P32 mapped-host block-epoch pilot

Reading a mapped host epoch once per block can avoid copy-stream starvation while amortizing coherent-memory cost.

Claim statusNO END-TO-END MINING ADVANTAGE DEMONSTRATED

Diagnostic work, exact algebra and local capabilities are not treated as end-to-end mining advantage.

Date2026-08-16

HardwareNVIDIA GeForce RTX 5080 · exact CUDA SHA-256d verifier

Run scopePreregistered cancellation campaign with sealed validation and audit

ReproducibilityBuild the sealed pilot, verify exact output and zero spills, then apply the latency opening gate before any extended benchmark.

01 / Question & hypothesis

What was tested?

Reading a mapped host epoch once per block can avoid copy-stream starvation while amortizing coherent-memory cost.

02 / Scientific basis

Why the test is meaningful

A block-level mapped read removes the separate control launch, but the running grid may cache, delay or serialize host-coherent access enough to miss the latency objective.

epoch_block=volatile mapped-host load once per blockretention pilot=throughput(candidate)open full campaign only if response≤15 ms

FIGURE / RESULT READINGVisual reading of the published metrics and gates for EXP-095; it summarizes the registered result, not a mining advantage.

03 / Method

How it was tested

Compile an exact, zero-spill P32 candidate with one mapped-host epoch load per block; run a bounded throughput and cancellation pilot before authorizing the 48-run campaign.

04 / Observed result

What happened

Pilot throughput5.153 GH/s

Cancel response203.54 ms

Full campaignNOT OPENED

The pilot sustained 5.153 GH/s and observed the signal, but response was 203.54 ms. The preregistered 48-run campaign was not opened.

05 / Validation

Exactness and statistical controls

The candidate compiled without spills and the control signal was eventually visible. The pilot latency failed by more than an order of magnitude.

06 / Interpretation

What the result means

Block-level mapped-host polling is rejected early. A future hypothesis must change load semantics, not repeat this campaign.

Limitations

Only a bounded pilot was warranted after the decisive failure.
No precise full-campaign retention interval was produced.
The result is not a statement about ASIC control paths.

07 / Reproduction

Evidence trail

Build the sealed pilot, verify exact output and zero spills, then apply the latency opening gate before any extended benchmark.

Canonical variants

CANONICAL-EXP-171PREREGISTERED-CAMPAIGNSEALED-AUDIT

Source: internally audited canonical reports. Local filesystem structure, private headers and operational identifiers are excluded from publication.