{
  "scenario": "F3",
  "backend": "efficient",
  "sequence": 41778,
  "batch": 1,
  "heads": 28,
  "head_dim": 128,
  "pid": 2815987,
  "torch": "2.11.0+cu130",
  "cuda": "13.0",
  "cudnn": 91900,
  "gpu": "NVIDIA RTX 6000D",
  "capability": [
    12,
    0
  ],
  "started_at": 1787624712.0639284,
  "results": [
    {
      "layout": "model_strided",
      "dtype": "torch.bfloat16",
      "q_shape": [
        1,
        28,
        41778,
        128
      ],
      "q_stride": [
        449197056,
        128,
        10752,
        1
      ],
      "k_stride": [
        449197056,
        128,
        10752,
        1
      ],
      "v_stride": [
        449197056,
        128,
        10752,
        1
      ],
      "q_contiguous": false,
      "warm": {
        "samples_ms": [
          355.5981750488281,
          355.6229248046875,
          355.778076171875,
          355.85919189453125,
          355.8716125488281,
          355.8981018066406,
          355.9027099609375,
          355.95330810546875,
          355.9855041503906,
          355.9941101074219,
          356.01824951171875,
          356.0218811035156
        ],
        "min_ms": 355.5981750488281,
        "median_ms": 355.90040588378906,
        "mean_ms": 355.8753204345703,
        "p90_ms": 356.01824951171875,
        "max_ms": 356.0218811035156,
        "stdev_ms": 0.14325910598831795,
        "cv": 0.0004025542030095817
      },
      "cold": {
        "samples_ms": [
          355.445556640625,
          355.473388671875,
          355.4867858886719,
          356.5604553222656,
          357.07843017578125
        ],
        "min_ms": 355.445556640625,
        "median_ms": 355.4867858886719,
        "mean_ms": 356.00892333984376,
        "p90_ms": 357.07843017578125,
        "max_ms": 357.07843017578125,
        "stdev_ms": 0.7623711917518489,
        "cv": 0.0021414384353060003
      },
      "cold_over_warm": 0.9988378209513696,
      "effective_tflops": 70.30639019724629,
      "io_lower_bound_gbps": 3.3657135428812426,
      "cuda_graph": {
        "supported": true,
        "samples_ms": [
          355.1769714355469,
          355.2284240722656,
          355.33447265625,
          355.8866271972656,
          355.9685974121094,
          355.9689025878906,
          356.0318603515625,
          356.0379333496094,
          356.03936767578125,
          356.0577087402344,
          356.1680908203125,
          357.1600341796875
        ],
        "min_ms": 355.1769714355469,
        "median_ms": 356.00038146972656,
        "mean_ms": 355.9215825398763,
        "p90_ms": 356.1680908203125,
        "max_ms": 357.1600341796875,
        "stdev_ms": 0.5255226851179629,
        "cv": 0.001476512554725689
      },
      "output_checksum": 2.1510920760192676e-06,
      "trace": "/data/wxy/profile_results/h3-sdpa-kernel-matrix-20260825-run1/jobs/gpu2-F3-efficient-model_strided.trace.json",
      "trace_error": null
    },
    {
      "layout": "contiguous",
      "dtype": "torch.bfloat16",
      "q_shape": [
        1,
        28,
        41778,
        128
      ],
      "q_stride": [
        149732352,
        5347584,
        128,
        1
      ],
      "k_stride": [
        149732352,
        5347584,
        128,
        1
      ],
      "v_stride": [
        149732352,
        5347584,
        128,
        1
      ],
      "q_contiguous": true,
      "warm": {
        "samples_ms": [
          354.3680114746094,
          354.7764587402344,
          355.1002197265625,
          355.64630126953125,
          355.6904296875,
          355.6906433105469,
          355.7364501953125,
          355.7505187988281,
          355.7789001464844,
          355.8151550292969,
          355.8532409667969,
          355.8674011230469
        ],
        "min_ms": 354.3680114746094,
        "median_ms": 355.7135467529297,
        "mean_ms": 355.5061442057292,
        "p90_ms": 355.8532409667969,
        "max_ms": 355.8674011230469,
        "stdev_ms": 0.48746400713134436,
        "cv": 0.001371183072575117
      },
      "cold": {
        "samples_ms": [
          353.8484802246094,
          353.9283447265625,
          354.0924072265625,
          354.1081237792969,
          354.134521484375
        ],
        "min_ms": 353.8484802246094,
        "median_ms": 354.0924072265625,
        "mean_ms": 354.0223754882812,
        "p90_ms": 354.134521484375,
        "max_ms": 354.134521484375,
        "stdev_ms": 0.12640736241965148,
        "cv": 0.00035706037576101115
      },
      "cold_over_warm": 0.9954425701771398,
      "effective_tflops": 70.34332269837265,
      "io_lower_bound_gbps": 3.3674815787434844,
      "cuda_graph": {
        "supported": true,
        "samples_ms": [
          354.05169677734375,
          354.11456298828125,
          354.2168273925781,
          354.2208251953125,
          354.23431396484375,
          354.2758483886719,
          354.3400573730469,
          354.37261962890625,
          354.38580322265625,
          354.4937744140625,
          354.5145263671875,
          354.607666015625
        ],
        "min_ms": 354.05169677734375,
        "median_ms": 354.3079528808594,
        "mean_ms": 354.3190434773763,
        "p90_ms": 354.5145263671875,
        "max_ms": 354.607666015625,
        "stdev_ms": 0.16576069135605714,
        "cv": 0.00046782890845843336
      },
      "output_checksum": 2.1056371224403847e-06,
      "trace": "/data/wxy/profile_results/h3-sdpa-kernel-matrix-20260825-run1/jobs/gpu2-F3-efficient-contiguous.trace.json",
      "trace_error": null
    }
  ],
  "status": "success",
  "finished_at": 1787624738.742594
}