t81-foundation

TISC Opcode Listing

Status: Canonical / Frozen (v1.1.0) Version: 1.1.0 Source of Truth: VM + opcode headers

1. Encoding Overview

2. Opcode Categories

2.1 Arithmetic

| Mnemonic | Numeric Encoding | Operands | Description | Deterministic | Implementation Location | | :— | :— | :— | :— | :— | :— | | Add | 5 (0x05) | A, B, C | Integer Addition | Yes | core/vm/vm.cpp | | Sub | 6 (0x06) | A, B, C | Integer Subtraction | Yes | core/vm/vm.cpp | | Mul | 7 (0x07) | A, B, C | Integer Multiplication | Yes | core/vm/vm.cpp | | Div | 8 (0x08) | A, B, C | Integer Division | Yes | core/vm/vm.cpp | | Mod | 9 (0x09) | A, B, C | Integer Modulo | Yes | core/vm/vm.cpp | | Inc | 13 (0x0D) | A, B, C | Integer Increment | Yes | core/vm/vm.cpp | | Dec | 14 (0x0E) | A, B, C | Integer Decrement | Yes | core/vm/vm.cpp | | Neg | 59 (0x3B) | A, B, C | Integer Negation | Yes | core/vm/vm.cpp | | FAdd | 36 (0x24) | A, B, C | Float Addition | Yes | core/vm/vm.cpp | | FSub | 37 (0x25) | A, B, C | Float Subtraction | Yes | core/vm/vm.cpp | | FMul | 38 (0x26) | A, B, C | Float Multiplication | Yes | core/vm/vm.cpp | | FDiv | 39 (0x27) | A, B, C | Float Division | Yes | core/vm/vm.cpp | | FracAdd | 40 (0x28) | A, B, C | Fraction Addition | Yes | core/vm/vm.cpp | | FracSub | 41 (0x29) | A, B, C | Fraction Subtraction | Yes | core/vm/vm.cpp | | FracMul | 42 (0x2A) | A, B, C | Fraction Multiplication | Yes | core/vm/vm.cpp | | FracDiv | 43 (0x2B) | A, B, C | Fraction Division | Yes | core/vm/vm.cpp | | TVecAdd | 33 (0x21) | A, B, C | Tensor Vector Addition | Yes | core/vm/vm.cpp | | TVecMul | 79 (0x4F) | A, B, C | Tensor Vector Multiplication | Yes | core/vm/vm.cpp | | TMatMul | 34 (0x22) | A, B, C | Tensor Matrix Multiplication | Yes | core/vm/vm.cpp | | TTenDot | 35 (0x23) | A, B, C | Tensor Dot Product | Yes | core/vm/vm.cpp | | TNorm | 127 (0x7F) | A, B, C | Ternary Normalization | Yes | core/vm/vm.cpp | | BitAnd | 167 (0xA7) | A, B, C | Bitwise AND | Yes | core/vm/vm.cpp | | BitOr | 168 (0xA8) | A, B, C | Bitwise OR | Yes | core/vm/vm.cpp | | BitXor | 169 (0xA9) | A, B, C | Bitwise XOR | Yes | core/vm/vm.cpp | | BitNot | 170 (0xAA) | A, B, C | Bitwise NOT | Yes | core/vm/vm.cpp | | BitShl | 171 (0xAB) | A, B, C | Bitwise Shift Left | Yes | core/vm/vm.cpp | | BitShr | 172 (0xAC) | A, B, C | Bitwise Shift Right (Arithmetic) | Yes | core/vm/vm.cpp | | BitUShr | 173 (0xAD) | A, B, C | Bitwise Shift Right (Logical) | Yes | core/vm/vm.cpp |

2.2 Control Flow

| Mnemonic | Numeric Encoding | Operands | Description | Deterministic | Implementation Location | | :— | :— | :— | :— | :— | :— | | Halt | 1 (0x01) | A, B, C | Halt Execution | Yes | core/vm/vm.cpp | | Jump | 10 (0x0A) | A, B, C | Unconditional Jump | Yes | core/vm/vm.cpp | | JumpIfZero | 11 (0x0B) | A, B, C | Jump if Zero | Yes | core/vm/vm.cpp | | JumpIfNotZero | 25 (0x19) | A, B, C | Jump if Not Zero | Yes | core/vm/vm.cpp | | Call | 26 (0x1A) | A, B, C | Function Call | Yes | core/vm/vm.cpp | | Ret | 27 (0x1B) | A, B, C | Return from Call | Yes | core/vm/vm.cpp | | JumpIfNegative | 60 (0x3C) | A, B, C | Jump if Negative | Yes | core/vm/vm.cpp | | JumpIfPositive | 61 (0x3D) | A, B, C | Jump if Positive | Yes | core/vm/vm.cpp | | AxHalt | 131 (0x83) | A, B, C | Axion Halt (Policy Deny) | Yes | core/vm/vm.cpp |

2.3 Comparison

| Mnemonic | Numeric Encoding | Operands | Description | Deterministic | Implementation Location | | :— | :— | :— | :— | :— | :— | | Cmp | 15 (0x0F) | A, B, C | Compare | Yes | core/vm/vm.cpp | | Less | 62 (0x3E) | A, B, C | Less Than | Yes | core/vm/vm.cpp | | LessEqual | 63 (0x3F) | A, B, C | Less Than or Equal | Yes | core/vm/vm.cpp | | Greater | 64 (0x40) | A, B, C | Greater Than | Yes | core/vm/vm.cpp | | GreaterEqual | 65 (0x41) | A, B, C | Greater Than or Equal | Yes | core/vm/vm.cpp | | Equal | 66 (0x42) | A, B, C | Equal | Yes | core/vm/vm.cpp | | NotEqual | 67 (0x43) | A, B, C | Not Equal | Yes | core/vm/vm.cpp |

2.4 Memory / Stack / Heap

| Mnemonic | Numeric Encoding | Operands | Description | Deterministic | Implementation Location | | :— | :— | :— | :— | :— | :— | | LoadImm | 2 (0x02) | A, B, C | Load Immediate | Yes | core/vm/vm.cpp | | Load | 3 (0x03) | A, B, C | Load from Memory | Yes | core/vm/vm.cpp | | Store | 4 (0x04) | A, B, C | Store to Memory | Yes | core/vm/vm.cpp | | Mov | 12 (0x0C) | A, B, C | Move Register | Yes | core/vm/vm.cpp | | Push | 16 (0x10) | A, B, C | Push to Stack | Yes | core/vm/vm.cpp | | Pop | 17 (0x11) | A, B, C | Pop from Stack | Yes | core/vm/vm.cpp | | SetF | 44 (0x2C) | A, B, C | Set Flag Register | Yes | core/vm/vm.cpp | | StackAlloc | 68 (0x44) | A, B, C | Allocate Stack Frame | Yes | core/vm/vm.cpp | | StackFree | 69 (0x45) | A, B, C | Free Stack Frame | Yes | core/vm/vm.cpp | | HeapAlloc | 70 (0x46) | A, B, C | Allocate Heap Block | Yes | core/vm/vm.cpp | | HeapFree | 71 (0x47) | A, B, C | Free Heap Block | Yes | core/vm/vm.cpp | | Canon | 128 (0x80) | A, B, C | Canonicalize Memory | Yes | core/vm/vm.cpp | | MemZero | 129 (0x81) | A, B, C | Zero Memory Region | Yes | core/vm/vm.cpp | | Copy | 130 (0x82) | A, B, C | Copy Memory Region | Yes | core/vm/vm.cpp |

2.5 Type Conversion

| Mnemonic | Numeric Encoding | Operands | Description | Deterministic | Implementation Location | | :— | :— | :— | :— | :— | :— | | I2F | 29 (0x1D) | A, B, C | Int to Float | Yes | core/vm/vm.cpp | | F2I | 30 (0x1E) | A, B, C | Float to Int | Yes | core/vm/vm.cpp | | I2Frac | 31 (0x1F) | A, B, C | Int to Fraction | Yes | core/vm/vm.cpp | | Frac2I | 32 (0x20) | A, B, C | Fraction to Int | Yes | core/vm/vm.cpp | | F2Frac | 163 (0xA3) | A, B, C | Float to Fraction | Yes | core/vm/vm.cpp | | Frac2F | 164 (0xA4) | A, B, C | Fraction to Float | Yes | core/vm/vm.cpp |

2.6 Logic

| Mnemonic | Numeric Encoding | Operands | Description | Deterministic | Implementation Location | | :— | :— | :— | :— | :— | :— | | TNot | 18 (0x12) | A, B, C | Ternary Not | Yes | core/vm/vm.cpp | | TAnd | 19 (0x13) | A, B, C | Ternary And | Yes | core/vm/vm.cpp | | TOr | 20 (0x14) | A, B, C | Ternary Or | Yes | core/vm/vm.cpp | | TXor | 21 (0x15) | A, B, C | Ternary Xor | Yes | core/vm/vm.cpp |

2.7 Axion / Policy

| Mnemonic | Numeric Encoding | Operands | Description | Deterministic | Implementation Location | | :— | :— | :— | :— | :— | :— | | AxRead | 22 (0x16) | A, B, C | Read Axion Guard | Yes | core/vm/vm.cpp | | AxSet | 23 (0x17) | A, B, C | Set Axion Guard | Yes | core/vm/vm.cpp | | AxVerify | 24 (0x18) | A, B, C | Verify Axion State | Yes | core/vm/vm.cpp | | Trap | 28 (0x1C) | A, B, C | Raise Trap | Yes | core/vm/vm.cpp | | Assert | 132 (0x84) | A, B, C | Runtime Assertion | Yes | core/vm/vm.cpp | | AxCheck | 158 (0x9E) | A, B, C | Axion Check (deny fails closed) | Yes | core/vm/vm.cpp | | AxSign | 159 (0x9F) | A, B, C | Axion Sign (Unimplemented; fail-closed) | Yes | core/vm/vm.cpp | | AxLineage | 160 (0xA0) | A, B, C | Axion Lineage (Unimplemented; fail-closed) | Yes | core/vm/vm.cpp | | AxCanon | 161 (0xA1) | A, B, C | Axion Canon (Unimplemented; fail-closed) | Yes | core/vm/vm.cpp | | AxReport | 162 (0xA2) | A, B, C | Axion Report (policy-checked) | Yes | core/vm/vm.cpp |

2.8 Tensor / Neural

| Mnemonic | Numeric Encoding | Operands | Description | Deterministic | Implementation Location | | :— | :— | :— | :— | :— | :— | | ChkShape | 45 (0x2D) | A, B, C | Check Tensor Shape | Yes | core/vm/vm.cpp | | WeightsLoad | 72 (0x48) | A, B, C | Load Weights Tensor | Yes | core/vm/vm.cpp | | TExp | 73 (0x49) | A, B, C | Tensor Exp | Yes | core/vm/vm.cpp | | TSqrt | 74 (0x4A) | A, B, C | Tensor Sqrt | Yes | core/vm/vm.cpp | | TSiLU | 75 (0x4B) | A, B, C | Tensor SiLU | Yes | core/vm/vm.cpp | | TSoftmax | 76 (0x4C) | A, B, C | Tensor Softmax | Yes | core/vm/vm.cpp | | TRMSNorm | 77 (0x4D) | A, B, C | Tensor RMSNorm | Yes | core/vm/vm.cpp | | TRoPE | 78 (0x4E) | A, B, C | Tensor RoPE | Yes | core/vm/vm.cpp | | TTranspose | 80 (0x50) | A, B, C | Tensor Transpose | Yes | core/vm/vm.cpp | | TGet | 107 (0x6B) | A, B, C | Tensor Get Element | Yes | core/vm/vm.cpp | | TNew | 108 (0x6C) | A, B, C | Tensor New | Yes | core/vm/vm.cpp | | TSet | 109 (0x6D) | A, B, C | Tensor Set Element | Yes | core/vm/vm.cpp | | TLoadHash | 121 (0x79) | A, B, C | Load Tensor by Hash | Yes | core/vm/vm.cpp | | TID | 122 (0x7A) | A, B, C | Tensor Identity | Yes | core/vm/vm.cpp | | TNeuralFwd | 165 (0xA5) | A, B, C | Neural Forward (Unimplemented; fail-closed) | Yes | core/vm/vm.cpp | | TNeuralBwd | 166 (0xA6) | A, B, C | Neural Backward (Unimplemented; fail-closed) | Yes | core/vm/vm.cpp |

2.9 Math (Float)

| Mnemonic | Numeric Encoding | Operands | Description | Deterministic | Implementation Location | | :— | :— | :— | :— | :— | :— | | FSin | 81 (0x51) | A, B, C | Float Sin | Yes | core/vm/vm.cpp | | FCos | 82 (0x52) | A, B, C | Float Cos | Yes | core/vm/vm.cpp | | FTan | 83 (0x53) | A, B, C | Float Tan | Yes | core/vm/vm.cpp | | FAsin | 84 (0x54) | A, B, C | Float Asin | Yes | core/vm/vm.cpp | | FAcos | 85 (0x55) | A, B, C | Float Acos | Yes | core/vm/vm.cpp | | FAtan | 86 (0x56) | A, B, C | Float Atan | Yes | core/vm/vm.cpp | | FSinh | 87 (0x57) | A, B, C | Float Sinh | Yes | core/vm/vm.cpp | | FCosh | 88 (0x58) | A, B, C | Float Cosh | Yes | core/vm/vm.cpp | | FTanh | 89 (0x59) | A, B, C | Float Tanh | Yes | core/vm/vm.cpp | | FSqrt | 90 (0x5A) | A, B, C | Float Sqrt | Yes | core/vm/vm.cpp | | FExp | 91 (0x5B) | A, B, C | Float Exp | Yes | core/vm/vm.cpp | | FLog | 92 (0x5C) | A, B, C | Float Log | Yes | core/vm/vm.cpp | | FPow | 93 (0x5D) | A, B, C | Float Pow | Yes | core/vm/vm.cpp |

2.10 Meta / System

| Mnemonic | Numeric Encoding | Operands | Description | Deterministic | Implementation Location | | :— | :— | :— | :— | :— | :— | | MetaRead | 94 (0x5E) | A, B, C | Read Meta Memory | Yes | core/vm/vm.cpp | | MetaWrite | 95 (0x5F) | A, B, C | Write Meta Memory | Yes | core/vm/vm.cpp | | MetaReflect | 96 (0x60) | A, B, C | Reflection Capture | Yes | core/vm/vm.cpp | | MetaRefine | 97 (0x61) | A, B, C | Meta Refinement | Yes | core/vm/vm.cpp | | Print | 98 (0x62) | A, B, C | System Print | Yes | core/vm/vm.cpp |

2.11 String / Vector

| Mnemonic | Numeric Encoding | Operands | Description | Deterministic | Implementation Location | | :— | :— | :— | :— | :— | :— | | StrLen | 99 (0x63) | A, B, C | String Length | Yes | core/vm/vm.cpp | | StrEmpty | 100 (0x64) | A, B, C | String Is Empty | Yes | core/vm/vm.cpp | | VecLen | 101 (0x65) | A, B, C | Vector Length | Yes | core/vm/vm.cpp | | VecEmpty | 102 (0x66) | A, B, C | Vector Is Empty | Yes | core/vm/vm.cpp | | VecFirst | 103 (0x67) | A, B, C | Vector First Element | Yes | core/vm/vm.cpp | | VecLast | 104 (0x68) | A, B, C | Vector Last Element | Yes | core/vm/vm.cpp | | VecPush | 105 (0x69) | A, B, C | Vector Push | Yes | core/vm/vm.cpp | | VecPop | 106 (0x6A) | A, B, C | Vector Pop | Yes | core/vm/vm.cpp | | StrConcat | 110 (0x6E) | A, B, C | String Concat | Yes | core/vm/vm.cpp | | StrStartsWith | 111 (0x6F) | A, B, C | String Starts With | Yes | core/vm/vm.cpp | | StrEndsWith | 112 (0x70) | A, B, C | String Ends With | Yes | core/vm/vm.cpp | | StrContains | 113 (0x71) | A, B, C | String Contains | Yes | core/vm/vm.cpp | | StrIndexOf | 114 (0x72) | A, B, C | String Index Of | Yes | core/vm/vm.cpp | | StrReplace | 115 (0x73) | A, B, C | String Replace | Yes | core/vm/vm.cpp | | StrVecNew | 116 (0x74) | A, B, C | New String Vector | Yes | core/vm/vm.cpp | | StrVecPush | 117 (0x75) | A, B, C | String Vector Push | Yes | core/vm/vm.cpp | | StrSplit | 118 (0x76) | A, B, C | String Split | Yes | core/vm/vm.cpp | | StrJoin | 119 (0x77) | A, B, C | String Join | Yes | core/vm/vm.cpp |

2.12 Options / Results / Enums

| Mnemonic | Numeric Encoding | Operands | Description | Deterministic | Implementation Location | | :— | :— | :— | :— | :— | :— | | MakeOptionSome | 46 (0x2E) | A, B, C | Create Option Some | Yes | core/vm/vm.cpp | | MakeOptionNone | 47 (0x2F) | A, B, C | Create Option None | Yes | core/vm/vm.cpp | | MakeResultOk | 48 (0x30) | A, B, C | Create Result Ok | Yes | core/vm/vm.cpp | | MakeResultErr | 49 (0x31) | A, B, C | Create Result Err | Yes | core/vm/vm.cpp | | OptionIsSome | 50 (0x32) | A, B, C | Check Option Some | Yes | core/vm/vm.cpp | | OptionUnwrap | 51 (0x33) | A, B, C | Unwrap Option | Yes | core/vm/vm.cpp | | ResultIsOk | 52 (0x34) | A, B, C | Check Result Ok | Yes | core/vm/vm.cpp | | ResultUnwrapOk | 53 (0x35) | A, B, C | Unwrap Result Ok | Yes | core/vm/vm.cpp | | ResultUnwrapErr | 54 (0x36) | A, B, C | Unwrap Result Err | Yes | core/vm/vm.cpp | | MakeEnumVariant | 55 (0x37) | A, B, C | Create Enum Variant | Yes | core/vm/vm.cpp | | MakeEnumVariantPayload | 56 (0x38) | A, B, C | Create Enum Payload | Yes | core/vm/vm.cpp | | EnumIsVariant | 57 (0x39) | A, B, C | Check Enum Variant | Yes | core/vm/vm.cpp | | EnumUnwrapPayload | 58 (0x3A) | A, B, C | Unwrap Enum Payload | Yes | core/vm/vm.cpp |

2.13 Complex

| Mnemonic | Numeric Encoding | Operands | Description | Deterministic | Implementation Location | | :— | :— | :— | :— | :— | :— | | MakeComplex | 120 (0x78) | A, B, C | Create Complex Num | Yes | core/vm/vm.cpp |

2.14 Cognitive Tier / Reflection

| Mnemonic | Numeric Encoding | Operands | Description | Deterministic | Implementation Location | | :— | :— | :— | :— | :— | :— | | SymLoad | 133 (0x85) | A, B, C | Symbol Load | Yes | core/vm/vm.cpp | | SymRewrite | 134 (0x86) | A, B, C | Symbol Rewrite | Yes | core/vm/vm.cpp | | SymConfluence | 135 (0x87) | A, B, C | Symbol Confluence | Yes | core/vm/vm.cpp | | SymCanon | 136 (0x88) | A, B, C | Symbol Canonicalize | Yes | core/vm/vm.cpp | | SymBind | 137 (0x89) | A, B, C | Symbol Bind | Yes | core/vm/vm.cpp | | ReflCap | 138 (0x8A) | A, B, C | Reflection Capture | Yes | core/vm/vm.cpp | | ReflJustify | 139 (0x8B) | A, B, C | Reflection Justify | Yes | core/vm/vm.cpp | | ReflCheck | 140 (0x8C) | A, B, C | Reflection Check | Yes | core/vm/vm.cpp | | ReflTrace | 141 (0x8D) | A, B, C | Reflection Trace | Yes | core/vm/vm.cpp | | ReflSeal | 142 (0x8E) | A, B, C | Reflection Seal | Yes | core/vm/vm.cpp | | Recurse | 143 (0x8F) | A, B, C | Recursive Descent | Yes | core/vm/vm.cpp | | Contract | 144 (0x90) | A, B, C | Recursive Contract | Yes | core/vm/vm.cpp | | Entropy | 145 (0x91) | A, B, C | Get Entropy | Yes | core/vm/vm.cpp | | Depth | 146 (0x92) | A, B, C | Get Depth | Yes | core/vm/vm.cpp | | Terminate | 147 (0x93) | A, B, C | Terminate Recursion | Yes | core/vm/vm.cpp | | Merge | 148 (0x94) | A, B, C | Merge State | Yes | core/vm/vm.cpp | | Gossip | 149 (0x95) | A, B, C | Gossip State | Yes | core/vm/vm.cpp | | TickSync | 150 (0x96) | A, B, C | Tick Synchronization | Yes | core/vm/vm.cpp | | Coherence | 151 (0x97) | A, B, C | Coherence Check | Yes | core/vm/vm.cpp | | DistSeal | 152 (0x98) | A, B, C | Distributed Seal | Yes | core/vm/vm.cpp | | InfSeed | 153 (0x99) | A, B, C | Infinite Seed | Yes | core/vm/vm.cpp | | InfExpand | 154 (0x9A) | A, B, C | Infinite Expand | Yes | core/vm/vm.cpp | | InfCollapse | 155 (0x9B) | A, B, C | Infinite Collapse | Yes | core/vm/vm.cpp | | InfConverge | 156 (0x9C) | A, B, C | Infinite Converge | Yes | core/vm/vm.cpp | | InfSignature | 157 (0x9D) | A, B, C | Infinite Signature | Yes | core/vm/vm.cpp |

2.15 Network / Async

| Mnemonic | Numeric Encoding | Operands | Description | Deterministic | Implementation Location | | :— | :— | :— | :— | :— | :— | | NSend | 123 (0x7B) | A, B, C | Network Send (Unimplemented; fail-closed) | Yes | core/vm/vm.cpp | | NRecv | 124 (0x7C) | A, B, C | Network Recv (Unimplemented; fail-closed) | Yes | core/vm/vm.cpp | | VWait | 125 (0x7D) | A, B, C | Async Wait (Unimplemented; fail-closed) | Yes | core/vm/vm.cpp | | VYield | 126 (0x7E) | A, B, C | Async Yield (Unimplemented; fail-closed) | Yes | core/vm/vm.cpp |

2.16 Map / Set Scaffolding

Mnemonic Numeric Encoding Operands Description Deterministic Implementation Location
MapNew 174 (0xAE) A, B, C Create new Map Yes core/vm/vm.cpp
MapPut 175 (0xAF) A, B, C Insert key-value into Map Yes core/vm/vm.cpp
MapGet 176 (0xB0) A, B, C Get value from Map by key Yes core/vm/vm.cpp
MapHas 177 (0xB1) A, B, C Check key presence in Map Yes core/vm/vm.cpp
MapRemove 178 (0xB2) A, B, C Remove key from Map Yes core/vm/vm.cpp
MapKeys 179 (0xB3) A, B, C Get all keys from Map Yes core/vm/vm.cpp
MapSize 180 (0xB4) A, B, C Get Map size Yes core/vm/vm.cpp
SetNew 181 (0xB5) A, B, C Create new Set Yes core/vm/vm.cpp
SetAdd 182 (0xB6) A, B, C Add element to Set Yes core/vm/vm.cpp
SetRemove 183 (0xB7) A, B, C Remove element from Set Yes core/vm/vm.cpp
SetHas 184 (0xB8) A, B, C Check element in Set Yes core/vm/vm.cpp
SetSize 185 (0xB9) A, B, C Get Set size Yes core/vm/vm.cpp

2.17 Tensor Shape

Mnemonic Numeric Encoding Operands Description Deterministic Implementation Location
TShape 186 (0xBA) A, B, C Get tensor dimension size: R[A] = shape[R[B]][C] Yes core/vm/vm.cpp

2.18 AI-Native Inference (RFC-0026)

All opcodes in this class operate on TensorHandle registers. All are Tier 2+ only. All are subject to Axion pre-instruction verification.

Mnemonic Numeric Encoding Operands Description Deterministic Implementation Location
ATTN 187 (0xBB) A, PACK(B,C) Scaled dot-product attention: R[A] = softmax(Q·Kᵀ/√dₖ)·V Yes core/vm/vm.cpp
QMATMUL 188 (0xBC) A, PACK(B,C) Quantized matmul: R[A] = dequantize(R_WT, R_SCALE) · R_ACT Yes core/vm/vm.cpp
EMBED 189 (0xBD) A, B, C Embedding lookup: gather rows from table R[B] at indices R[C] Yes core/vm/vm.cpp
WLOAD 190 (0xBE) A, B, C Weight load with Axion policy gate; emits CanonFS WeightLoad audit event (AI-M4) Yes core/vm/vm.cpp
GATHER 191 (0xBF) A, B, PACK(IDX,AXIS) Sparse gather: gather index R[IDX] from tensor R[B] along axis R[AXIS] (AI-M5) Yes core/vm/vm.cpp
SCATTER 192 (0xC0) A, B, PACK(IDX,SRC) Sparse scatter-add: scatter R[SRC] into R[B] at R[IDX]; aliasing detection enforced (AI-M5) Yes core/vm/vm.cpp
Int2BigInt 193 (0xC1) A, B, C Convert integer register to BigInt handle Yes core/vm/vm.cpp

2.19 Ternary-Native Inference (RFC-0034)

All opcodes in this class require Tier 2+. Weights must be in {−1, 0, +1} (T81Qutrit domain). TACT carries a post-execute Axion activation-ceiling gate (§5.17.6).

Mnemonic Numeric Encoding Operands Description Deterministic Implementation Location
TWMATMUL 194 (0xC2) A, B, C Ternary-weight matmul; T81BigInt accumulator; no FP multiply Yes core/vm/vm.cpp
TQUANT 195 (0xC3) A, B, C Quantize tensor to ternary {−1,0,+1} using threshold R[C] Yes core/vm/vm.cpp
TATTN 196 (0xC4) A, B, PACK(C) Ternary Q/K attention; ternary matmul + softmax + float V-proj Yes core/vm/vm.cpp
TWEMBED 197 (0xC5) A, B, C Row-gather from T81Qutrit embedding table at index R[C] Yes core/vm/vm.cpp
TERNACCUM 198 (0xC6) A, B, C Scalar ternary dot product; result stored as T81BigInt handle Yes core/vm/vm.cpp
TACT 199 (0xC7) A, B, C Ternary activation (mode R[C]); activation-ceiling Axion gate Yes core/vm/vm.cpp

TACT Modes

Mode Byte Name Semantics
0x01 TernaryStep x > 0.5 → +1; x < −0.5 → −1; else 0
0x02 TanhQuantized tanh(x) > 0.5 → +1; tanh(x) < −0.5 → −1; else 0

Post-execute Axion activation-ceiling verdict model:

Verdict Effect
Allow RD committed; PC advances normally
Quarantine RD not committed; PC does not advance; SecurityFault raised
Deny RD not committed; ActivationFault raised

2.20 Governed Foreign Function Interface (RFC-0036 + RFC-00B8)

Opcodes for governed calls to external (non-T81) functions. All are subject to FFIDispatcher policy enforcement, resource quotas, and Axion audit events.

Mnemonic Numeric Encoding Operands Description Deterministic Implementation Location
FFICall 200 (0xC8) A, B, C Call foreign function; text_literal carries function name; policy check before dispatch Conditional (policy-dependent) core/vm/vm.cpp
FFIRegister 201 (0xC9) A, B Register foreign library by name R[A] and version hash R[B] Yes core/vm/vm.cpp
FFIPolicySet 202 (0xCA) A, B Set per-call FFI policy type R[A] to value R[B] Yes core/vm/vm.cpp

2.21 Ternary Lattice Cryptography (RFC-0038)

Negacyclic polynomial arithmetic over {−1, 0, +1} coefficients in Z[x]/(x^n + 1). No integer multiplications — only add/sub/trit-flip. T81BigInt-exact. Tier 2+.

Mnemonic Numeric Encoding Operands Description Deterministic Implementation Location
POLYMUL 203 (0xCB) A, B, C Negacyclic poly multiply: R[A] = polymul(*tensor(R[B]), *tensor(R[C])) in Z[x]/(x^n+1) Yes core/vm/vm.cpp
POLYMOD 204 (0xCC) A, B, C Centered reduction mod q: every coefficient c → ((c%q)+q)%q, shifted to (−q/2, q/2]; q = R[C] Yes core/vm/vm.cpp

2.22 NTRU-KEM Polynomial Ring Arithmetic (RFC-0039)

Elementwise polynomial subtraction over tensor handles; completes the {+, −, ×, mod} ring surface.

Mnemonic Numeric Encoding Operands Description Deterministic Implementation Location
TVecSub 212 (0xD4) A, B, C Elementwise subtraction: R[A][k] = tensor(R[B])[k] − tensor(R[C])[k] for all k Yes core/vm/vm.cpp

2.23 SWAR Tensor Operations (RFC-0040)

Explicit SWAR dispatch over ExactTrit tensor handles. These opcodes are stable entry points for RFC-0040 semantics and bypass scalar ternary-trit execution.

Mnemonic Numeric Encoding Operands Description Deterministic Implementation Location
TNOT_SWAR 213 (0xD5) A, B, C Exact-trit tensor unary negation via stable SWAR kernels: R[A] = swar_not(tensor(R[B])) Yes core/vm/vm.cpp
TAND_SWAR 214 (0xD6) A, B, C Exact-trit tensor elementwise ternary min via SWAR: R[A] = swar_and(tensor(R[B]), tensor(R[C])) Yes core/vm/vm.cpp
TOR_SWAR 215 (0xD7) A, B, C Exact-trit tensor elementwise ternary max via SWAR: R[A] = swar_or(tensor(R[B]), tensor(R[C])) Yes core/vm/vm.cpp

3. Reserved / Unused Opcodes

4. Implementation Consistency Audit

This site is open source. Improve this page.