{
  "scenario": "RVA",
  "backend": "auto",
  "sequence": 81547,
  "batch": 1,
  "heads": 28,
  "head_dim": 128,
  "pid": 2815992,
  "torch": "2.11.0+cu130",
  "cuda": "13.0",
  "cudnn": 91900,
  "gpu": "NVIDIA RTX 6000D",
  "capability": [
    12,
    0
  ],
  "started_at": 1787624711.8685164,
  "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": [
          667.9579467773438,
          668.0010986328125,
          668.0359497070312,
          668.04931640625,
          668.0708618164062,
          668.1140747070312,
          668.131591796875,
          668.1500854492188,
          668.1597290039062,
          668.1597900390625,
          668.2352905273438,
          668.3551025390625
        ],
        "min_ms": 667.9579467773438,
        "median_ms": 668.1228332519531,
        "mean_ms": 668.1184031168619,
        "p90_ms": 668.2352905273438,
        "max_ms": 668.3551025390625,
        "stdev_ms": 0.10772217149971701,
        "cv": 0.00016123215734992275
      },
      "cold": {
        "samples_ms": [
          667.9595336914062,
          667.996337890625,
          668.0238647460938,
          668.0654296875,
          668.4395751953125
        ],
        "min_ms": 667.9595336914062,
        "median_ms": 668.0238647460938,
        "mean_ms": 668.0969482421875,
        "p90_ms": 668.4395751953125,
        "max_ms": 668.4395751953125,
        "stdev_ms": 0.19540537819270587,
        "cv": 0.00029248057292707575
      },
      "cold_over_warm": 0.9998518707924146,
      "effective_tflops": 142.68806725285694,
      "io_lower_bound_gbps": 3.499529529053354,
      "cuda_graph": {
        "supported": true,
        "samples_ms": [
          665.370361328125,
          665.5252685546875,
          665.6463623046875,
          665.66748046875,
          665.778564453125,
          665.8983764648438,
          665.9188842773438,
          665.9459228515625,
          666.39453125,
          666.6050415039062,
          666.6434936523438,
          669.261962890625
        ],
        "min_ms": 665.370361328125,
        "median_ms": 665.9086303710938,
        "mean_ms": 666.2213541666666,
        "p90_ms": 666.6434936523438,
        "max_ms": 669.261962890625,
        "stdev_ms": 1.040740684982182,
        "cv": 0.001562154497860516
      },
      "output_checksum": 8.005773111108283e-07,
      "trace": "/data/wxy/profile_results/h3-sdpa-kernel-matrix-20260825-run1/jobs/gpu7-RVA-auto-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": [
          665.2757568359375,
          665.328369140625,
          665.515625,
          665.9142456054688,
          665.9380493164062,
          665.9698486328125,
          665.9846801757812,
          666.00048828125,
          666.1372680664062,
          666.1569213867188,
          666.6114501953125,
          666.7260131835938
        ],
        "min_ms": 665.2757568359375,
        "median_ms": 665.9772644042969,
        "mean_ms": 665.9632263183594,
        "p90_ms": 666.6114501953125,
        "max_ms": 666.7260131835938,
        "stdev_ms": 0.44218226807130984,
        "cv": 0.0006639740012610359
      },
      "cold": {
        "samples_ms": [
          665.6121215820312,
          665.6764526367188,
          665.8822631835938,
          665.883544921875,
          665.9204711914062
        ],
        "min_ms": 665.6121215820312,
        "median_ms": 665.8822631835938,
        "mean_ms": 665.794970703125,
        "p90_ms": 665.9204711914062,
        "max_ms": 665.9204711914062,
        "stdev_ms": 0.14026410364186667,
        "cv": 0.00021067161786118358
      },
      "cold_over_warm": 0.9998573506547733,
      "effective_tflops": 143.14776323407608,
      "io_lower_bound_gbps": 3.5108039102376813,
      "cuda_graph": {
        "supported": true,
        "samples_ms": [
          665.3902587890625,
          665.5931396484375,
          665.6646728515625,
          665.6961669921875,
          665.746826171875,
          665.7867431640625,
          665.7907104492188,
          665.8287353515625,
          665.9599609375,
          665.9778442382812,
          666.2161865234375,
          666.3693237304688
        ],
        "min_ms": 665.3902587890625,
        "median_ms": 665.7887268066406,
        "mean_ms": 665.8350474039713,
        "p90_ms": 666.2161865234375,
        "max_ms": 666.3693237304688,
        "stdev_ms": 0.2668853979731874,
        "cv": 0.0004008281015151555
      },
      "output_checksum": 5.333090484782588e-07,
      "trace": "/data/wxy/profile_results/h3-sdpa-kernel-matrix-20260825-run1/jobs/gpu7-RVA-auto-contiguous.trace.json",
      "trace_error": null
    }
  ],
  "status": "success",
  "finished_at": 1787624759.8206172
}