs a 257-hit chunk: prod only) is reported, not failed
# and stale is masked by the overflow fallback once the un-reset count passes 256 (prod, fuzz): kf is its shape
reachable = (fault in ('', 'stale', 'digest')) if shape == 'kf' else not ((fault == 'drop257' and shape != 'prod') or fault == 'stale')
if reachable and got_equal != want_equal:
print(f'SELF-TEST FAIL: shape {shape} fault "{fault or "none"}": equal={got_equal}, wanted {want_equal}'); fails += 1
else:
print(f'self-test: shape {shape} fault "{fault or "none"}": {"EQUAL" if got_equal else "DIFFERENT"} at {d0}')
os.environ.pop('X2_FAULT', None)
print('SELF-TEST ' + ('PASS: the clean fake is EQUAL on every shape; drop-257, a wrong digest, stale hits, a missed day switch and a < for <= are each DIFFERENT' if not fails else f'FAIL ({fails})'))
return 1 if fails else 0
if __name__ == '__main__':
if '--self-test' in sys.argv:
sys.exit(self_test())
import argparse
ap = argparse.ArgumentParser()
ap.add_argument('cmd', choices=['run', 'expect', 'compare'])
ap.add_argument('--worker'); ap.add_argument('--ctx'); ap.add_argument('--mode', choices=['full', 'select', 'fault'])
ap.add_argument('--shape', choices=['prod', 'fuzz', 'kf'], required=True); ap.add_argument('--out', required=True)
ap.add_argument('--ctx-id'); ap.add_argument('--pid-dir')
a = ap.parse_args()
sys.exit({'run': run, 'expect': expect, 'compare': compare}[a.cmd](a))