{
  "domain": "ai.onnx",
  "name": "Less",
  "sinceVersion": 13,
  "description": "Performs an elementwise `less-than` comparison between tensors `A` and `B` with NumPy-style multidirectional broadcasting, producing a boolean result tensor `C`. Each output element is `true` where the corresponding element of `A` is strictly less than that of `B`.",
  "inputs": [
    { "role": "A", "dtype": "T", "description": "First input operand for the less-than comparison." },
    { "role": "B", "dtype": "T", "description": "Second input operand for the less-than comparison." }
  ],
  "outputs": [
    {
      "role": "C",
      "dtype": "B",
      "rank": "max(ranks.A, ranks.B)",
      "description": "Boolean result tensor; true where A < B.",
      "shape": "broadcastShape(shapes.A, shapes.B)"
    }
  ],
  "typeConstraints": { "T": ["float32", "float16", "int32", "int16", "uint32", "int8", "uint8"], "B": ["bool"] },
  "args": {
    "a": { "kind": "tensor", "semantic": "A", "role": "input" },
    "b": { "kind": "tensor", "semantic": "B", "role": "input" },
    "c": { "kind": "tensor", "semantic": "C", "role": "output" }
  },
  "tunables": { "WORKGROUP_SIZE": 256 },
  "variants": [
    {
      "id": "same_shape_vec4",
      "priority": 20,
      "when": ["sameShape(shapes.A, shapes.C)", "sameShape(shapes.B, shapes.C)", "numel(shapes.C) > 0", "numel(shapes.C) % 4 == 0", "f16Ok(dtypes.T)"],
      "constants": { "vectorScalar": "\"vec4<\" ~ dtypes.T ~ \">\"", "usesF16": "dtypes.T == \"f16\"" },
      "passes": [
        {
          "id": "main",
          "name": "Less.vec4",
          "source": { "shader": "compare-vec4.wgsl.jinja", "inputs": { "op": "\"less\"" } },
          "bindings": [
            {
              "name": "a",
              "arg": "a",
              "semantic": "A",
              "buffer": { "type": "read-only-storage" },
              "elementType": "$vectorScalar"
            },
            {
              "name": "b",
              "arg": "b",
              "semantic": "B",
              "buffer": { "type": "read-only-storage" },
              "elementType": "$vectorScalar"
            },
            { "name": "c", "arg": "c", "semantic": "C", "buffer": { "type": "storage" }, "elementType": "vec4<u32>" },
            {
              "name": "params",
              "semantic": "kernel.params",
              "buffer": { "type": "uniform" },
              "struct": {
                "name": "Params",
                "fields": [{ "name": "count", "type": "u32", "value": "numel(shapes.C) / 4" }]
              }
            }
          ],
          "dispatch": { "threads": "numel(shapes.C) / 4", "workgroupSize": "tunables.WORKGROUP_SIZE" }
        }
      ]
    },
    {
      "id": "broadcast_vec4",
      "when": ["ranks.A <= ranks.C", "ranks.B <= ranks.C", "ranks.C >= 1", "dim(shapes.C, ranks.C - 1) % 4 == 0", "numel(shapes.C) % 4 == 0", "numel(shapes.C) >= 4", "f16Ok(dtypes.T)"],
      "constants": { "scalar": "dtypes.T", "usesF16": "dtypes.T == \"f16\"" },
      "passes": [
        {
          "id": "main",
          "name": "Less",
          "source": {
            "shader": "compare-broadcast-vec4.wgsl.jinja",
            "inputs": {
              "aShape": "shapes.A",
              "bShape": "shapes.B",
              "cShape": "shapes.C",
              "aRank": "ranks.A",
              "bRank": "ranks.B",
              "cRank": "ranks.C",
              "op": "\"less\""
            }
          },
          "bindings": [
            {
              "name": "a",
              "arg": "a",
              "semantic": "A",
              "buffer": { "type": "read-only-storage" },
              "elementType": "$scalar"
            },
            {
              "name": "b",
              "arg": "b",
              "semantic": "B",
              "buffer": { "type": "read-only-storage" },
              "elementType": "$scalar"
            },
            { "name": "c", "arg": "c", "semantic": "C", "buffer": { "type": "storage" }, "elementType": "vec4<u32>" },
            {
              "name": "params",
              "semantic": "kernel.params",
              "buffer": { "type": "uniform" },
              "struct": {
                "name": "Params",
                "fields": [{ "name": "count", "type": "u32", "value": "numel(shapes.C) / 4" }]
              }
            }
          ],
          "dispatch": { "gridStride": "numel(shapes.C) / 4", "workgroupSize": "tunables.WORKGROUP_SIZE" }
        }
      ],
      "priority": 10
    },
    {
      "id": "broadcast",
      "when": ["ranks.A <= ranks.C", "ranks.B <= ranks.C", "f16Ok(dtypes.T)"],
      "constants": { "scalar": "dtypes.T", "usesF16": "dtypes.T == \"f16\"" },
      "passes": [
        {
          "id": "main",
          "name": "Less",
          "source": {
            "shader": "compare-broadcast.wgsl.jinja",
            "inputs": {
              "aShape": "shapes.A",
              "bShape": "shapes.B",
              "cShape": "shapes.C",
              "aRank": "ranks.A",
              "bRank": "ranks.B",
              "cRank": "ranks.C",
              "op": "\"less\"",
              "itemsPerInvocation": 4
            }
          },
          "bindings": [
            {
              "name": "a",
              "arg": "a",
              "semantic": "A",
              "buffer": { "type": "read-only-storage" },
              "elementType": "$scalar"
            },
            {
              "name": "b",
              "arg": "b",
              "semantic": "B",
              "buffer": { "type": "read-only-storage" },
              "elementType": "$scalar"
            },
            { "name": "c", "arg": "c", "semantic": "C", "buffer": { "type": "storage" }, "elementType": "u32" },
            {
              "name": "params",
              "semantic": "kernel.params",
              "buffer": { "type": "uniform" },
              "struct": { "name": "Params", "fields": [{ "name": "count", "type": "u32", "value": "numel(shapes.C)" }] }
            }
          ],
          "dispatch": { "threads": "ceilDiv(numel(shapes.C), 4)", "workgroupSize": "tunables.WORKGROUP_SIZE" }
        }
      ]
    },
    {
      "id": "same_shape_scalar_x4",
      "priority": 15,
      "when": ["sameShape(shapes.A, shapes.C)", "sameShape(shapes.B, shapes.C)", "numel(shapes.C) > 0", "numel(shapes.C) % 4 != 0", "ranks.A <= ranks.C", "ranks.B <= ranks.C", "f16Ok(dtypes.T)"],
      "constants": { "scalar": "dtypes.T", "usesF16": "dtypes.T == \"f16\"" },
      "passes": [
        {
          "id": "main",
          "name": "Less",
          "source": {
            "shader": "compare-broadcast.wgsl.jinja",
            "inputs": {
              "aShape": "shapes.A",
              "bShape": "shapes.B",
              "cShape": "shapes.C",
              "aRank": "ranks.A",
              "bRank": "ranks.B",
              "cRank": "ranks.C",
              "op": "\"less\"",
              "itemsPerInvocation": 4
            }
          },
          "bindings": [
            {
              "name": "a",
              "arg": "a",
              "semantic": "A",
              "buffer": { "type": "read-only-storage" },
              "elementType": "$scalar"
            },
            {
              "name": "b",
              "arg": "b",
              "semantic": "B",
              "buffer": { "type": "read-only-storage" },
              "elementType": "$scalar"
            },
            { "name": "c", "arg": "c", "semantic": "C", "buffer": { "type": "storage" }, "elementType": "u32" },
            {
              "name": "params",
              "semantic": "kernel.params",
              "buffer": { "type": "uniform" },
              "struct": { "name": "Params", "fields": [{ "name": "count", "type": "u32", "value": "numel(shapes.C)" }] }
            }
          ],
          "dispatch": { "threads": "ceilDiv(numel(shapes.C), 4)", "workgroupSize": "tunables.WORKGROUP_SIZE" }
        }
      ]
    }
  ]
}