EXP-063NEGATIVE
Canonical GPU engineering / SHA-256d

Forced 2×/4× inter-nonce unrolling

Forcing two or four nonces into one unrolled body can expose enough instruction-level parallelism to exceed the baseline.

Claim statusNO END-TO-END MINING ADVANTAGE DEMONSTRATED

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

Date2026-08-14
HardwareNVIDIA GeForce RTX 5080 · exact SHA-256d verifier
Run scopePreregistered campaign with sealed result and audit
ReproducibilityFreeze U2 after discovery, use unseen headers and audit exact work, SASS and register counts.
01 / Question & hypothesis

What was tested?

Forcing two or four nonces into one unrolled body can expose enough instruction-level parallelism to exceed the baseline.

02 / Scientific basis

Why the test is meaningful

ILP is useful only if it outweighs duplicated instructions and register pressure under identical arithmetic and nonce assignment.

speedup_Uk=Tbaseline/TUkSASS_Uk≈k·SASS_basepromote on unseen headers only
Forced 2×/4× inter-nonce unrollingVisual reading of the published metrics and gates for EXP-063; it summarizes the registered result, not a mining advantage.EXP-063 / FORCED 2×/4× INTER-NONCE UNROLLINGU2 KERNEL RATIO0.996673REPLICATION DIRECTION0 / 4U2 REGISTERS64
FIGURE / RESULT READINGVisual reading of the published metrics and gates for EXP-063; it summarizes the registered result, not a mining advantage.
03 / Method

How it was tested

Choose U2 on two discovery headers, freeze it, and evaluate four unseen headers in 44 balanced runs of 134,217,728 × 48 hashes.

04 / Observed result

What happened

U2 kernel ratio0.996673
Replication direction0 / 4
U2 registers64

U2 lost on 4/4 replication headers: kernel ratio 0.996673 and 95% interval 0.995713–0.997634. U2/U4 raised SASS from 2,507 to 4,968/9,874 and registers from 44 to 64.

05 / Validation

Exactness and statistical controls

All variants recovered exact B32 hits with no discrepancies or spills. Selection was completed before replication.

06 / Interpretation

What the result means

The compiler materialized duplication, but extra ILP did not repay footprint and register pressure.

Limitations

  • Only forced factors 2 and 4 were tested.
  • The conclusion is architecture-specific.
  • Larger factors were closed by the negative mechanism.
07 / Reproduction

Evidence trail

Freeze U2 after discovery, use unseen headers and audit exact work, SASS and register counts.

Canonical variants

CANONICAL-EXP-139PREREGISTERED-CAMPAIGNSEALED-AUDIT

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