6cb0140fae
The public history carried kat/__pycache__/mlkem768_reference.cpython-314.pyc, a compiled Python artifact embedding the operator's absolute local path. Text secret scanners do not read compiled binaries, which is exactly how it slipped through, and removing it from the tip would have left it reachable through the old root commits. So this repository is republished from a single clean root. This root also carries, from the previously unpublished line of work: - corrected LICENSE year, LICENSING.md, VERIFY-POLICY.md, and CITATIONS-UNRESOLVED.md remeasured 2026-08-11 (101 paths, README aligned) - O-018: run_consensus_verification.py ran 19 of 29 models and reported PASS; it now runs all 29, and computemarket_smt.py gains resolveByTimeout / reclaimUnsettled cases plus a negative control - O-006: the word 'audited' removed from next to Bouncy Castle, twice, after a concurrent edit resurrected it - O-014: prior art named and dated - Algorand's native falcon_verify shipped about ten months before AERE's precompiles; the primacy claim is withdrawn where it was implied - bench/ scripts parametrized so they actually run for an outsider (the earlier textual sanitization left $STAGING unexpanded inside Python strings) - AIP-2/AIP-3 errata with measured figures, spec remeasurements at 2026-08-01, and the spec-zk-stack retractions (owner is an operational key, not the Foundation; 'maximally sound' withdrawn; aggregator V1 deprecated) The redacted bench-host environment files from the sanitized line are kept exactly as published; the unredacted local variants are not carried.
74 lines
5.2 KiB
Plaintext
74 lines
5.2 KiB
Plaintext
=======================================================
|
|
BLOCK-STM END-TO-END BLOCK-PRODUCTION THROUGHPUT
|
|
machine threads available : 16
|
|
txns per batch : 20000
|
|
simulated EVM cost per tx : 20 keccak-rounds
|
|
timing : median of 5 runs (ms)
|
|
pipeline : exec (parallel) + commit (serial) + state-root (serial)
|
|
serial-equivalence : par state == seq state AND keccak roots identical, EVERY run
|
|
=======================================================
|
|
|
|
### 0% designed conflict (disjoint accounts, 64 contracts spread)
|
|
committed state keys : 49637
|
|
sequential exec 166.95 ms | root 9.45 ms | END-TO-END 176.40 ms | root-share 5.4%
|
|
sequential end-to-end TPS: 113378
|
|
W | exec commit root e2e_ms | execSpd e2eSpd | TPS abort% | root%e2e | equiv
|
|
1 | 192.12 8.69 9.24 210.04 | 0.87x 0.84x | 95220 0.0% | 8.5% | PASS
|
|
2 | 108.62 8.72 10.23 127.57 | 1.54x 1.38x | 156775 0.0% | 14.9% | PASS
|
|
4 | 71.23 8.45 8.97 88.65 | 2.34x 1.99x | 225609 0.0% | 19.7% | PASS
|
|
8 | 38.84 7.82 8.10 54.76 | 4.30x 3.22x | 365254 0.1% | 29.1% | PASS
|
|
16 | 32.50 7.33 8.23 48.06 | 5.14x 3.67x | 416134 0.2% | 32.4% | PASS
|
|
Amdahl: serial tail (commit+root) = 16.54 ms; end-to-end speedup ceiling @16w = 3.65x, @infinite cores = 10.67x
|
|
|
|
### 10% designed conflict (100 hot accts, 16 contracts)
|
|
committed state keys : 18342
|
|
sequential exec 170.92 ms | root 2.87 ms | END-TO-END 173.79 ms | root-share 1.7%
|
|
sequential end-to-end TPS: 115081
|
|
W | exec commit root e2e_ms | execSpd e2eSpd | TPS abort% | root%e2e | equiv
|
|
1 | 208.57 4.91 3.13 216.61 | 0.82x 0.80x | 92331 0.0% | 3.7% | PASS
|
|
2 | 111.66 5.00 3.57 120.24 | 1.53x 1.45x | 166336 0.1% | 7.1% | PASS
|
|
4 | 68.63 5.13 3.34 77.11 | 2.49x 2.25x | 259374 0.2% | 11.0% | PASS
|
|
8 | 41.14 3.93 3.27 48.33 | 4.15x 3.60x | 413814 0.5% | 14.9% | PASS
|
|
16 | 31.59 5.02 3.26 39.87 | 5.41x 4.36x | 501666 0.8% | 20.8% | PASS
|
|
Amdahl: serial tail (commit+root) = 9.54 ms; end-to-end speedup ceiling @16w = 4.20x, @infinite cores = 18.21x
|
|
|
|
### 50% designed conflict (20 hot accts, 4 contracts)
|
|
committed state keys : 9006
|
|
sequential exec 164.90 ms | root 1.52 ms | END-TO-END 166.41 ms | root-share 0.9%
|
|
sequential end-to-end TPS: 120182
|
|
W | exec commit root e2e_ms | execSpd e2eSpd | TPS abort% | root%e2e | equiv
|
|
1 | 194.05 2.72 1.57 198.34 | 0.85x 0.84x | 100839 0.0% | 2.2% | PASS
|
|
2 | 111.70 2.63 1.73 116.05 | 1.48x 1.43x | 172332 2.0% | 3.8% | PASS
|
|
4 | 64.83 2.23 1.40 68.46 | 2.54x 2.43x | 292161 5.2% | 5.3% | PASS
|
|
8 | 44.69 2.25 1.57 48.51 | 3.69x 3.43x | 412270 9.1% | 7.9% | PASS
|
|
16 | 38.28 2.03 1.59 41.90 | 4.31x 3.97x | 477368 13.5% | 8.6% | PASS
|
|
Amdahl: serial tail (commit+root) = 3.19 ms; end-to-end speedup ceiling @16w = 3.92x, @infinite cores = 52.13x
|
|
|
|
### PATHOLOGICAL ~100% (single hot slot, no parallelism)
|
|
committed state keys : 1996
|
|
sequential exec 164.73 ms | root 0.29 ms | END-TO-END 165.03 ms | root-share 0.2%
|
|
sequential end-to-end TPS: 121191
|
|
W | exec commit root e2e_ms | execSpd e2eSpd | TPS abort% | root%e2e | equiv
|
|
1 | 191.33 0.06 0.35 191.74 | 0.86x 0.86x | 104306 0.0% | 0.2% | PASS
|
|
2 | 115.17 0.05 0.30 115.51 | 1.43x 1.43x | 173140 6.4% | 0.3% | PASS
|
|
4 | 71.82 0.07 0.35 72.23 | 2.29x 2.28x | 276876 14.0% | 0.6% | PASS
|
|
8 | 57.49 0.04 0.30 57.83 | 2.87x 2.85x | 345848 20.6% | 0.6% | PASS
|
|
16 | 61.98 0.04 0.41 62.44 | 2.66x 2.64x | 320331 24.2% | 0.7% | PASS
|
|
Amdahl: serial tail (commit+root) = 0.39 ms; end-to-end speedup ceiling @16w = 2.66x, @infinite cores = 422.21x
|
|
|
|
=======================================================
|
|
SERIAL-EQUIVALENCE: 100 checks, 0 mismatches
|
|
RESULT: PASS -- every parallel run committed the exact sequential state + root
|
|
=======================================================
|
|
|
|
HONEST READING:
|
|
- execSpd is the EXECUTION-PHASE speedup (comparable to the documented ~9.4x).
|
|
- e2eSpd is the END-TO-END block-production speedup. It is LOWER, because the
|
|
commit (state materialization) and state-root (keccak) phases are serial and
|
|
do not shrink with more workers -- the Amdahl serial tail.
|
|
- TPS is the real end-to-end transactions/second of THIS pipeline on THIS box
|
|
(shared-vCPU cloud). It is a measured rollup-batch number, NOT the 273k TPS
|
|
documented-conditions marketing ceiling and does not reach or imply it.
|
|
- Under total conflict (pathological) there is no parallelism; end-to-end can be
|
|
<= 1x. That is the correct, honest behavior of Block-STM.
|