HL.MADD

assembly

[
  "hl.madd SrcL, SrcR, SrcD, ->Dst0, Dst1"
]

block

[]

catalog_indices

[
  220
]

catalog_records

[
  {
    "asm": "hl.madd SrcL, SrcR, SrcD, ->Dst0, Dst1",
    "constraints": [],
    "encoding": [
      {
        "index": 0,
        "mask": "0x0600707f07ff",
        "match": "0x00006047000e",
        "width_bits": 48
      }
    ],
    "encoding_kind": "HL48",
    "fields": [
      {
        "name": "RegDst0",
        "pieces": [
          {
            "instruction_lsb": 23,
            "value_lsb": 0,
            "width": 5
          }
        ],
        "signedness": "encoding-defined",
        "width": 5
      },
      {
        "name": "RegDst1",
        "pieces": [
          {
            "instruction_lsb": 11,
            "value_lsb": 0,
            "width": 5
          }
        ],
        "signedness": "encoding-defined",
        "width": 5
      },
      {
        "name": "SrcD",
        "pieces": [
          {
            "instruction_lsb": 43,
            "value_lsb": 0,
            "width": 5
          }
        ],
        "signedness": "encoding-defined",
        "width": 5
      },
      {
        "name": "SrcL",
        "pieces": [
          {
            "instruction_lsb": 31,
            "value_lsb": 0,
            "width": 5
          }
        ],
        "signedness": "encoding-defined",
        "width": 5
      },
      {
        "name": "SrcR",
        "pieces": [
          {
            "instruction_lsb": 36,
            "value_lsb": 0,
            "width": 5
          }
        ],
        "signedness": "encoding-defined",
        "width": 5
      }
    ],
    "form_id": "hl_madd_48_b062d741fd99",
    "length_bits": 48,
    "mnemonic": "HL.MADD",
    "semantic_family": "ALU",
    "semantic_group": "ALU",
    "semantic_handler": "ExecuteScalarMultiplyAddPair",
    "semantic_summary": "HL.MADD computes a signed 128-bit product plus a sign-extended XLEN addend and publishes low then high halves.",
    "status": "accepted"
  }
]

classification

[
  "alu"
]

contract

{
  "block_composition": [
    "none"
  ],
  "canonical_assembly": [
    "hl.madd SrcL, SrcR, SrcD, ->Dst0, Dst1"
  ],
  "defaults": [
    "Every encoded operand and destination field is required; no field can be omitted.",
    "The mnemonic fixes signedness, effective operand width, single-versus-pair result shape, and add-versus-subtract behavior; there is no encoded arithmetic mode."
  ],
  "encoding_class": "standalone-encoded",
  "examples": [
    "hl.madd srcl, srcr, srcd, ->dst0, dst1"
  ],
  "exceptions": [
    "Multiplication and accumulation are fixed-width and raise no arithmetic exception; discarded overflow wraps modulo the defined result width.",
    "An unavailable selected T/U source raises Fault_IllegalInstruction before any destination effect and before TPC advances."
  ],
  "field_contracts": {},
  "field_zero_meanings": {
    "SrcD": "Encoded zero reads the architectural zero GPR.",
    "SrcL": "Encoded zero reads the architectural zero GPR.",
    "SrcR": "Encoded zero reads the architectural zero GPR.",
    "RegDst0": "Encoded zero discards the low result.",
    "RegDst1": "Encoded zero discards the high result."
  },
  "legality": [
    "Every source Reg5 code is assigned: 0..23 select absolute GPRs, 24..27 select T#1..T#4, and 28..31 select U#1..U#4 without consumption.",
    "Each destination independently uses the common map: codes 0 and 24..29 discard, codes 1..23 write GPRs, code 30 pushes U, and code 31 pushes T.",
    "Fixed encoding bits must match the canonical form; every encoded source, destination, and immediate value otherwise has assigned behavior."
  ],
  "memory_effects": [
    "none"
  ],
  "operands": [
    {
      "field": "RegDst0",
      "role": "low product or accumulator Reg5 destination"
    },
    {
      "field": "RegDst1",
      "role": "high product or accumulator Reg5 destination"
    },
    {
      "field": "SrcD",
      "role": "addend Reg5 source"
    },
    {
      "field": "SrcL",
      "role": "left multiplicand or additive Reg5 source"
    },
    {
      "field": "SrcR",
      "role": "right multiplicand Reg5 source"
    }
  ],
  "ordering": [
    "Snapshot every source before any destination effect so duplicate selectors and destination aliases observe pre-instruction values.",
    "Publish the low result to RegDst0, publish the high result to RegDst1, then advance TPC by six bytes."
  ],
  "standalone_opcode": true,
  "state_effects": [
    "Compute the signed 128-bit product of SrcL and SrcR, sign-extend SrcD to 128 bits, and add modulo 2^128.",
    "Snapshot every source and compute the complete 128-bit result before destinations. Publish bits 63:0 to RegDst0, then bits 127:64 to RegDst1. Duplicate destinations are legal; the second high result is final/newest.",
    "No memory, reservation, descriptor, numeric-status, block, privilege, branch-target, or other control state changes. Successful execution advances TPC by six bytes."
  ]
}

depends_on

[
  "PTO-BLOCK-MODEL-SCHEMA-PROFILE-ENCODING"
]

id

PTO-SCALAR-HL-MADD

mnemonic

HL.MADD

summary

HL.MADD computes a signed 128-bit product plus a sign-extended XLEN addend and publishes low then high halves.

surface

scalar

{
  "classification": [
    "alu"
  ],
  "documentation": "docs/scalar/alu/HL.MADD.md",
  "id": "PTO-SCALAR-HL-MADD",
  "instruction_contract": {
    "artifact": "spec/evidence/instruction-contract-closure.json",
    "mnemonic": "HL.MADD",
    "ndf_clause": "PTO-INST-SCALAR-HL-MADD"
  },
  "mnemonic": "HL.MADD",
  "readiness_subjects": [
    "ADR-0026",
    "ADR-0059"
  ],
  "semantic_tests": [
    "PTO-AVS-SCALAR-HL-MADD-ACCUMULATE-001"
  ],
  "source": "asl/scalar/alu/HL.MADD.asl",
  "surface": "scalar",
  "tests": [
    "PTO-AVS-SCALAR-HL-MADD-ACCUMULATE-001",
    "PTO-AVS-SCALAR-HL-MADD-DECODE-001"
  ]
}