mirror of
https://github.com/allaunthefox/Research-Stack.git
synced 2026-07-31 03:05:21 +00:00
5290413080
AVMIsa fixes (all pre-existing errors from the upstream merge):
- Types.lean: add Repr to AvmTy
- Value.lean: replace `deriving Inhabited` with explicit instance (AnyVal
is a dependent structure; auto-derive can't pick a default ty+val pair);
add Repr instance that delegates to AvmVal.repr
- Instr.lean: add Repr to Prim and Instr
- State.lean: fix `List.set ⟨i, h⟩` → `List.set i` (List.set takes Nat,
not Fin); drop now-dead `h` binding; add Repr to State
- Step.lean: rewrite evalPrim branches to pattern-match directly on AnyVal
`⟨ty, val⟩` pairs instead of `if v.ty = T` + separate val match (Lean
can't unify `AvmVal v.ty` with `AvmVal T` from a propositional if-guard);
replace `List.get? pc` (removed in Lean 4.30) with `list[pc]?` subscript;
rename Q0_16.addSat/subSat → Q0_16.add/sub (no sat variants exist);
add Repr to StepError and Outcome
PIST.Trace fixes:
- Drop invalid `set_option pp.pretty true`
- MVarId.toNat → MVarId.name.toString
- List.size → List.length (then .toArray for Json.arr)
- Json.num takes JsonNumber {mantissa : Int, exponent : Int}; cast Nat → Int
- goals.mapM goalToJson: lift MetaM → TacticM via liftMetaM
Canary result: `#eval run 8 canaryNot canaryState` →
Outcome.ok { pc := 2, stack := [AvmVal.b true], halted := true }
Generated with [Devin](https://cli.devin.ai/docs)
Co-Authored-By: Devin <158243242+devin-ai-integration[bot]@users.noreply.github.com>
|
||
|---|---|---|
| .. | ||
| FAMM | ||
| GhostPivot | ||
| GhostPivot2 | ||
| manifold | ||
| PIST | ||
| rust | ||
| search | ||
| simulations | ||
| tardygrada | ||
| README.md | ||
| SearchSpace.lean | ||
2-Search-Space
Purpose: Quaternion, braid, and manifold primitives for high-dimensional search space navigation.
Depends on: 0-Core-Formalism
Components
| Component | Status | Description |
|---|---|---|
| PIST | In core/, scratch/ |
Perfectly Imperfect Square Theory |
| SVQF | In core/ |
Spherical Quaternion Vector Field |
| FAMM | In core/, scratch/ |
Frustration-Aware Manifold Mesh |
| Braid Theory | Planned | Topological constraints via braid brackets |
Key Optimizations
- Quaternion Sieve — Counter-rotation as band-pass filter
- Non-linear Gearbox — φ-based fractional steps
- Calabi-Yau Compactification — Discarded space folded into compactified dimensions
- Ricci Flow — Self-smoothing via ∂g_ij/∂t = -2R_ij
Hardware Range
Framework uses fixed-point arithmetic (Q16_16) — runnable on NES (1983) to modern FPGA.