Experiment: VFPU Scalar and Wide Register Addressing
Experiment ID: EXP-VFPU-ADDRESS-20260805-A · State: ACCEPTANCE ELIGIBLE · Classification: prior-art remeasurement / hardware confirmation
Question
Do the documented VFPU E-field register mappings survive direct measurement on real PSP hardware for scalar encodings and selected wide discriminators?
Fixture and environment
Source-owned register-address probe; PSP-3001, firmware 6.61 with ARK, PSPLink-qualified session. The exact binary/capture is retained privately; this page publishes the bounded case set and measured outcomes.
Procedure and observation
Measure all 128 scalar E-field encodings, then run 14 selected wide cases chosen to distinguish transpose and wrap behavior. Public discriminators include triple E=0x20 → [0,4,8], triple E=0x40 → [1,2,3], and quad E=0x60 → [8,12,0,4]. All measured cases matched the current physical-index rule.
Interpretation
This is PRIOR ART + HARDWARE RECONFIRMATION. Hardware is the behavioral authority for the measured domain; the implementation’s historical emulator lineage remains a separate provenance fact.
Claims not supported
- Only 14 of 512 wide-vector cases were measured; the unmeasured remainder stays NOT TESTED.
- No model-wide or firmware-wide claim; no matrix-register claim.
Implementation comparison
Compare the public decoder/code generator at tools/codegen.py and the public VFPU register-addressing synthesis.
Replication
Run the same scalar digest and selected wide discriminators on PSP-1000 and PSP-2000, preserving model generation, firmware/CFW, fixture identity, and repeat count. Do not submit device identifiers or private captures.
Related: FIND-VFPU-ADDR-001 · CONFLICT-REAL-HARDWARE-AUTHORITY-001 · EXP-VFPU-WIDE-COVERAGE-001 (planned).
Publication safety: public-safe aggregate; exact private fixture withheld. Last reviewed: 2026-08-11.