Independent replication of P32 warp cancellation
The exact P32 warp-broadcast primitive replicates responsive cancellation with throughput retention high enough for default selection.
Diagnostic work, exact algebra and local capabilities are not treated as end-to-end mining advantage.
What was tested?
The exact P32 warp-broadcast primitive replicates responsive cancellation with throughput retention high enough for default selection.
Why the test is meaningful
A sub-percent control cost requires unseen headers, frozen binaries and repeated paired measurements to distinguish a real penalty from launch noise.
80 paired runs=8 headers×5 repetitions×2 armsresponse gate≤15 msdefault gate: retention CI95 supports ≥0.9995How it was tested
Freeze control and warp-epoch binaries, then use eight fresh headers, five repetitions per header, 80 paired runs, eight cancellations and full 2^32 exactness per variant.
What happened
Cancellation passed 8/8 at 6.49–14.66 ms. Retention was 0.998948 with CI95 0.998676–0.999219; the candidate was favorable on 0/8 datasets.
Exactness and statistical controls
Both variants covered 2^32, produced zero discrepancies and compiled with 44 registers and zero spills. The ~0.105% penalty reproduced independently.
What the result means
P32 warp cancellation is technically valid and responsive, but not throughput-neutral enough to become the default. It remains engineering evidence, not a pool advantage.
Limitations
- The result is specific to the tested GPU stack.
- Energy was not a promotion endpoint here.
- A latency option may have operational value without being a mining advantage.
Evidence trail
Use the frozen binaries and eight unseen headers; require all exactness and latency gates before interpreting the paired retention interval.
Canonical variants
CANONICAL-EXP-167PREREGISTERED-CAMPAIGNSEALED-AUDITSource: internally audited canonical reports. Local filesystem structure, private headers and operational identifiers are excluded from publication.