{
  "scenario": "RVA",
  "backend": "efficient",
  "sequence": 81547,
  "batch": 1,
  "heads": 28,
  "head_dim": 128,
  "pid": 2815991,
  "torch": "2.11.0+cu130",
  "cuda": "13.0",
  "cudnn": 91900,
  "gpu": "NVIDIA RTX 6000D",
  "capability": [
    12,
    0
  ],
  "started_at": 1787624711.4254694,
  "results": [
    {
      "layout": "model_strided",
      "dtype": "torch.bfloat16",
      "q_shape": [
        1,
        28,
        81547,
        128
      ],
      "q_stride": [
        876793344,
        128,
        10752,
        1
      ],
      "k_stride": [
        876793344,
        128,
        10752,
        1
      ],
      "v_stride": [
        876793344,
        128,
        10752,
        1
      ],
      "q_contiguous": false,
      "warm": {
        "samples_ms": [
          1349.029296875,
          1351.04248046875,
          1352.032958984375,
          1352.2147216796875,
          1352.338623046875,
          1352.349853515625,
          1352.5400390625,
          1352.6226806640625,
          1352.8155517578125,
          1353.0452880859375,
          1353.33203125,
          1353.4384765625
        ],
        "min_ms": 1349.029296875,
        "median_ms": 1352.4449462890625,
        "mean_ms": 1352.2335001627605,
        "p90_ms": 1353.33203125,
        "max_ms": 1353.4384765625,
        "stdev_ms": 1.1928786749511278,
        "cv": 0.0008821543578143478
      },
      "cold": {
        "samples_ms": [
          1347.048583984375,
          1347.3988037109375,
          1347.517333984375,
          1347.9693603515625,
          1348.06005859375
        ],
        "min_ms": 1347.048583984375,
        "median_ms": 1347.517333984375,
        "mean_ms": 1347.598828125,
        "p90_ms": 1348.06005859375,
        "max_ms": 1348.06005859375,
        "stdev_ms": 0.418164845394166,
        "cv": 0.00031030365763673554
      },
      "cold_over_warm": 0.9963565154217862,
      "effective_tflops": 70.48949092220434,
      "io_lower_bound_gbps": 1.7288064777908283,
      "cuda_graph": {
        "supported": true,
        "samples_ms": [
          1347.1910400390625,
          1347.529541015625,
          1347.7379150390625,
          1347.979736328125,
          1348.06298828125,
          1348.1065673828125,
          1348.116943359375,
          1348.1597900390625,
          1348.1629638671875,
          1348.2376708984375,
          1348.326171875,
          1348.407958984375
        ],
        "min_ms": 1347.1910400390625,
        "median_ms": 1348.1117553710938,
        "mean_ms": 1348.0016072591145,
        "p90_ms": 1348.326171875,
        "max_ms": 1348.407958984375,
        "stdev_ms": 0.35100013696322385,
        "cv": 0.0002603855478161564
      },
      "output_checksum": 8.007740461835056e-07,
      "trace": "/data/wxy/profile_results/h3-sdpa-kernel-matrix-20260825-run1/jobs/gpu6-RVA-efficient-model_strided.trace.json",
      "trace_error": null
    },
    {
      "layout": "contiguous",
      "dtype": "torch.bfloat16",
      "q_shape": [
        1,
        28,
        81547,
        128
      ],
      "q_stride": [
        292264448,
        10438016,
        128,
        1
      ],
      "k_stride": [
        292264448,
        10438016,
        128,
        1
      ],
      "v_stride": [
        292264448,
        10438016,
        128,
        1
      ],
      "q_contiguous": true,
      "warm": {
        "samples_ms": [
          1347.3359375,
          1347.4439697265625,
          1347.51708984375,
          1347.669189453125,
          1347.7310791015625,
          1347.80322265625,
          1347.81591796875,
          1347.8663330078125,
          1347.9178466796875,
          1347.9669189453125,
          1347.970458984375,
          1348.2000732421875
        ],
        "min_ms": 1347.3359375,
        "median_ms": 1347.8095703125,
        "mean_ms": 1347.7698364257812,
        "p90_ms": 1347.970458984375,
        "max_ms": 1348.2000732421875,
        "stdev_ms": 0.24654366841173367,
        "cv": 0.00018292713024766546
      },
      "cold": {
        "samples_ms": [
          1346.7711181640625,
          1347.3646240234375,
          1347.4847412109375,
          1347.4930419921875,
          1348.5291748046875
        ],
        "min_ms": 1346.7711181640625,
        "median_ms": 1347.4847412109375,
        "mean_ms": 1347.5285400390626,
        "p90_ms": 1348.5291748046875,
        "max_ms": 1348.5291748046875,
        "stdev_ms": 0.6334439082826026,
        "cv": 0.0004700782873691418
      },
      "cold_over_warm": 0.9997589948100107,
      "effective_tflops": 70.7319178199041,
      "io_lower_bound_gbps": 1.7347521753075918,
      "cuda_graph": {
        "supported": true,
        "samples_ms": [
          1347.15966796875,
          1347.5003662109375,
          1347.50927734375,
          1347.5654296875,
          1347.6072998046875,
          1347.6678466796875,
          1347.7098388671875,
          1347.892333984375,
          1347.9625244140625,
          1347.9769287109375,
          1347.9835205078125,
          1348.0106201171875
        ],
        "min_ms": 1347.15966796875,
        "median_ms": 1347.6888427734375,
        "mean_ms": 1347.712137858073,
        "p90_ms": 1347.9835205078125,
        "max_ms": 1348.0106201171875,
        "stdev_ms": 0.2617636624382269,
        "cv": 0.00019422817015972673
      },
      "output_checksum": 5.330903150024824e-07,
      "trace": "/data/wxy/profile_results/h3-sdpa-kernel-matrix-20260825-run1/jobs/gpu6-RVA-efficient-contiguous.trace.json",
      "trace_error": null
    }
  ],
  "status": "success",
  "finished_at": 1787624805.9117982
}