{
  "op": "com.microsoft.MatMulBnb4",
  "fixtureArrays": {
    "bnb4_weight_cycle_b_t": [1, 35, 69, 103, 137, 171, 205, 239, 254, 220, 186, 152, 118, 84, 50, 16, 90]
  },
  "cases": [
    {
      "name": "nf4_scalar_small",
      "attrs": { "K": 16, "N": 4, "block_size": 16, "quant_type": 1 },
      "inputs": {
        "aT": {
          "dtype": "float32",
          "shape": [3, 16],
          "data": { "kind": "fillFloat32", "sinStep": 0.17, "cosStep": 0.31 }
        },
        "bT": {
          "dtype": "uint8",
          "shape": [32],
          "data": { "kind": "cycle", "values": { "$ref": "#/fixtureArrays/bnb4_weight_cycle_b_t" } }
        },
        "absmaxT": {
          "dtype": "float32",
          "shape": [4],
          "data": { "kind": "fillFloat32", "sinStep": 0.11, "cosStep": 0.07, "scale": 0.02, "offset": 0.06 }
        }
      },
      "outputs": { "yT": { "dtype": "float32", "shape": [3, 4], "tolerance": 0.00001 } }
    },
    {
      "name": "fp4_scalar_small",
      "attrs": { "K": 16, "N": 4, "block_size": 16, "quant_type": 0 },
      "inputs": {
        "aT": {
          "dtype": "float32",
          "shape": [3, 16],
          "data": { "kind": "fillFloat32", "sinStep": 0.17, "cosStep": 0.31 }
        },
        "bT": {
          "dtype": "uint8",
          "shape": [32],
          "data": { "kind": "cycle", "values": { "$ref": "#/fixtureArrays/bnb4_weight_cycle_b_t" } }
        },
        "absmaxT": {
          "dtype": "float32",
          "shape": [4],
          "data": { "kind": "fillFloat32", "sinStep": 0.11, "cosStep": 0.07, "scale": 0.02, "offset": 0.06 }
        }
      },
      "outputs": { "yT": { "dtype": "float32", "shape": [3, 4], "tolerance": 0.00001 } }
    },
    {
      "name": "nf4_scalar_block_straddle",
      "attrs": { "K": 24, "N": 3, "block_size": 16, "quant_type": 1 },
      "inputs": {
        "aT": {
          "dtype": "float32",
          "shape": [2, 24],
          "data": { "kind": "fillFloat32", "sinStep": 0.17, "cosStep": 0.31 }
        },
        "bT": {
          "dtype": "uint8",
          "shape": [36],
          "data": { "kind": "cycle", "values": { "$ref": "#/fixtureArrays/bnb4_weight_cycle_b_t" } }
        },
        "absmaxT": {
          "dtype": "float32",
          "shape": [5],
          "data": { "kind": "fillFloat32", "sinStep": 0.11, "cosStep": 0.07, "scale": 0.02, "offset": 0.06 }
        }
      },
      "outputs": { "yT": { "dtype": "float32", "shape": [2, 3], "tolerance": 0.00001 } }
    },
    {
      "name": "nf4_scalar_odd_k",
      "attrs": { "K": 33, "N": 5, "block_size": 16, "quant_type": 1 },
      "inputs": {
        "aT": {
          "dtype": "float32",
          "shape": [3, 33],
          "data": { "kind": "fillFloat32", "sinStep": 0.17, "cosStep": 0.31 }
        },
        "bT": {
          "dtype": "uint8",
          "shape": [83],
          "data": { "kind": "cycle", "values": { "$ref": "#/fixtureArrays/bnb4_weight_cycle_b_t" } }
        },
        "absmaxT": {
          "dtype": "float32",
          "shape": [11],
          "data": { "kind": "fillFloat32", "sinStep": 0.11, "cosStep": 0.07, "scale": 0.02, "offset": 0.06 }
        }
      },
      "outputs": { "yT": { "dtype": "float32", "shape": [3, 5], "tolerance": 0.00001 } }
    },
    {
      "name": "nf4_gemv_decode",
      "attrs": { "K": 128, "N": 64, "block_size": 64, "quant_type": 1 },
      "inputs": {
        "aT": {
          "dtype": "float32",
          "shape": [1, 128],
          "data": { "kind": "fillFloat32", "sinStep": 0.17, "cosStep": 0.31 }
        },
        "bT": {
          "dtype": "uint8",
          "shape": [4096],
          "data": { "kind": "cycle", "values": { "$ref": "#/fixtureArrays/bnb4_weight_cycle_b_t" } }
        },
        "absmaxT": {
          "dtype": "float32",
          "shape": [128],
          "data": { "kind": "fillFloat32", "sinStep": 0.11, "cosStep": 0.07, "scale": 0.02, "offset": 0.06 }
        }
      },
      "outputs": { "yT": { "dtype": "float32", "shape": [1, 64], "tolerance": 0.00002 } }
    },
    {
      "name": "fp4_gemv_decode",
      "attrs": { "K": 64, "N": 32, "block_size": 32, "quant_type": 0 },
      "inputs": {
        "aT": {
          "dtype": "float32",
          "shape": [1, 64],
          "data": { "kind": "fillFloat32", "sinStep": 0.17, "cosStep": 0.31 }
        },
        "bT": {
          "dtype": "uint8",
          "shape": [1024],
          "data": { "kind": "cycle", "values": { "$ref": "#/fixtureArrays/bnb4_weight_cycle_b_t" } }
        },
        "absmaxT": {
          "dtype": "float32",
          "shape": [64],
          "data": { "kind": "fillFloat32", "sinStep": 0.11, "cosStep": 0.07, "scale": 0.02, "offset": 0.06 }
        }
      },
      "outputs": { "yT": { "dtype": "float32", "shape": [1, 32], "tolerance": 0.00001 } }
    },
    {
      "name": "nf4_gemv_ragged_k",
      "attrs": { "K": 100, "N": 12, "block_size": 16, "quant_type": 1 },
      "inputs": {
        "aT": {
          "dtype": "float32",
          "shape": [1, 100],
          "data": { "kind": "fillFloat32", "sinStep": 0.17, "cosStep": 0.31 }
        },
        "bT": {
          "dtype": "uint8",
          "shape": [600],
          "data": { "kind": "cycle", "values": { "$ref": "#/fixtureArrays/bnb4_weight_cycle_b_t" } }
        },
        "absmaxT": {
          "dtype": "float32",
          "shape": [75],
          "data": { "kind": "fillFloat32", "sinStep": 0.11, "cosStep": 0.07, "scale": 0.02, "offset": 0.06 }
        }
      },
      "outputs": { "yT": { "dtype": "float32", "shape": [1, 12], "tolerance": 0.00001 } }
    },
    {
      "name": "nf4_tiled_prefill",
      "attrs": { "K": 64, "N": 48, "block_size": 32, "quant_type": 1 },
      "inputs": {
        "aT": {
          "dtype": "float32",
          "shape": [32, 64],
          "data": { "kind": "fillFloat32", "sinStep": 0.17, "cosStep": 0.31 }
        },
        "bT": {
          "dtype": "uint8",
          "shape": [1536],
          "data": { "kind": "cycle", "values": { "$ref": "#/fixtureArrays/bnb4_weight_cycle_b_t" } }
        },
        "absmaxT": {
          "dtype": "float32",
          "shape": [96],
          "data": { "kind": "fillFloat32", "sinStep": 0.11, "cosStep": 0.07, "scale": 0.02, "offset": 0.06 }
        }
      },
      "outputs": { "yT": { "dtype": "float32", "shape": [32, 48], "tolerance": 0.00002 } }
    },
    {
      "name": "fp4_tiled_prefill",
      "attrs": { "K": 64, "N": 48, "block_size": 32, "quant_type": 0 },
      "inputs": {
        "aT": {
          "dtype": "float32",
          "shape": [32, 64],
          "data": { "kind": "fillFloat32", "sinStep": 0.17, "cosStep": 0.31 }
        },
        "bT": {
          "dtype": "uint8",
          "shape": [1536],
          "data": { "kind": "cycle", "values": { "$ref": "#/fixtureArrays/bnb4_weight_cycle_b_t" } }
        },
        "absmaxT": {
          "dtype": "float32",
          "shape": [96],
          "data": { "kind": "fillFloat32", "sinStep": 0.11, "cosStep": 0.07, "scale": 0.02, "offset": 0.06 }
        }
      },
      "outputs": { "yT": { "dtype": "float32", "shape": [32, 48], "tolerance": 0.00002 } }
    },
    {
      "name": "nf4_sgmat_tall_aligned",
      "attrs": { "K": 64, "N": 64, "block_size": 32, "quant_type": 1 },
      "inputs": {
        "aT": {
          "dtype": "float32",
          "shape": [128, 64],
          "data": { "kind": "fillFloat32", "sinStep": 0.17, "cosStep": 0.31 }
        },
        "bT": {
          "dtype": "uint8",
          "shape": [2048],
          "data": { "kind": "cycle", "values": { "$ref": "#/fixtureArrays/bnb4_weight_cycle_b_t" } }
        },
        "absmaxT": {
          "dtype": "float32",
          "shape": [128],
          "data": { "kind": "fillFloat32", "sinStep": 0.11, "cosStep": 0.07, "scale": 0.02, "offset": 0.06 }
        }
      },
      "outputs": { "yT": { "dtype": "float32", "shape": [128, 64], "tolerance": 0.00002 } }
    },
    {
      "name": "fp4_sgmat_aligned",
      "attrs": { "K": 64, "N": 64, "block_size": 32, "quant_type": 0 },
      "inputs": {
        "aT": {
          "dtype": "float32",
          "shape": [64, 64],
          "data": { "kind": "fillFloat32", "sinStep": 0.17, "cosStep": 0.31 }
        },
        "bT": {
          "dtype": "uint8",
          "shape": [2048],
          "data": { "kind": "cycle", "values": { "$ref": "#/fixtureArrays/bnb4_weight_cycle_b_t" } }
        },
        "absmaxT": {
          "dtype": "float32",
          "shape": [128],
          "data": { "kind": "fillFloat32", "sinStep": 0.11, "cosStep": 0.07, "scale": 0.02, "offset": 0.06 }
        }
      },
      "outputs": { "yT": { "dtype": "float32", "shape": [64, 64], "tolerance": 0.00002 } }
    },
    {
      "name": "nf4_tiled_tail",
      "attrs": { "K": 33, "N": 19, "block_size": 16, "quant_type": 1 },
      "inputs": {
        "aT": {
          "dtype": "float32",
          "shape": [20, 33],
          "data": { "kind": "fillFloat32", "sinStep": 0.17, "cosStep": 0.31 }
        },
        "bT": {
          "dtype": "uint8",
          "shape": [314],
          "data": { "kind": "cycle", "values": { "$ref": "#/fixtureArrays/bnb4_weight_cycle_b_t" } }
        },
        "absmaxT": {
          "dtype": "float32",
          "shape": [40],
          "data": { "kind": "fillFloat32", "sinStep": 0.11, "cosStep": 0.07, "scale": 0.02, "offset": 0.06 }
        }
      },
      "outputs": { "yT": { "dtype": "float32", "shape": [20, 19], "tolerance": 0.00001 } }
    },
    {
      "name": "nf4_tiled_tail_bk32",
      "attrs": { "K": 33, "N": 19, "block_size": 16, "quant_type": 1 },
      "tunables": { "PORTABLE_TILE_K": 32 },
      "inputs": {
        "aT": {
          "dtype": "float32",
          "shape": [20, 33],
          "data": { "kind": "fillFloat32", "sinStep": 0.17, "cosStep": 0.31 }
        },
        "bT": {
          "dtype": "uint8",
          "shape": [314],
          "data": { "kind": "cycle", "values": { "$ref": "#/fixtureArrays/bnb4_weight_cycle_b_t" } }
        },
        "absmaxT": {
          "dtype": "float32",
          "shape": [40],
          "data": { "kind": "fillFloat32", "sinStep": 0.11, "cosStep": 0.07, "scale": 0.02, "offset": 0.06 }
        }
      },
      "outputs": { "yT": { "dtype": "float32", "shape": [20, 19], "tolerance": 0.00001 } }
    },
    {
      "name": "fp4_tiled_tail_bk64",
      "attrs": { "K": 33, "N": 19, "block_size": 16, "quant_type": 0 },
      "tunables": { "PORTABLE_TILE_K": 64 },
      "inputs": {
        "aT": {
          "dtype": "float32",
          "shape": [20, 33],
          "data": { "kind": "fillFloat32", "sinStep": 0.17, "cosStep": 0.31 }
        },
        "bT": {
          "dtype": "uint8",
          "shape": [314],
          "data": { "kind": "cycle", "values": { "$ref": "#/fixtureArrays/bnb4_weight_cycle_b_t" } }
        },
        "absmaxT": {
          "dtype": "float32",
          "shape": [40],
          "data": { "kind": "fillFloat32", "sinStep": 0.11, "cosStep": 0.07, "scale": 0.02, "offset": 0.06 }
        }
      },
      "outputs": { "yT": { "dtype": "float32", "shape": [20, 19], "tolerance": 0.00001 } }
    },
    {
      "name": "nf4_f16_scalar",
      "attrs": { "K": 32, "N": 8, "block_size": 16, "quant_type": 1 },
      "inputs": {
        "aT": {
          "dtype": "float16",
          "shape": [3, 32],
          "data": { "kind": "fillFloat32", "sinStep": 0.17, "cosStep": 0.31 }
        },
        "bT": {
          "dtype": "uint8",
          "shape": [128],
          "data": { "kind": "cycle", "values": { "$ref": "#/fixtureArrays/bnb4_weight_cycle_b_t" } }
        },
        "absmaxT": {
          "dtype": "float16",
          "shape": [16],
          "data": { "kind": "fillFloat32", "sinStep": 0.11, "cosStep": 0.07, "scale": 0.02, "offset": 0.06 }
        }
      },
      "outputs": { "yT": { "dtype": "float16", "shape": [3, 8], "tolerance": 0.02 } }
    },
    {
      "name": "nf4_f16_gemv",
      "attrs": { "K": 64, "N": 16, "block_size": 32, "quant_type": 1 },
      "inputs": {
        "aT": {
          "dtype": "float16",
          "shape": [1, 64],
          "data": { "kind": "fillFloat32", "sinStep": 0.17, "cosStep": 0.31 }
        },
        "bT": {
          "dtype": "uint8",
          "shape": [512],
          "data": { "kind": "cycle", "values": { "$ref": "#/fixtureArrays/bnb4_weight_cycle_b_t" } }
        },
        "absmaxT": {
          "dtype": "float16",
          "shape": [32],
          "data": { "kind": "fillFloat32", "sinStep": 0.11, "cosStep": 0.07, "scale": 0.02, "offset": 0.06 }
        }
      },
      "outputs": { "yT": { "dtype": "float16", "shape": [1, 16], "tolerance": 0.02 } }
    },
    {
      "name": "nf4_f16_tiled",
      "attrs": { "K": 32, "N": 16, "block_size": 32, "quant_type": 1 },
      "inputs": {
        "aT": {
          "dtype": "float16",
          "shape": [16, 32],
          "data": { "kind": "fillFloat32", "sinStep": 0.17, "cosStep": 0.31 }
        },
        "bT": {
          "dtype": "uint8",
          "shape": [256],
          "data": { "kind": "cycle", "values": { "$ref": "#/fixtureArrays/bnb4_weight_cycle_b_t" } }
        },
        "absmaxT": {
          "dtype": "float16",
          "shape": [16],
          "data": { "kind": "fillFloat32", "sinStep": 0.11, "cosStep": 0.07, "scale": 0.02, "offset": 0.06 }
        }
      },
      "outputs": { "yT": { "dtype": "float16", "shape": [16, 16], "tolerance": 0.02 } }
    },
    {
      "name": "nf4_f16_tiled_longk_m32_k512_n32",
      "attrs": { "K": 512, "N": 32, "block_size": 64, "quant_type": 1 },
      "inputs": {
        "aT": {
          "dtype": "float16",
          "shape": [32, 512],
          "data": { "kind": "fillFloat32", "sinStep": 0.17, "cosStep": 0.31 }
        },
        "bT": {
          "dtype": "uint8",
          "shape": [8192],
          "data": { "kind": "cycle", "values": { "$ref": "#/fixtureArrays/bnb4_weight_cycle_b_t" } }
        },
        "absmaxT": {
          "dtype": "float16",
          "shape": [256],
          "data": { "kind": "fillFloat32", "sinStep": 0.11, "cosStep": 0.07, "scale": 0.02, "offset": 0.06 }
        }
      },
      "outputs": { "yT": { "dtype": "float16", "shape": [32, 32], "tolerance": 0.001, "relTolerance": 0.001 } },
      "provenance": {
        "notes": "The other f16 fixtures all run K <= 64, where an f16 accumulator is indistinguishable from an f32 one. This one accumulates over K=512 so the tiled path's f32 accumulation is actually pinned."
      }
    },
    {
      "name": "nf4_sgmat_widened_f16",
      "provenance": {
        "notes": "A float16 request on the aligned subgroup-matrix shape. The matrix units take float32 operands and accumulate in float32, so this path widens the activations either side of the multiply rather than using the device's float16 configuration, whose result component is also float16. It is the only case that reaches the widening passes."
      },
      "attrs": { "K": 64, "N": 64, "block_size": 32, "quant_type": 1 },
      "inputs": {
        "aT": {
          "dtype": "float16",
          "shape": [128, 64],
          "data": { "kind": "fillFloat32", "sinStep": 0.17, "cosStep": 0.31 }
        },
        "bT": {
          "dtype": "uint8",
          "shape": [2048],
          "data": { "kind": "cycle", "values": { "$ref": "#/fixtureArrays/bnb4_weight_cycle_b_t" } }
        },
        "absmaxT": {
          "dtype": "float16",
          "shape": [128],
          "data": { "kind": "fillFloat32", "sinStep": 0.11, "cosStep": 0.07, "scale": 0.02, "offset": 0.06 }
        }
      },
      "outputs": { "yT": { "dtype": "float16", "shape": [128, 64], "tolerance": 0.002, "relTolerance": 0.002 } }
    }
  ]
}