executableCore-1 previewrevision 2

Fused Multiply Add

compute/fused-multiply-add

computecompiledinterpreterc

Public context

What an implementation receives

# Fused Multiply Add Implement the normative fused multiply add behavior for the ZergLang compute domain. Preserve deterministic ordering, typed failures, ownership transfer, and exactly-once cleanup required by the cited clauses. Submit the complete source closure matching `interface.zli`. Public cases are examples, not an exhaustive specification. Do not use ambient I/O, time, random state, or undeclared capabilities.

Public interfaceZergLang

module benchmark_compute_fused_multiply_add;

public value Kernel {
    public compiled compute message run(type: Class<Self>,
        lhs: TensorView<Float64,1,RowMajor>,
        rhs: TensorView<Float64,1,RowMajor>,
        addend: TensorView<Float64,1,RowMajor>)
        -> Tensor<Float64,1,RowMajor> effects { compute };
}

Public cases

1 inspectable case

publicpublic input + expected behavior

Input

{
  "fields": [
    {
      "name": "lhs",
      "value": {
        "items": [
          {
            "bits": "0x3ff0000000000001",
            "format": "binary64",
            "type": "float"
          },
          {
            "bits": "0xc008000000000000",
            "format": "binary64",
            "type": "float"
          },
          {
            "bits": "0x3fe0000000000000",
            "format": "binary64",
            "type": "float"
          }
        ],
        "type": "sequence"
      }
    },
    {
      "name": "rhs",
      "value": {
        "items": [
          {
            "bits": "0x3feffffffffffffe",
            "format": "binary64",
            "type": "float"
          },
          {
            "bits": "0x4000000000000000",
            "format": "binary64",
            "type": "float"
          },
          {
            "bits": "0xc020000000000000",
            "format": "binary64",
            "type": "float"
          }
        ],
        "type": "sequence"
      }
    },
    {
      "name": "addend",
      "value": {
        "items": [
          {
            "bits": "0xbff0000000000000",
            "format": "binary64",
            "type": "float"
          },
          {
            "bits": "0x401c000000000000",
            "format": "binary64",
            "type": "float"
          },
          {
            "bits": "0x4010000000000000",
            "format": "binary64",
            "type": "float"
          }
        ],
        "type": "sequence"
      }
    },
    {
      "name": "layout",
      "value": {
        "type": "text",
        "value": "row-major"
      }
    },
    {
      "name": "shape",
      "value": {
        "items": [
          {
            "type": "integer",
            "value": "3"
          }
        ],
        "type": "sequence"
      }
    },
    {
      "name": "scenario",
      "value": {
        "type": "text",
        "value": "fused-multiply-add"
      }
    }
  ],
  "type": "record"
}

Expected

{
  "kind": "value",
  "value": {
    "fields": [
      {
        "name": "layout",
        "value": {
          "type": "text",
          "value": "row-major"
        }
      },
      {
        "name": "shape",
        "value": {
          "items": [
            {
              "type": "integer",
              "value": "3"
            }
          ],
          "type": "sequence"
        }
      },
      {
        "name": "bits",
        "value": {
          "items": [
            {
              "bits": "0xb970000000000000",
              "format": "binary64",
              "type": "float"
            },
            {
              "bits": "0x3ff0000000000000",
              "format": "binary64",
              "type": "float"
            },
            {
              "bits": "0x0000000000000000",
              "format": "binary64",
              "type": "float"
            }
          ],
          "type": "sequence"
        }
      }
    ],
    "type": "record"
  }
}

Held-out inputs and expectations are excluded from this projection. Their published outcomes are aggregate-only.

Code entry

ZergLang source policy

Active reference source withheld

This task participates in model synthesis. Publishing its canonical solution would invalidate future evaluations; the prompt, interface, and public cases above are the complete public authoring context.

Execution contract

compute

Selector: benchmark_compute_fused_multiply_add.Kernel.run
Executors: compute-scalar
Platforms: portable
Performance lane: eligible

Normative authority

1 bound clauses

C1-COMPUTE-004

Official task history

Signed conformance observations

No official conformance run currently contains this task revision in its signed catalog.

Provenance

Where this task came from

[
  {
    "kind": "original",
    "license": "Apache-2.0",
    "source": "ZergLang Core-0/Core-1 normative specification"
  }
]