{
  "schema_version": 1,
  "protocol": {
    "scope": "complete mixer preparation, recurrence, input and parameter gradients",
    "dtype": "BF16 autocast and FP32 recurrent carry",
    "input_gradient": true,
    "order": "AB/BA counterbalanced",
    "compile_excluded": true,
    "baseline_policy": "strongest previously calibrated legal RG-LRU backend per shape",
    "samu_policy": "strongest previously calibrated legal SAMU backend per shape",
    "resets": "independent sequences reset at first token; no internal packed reset"
  },
  "environment": {
    "torch": "2.1.2+cu118",
    "cuda": "11.8",
    "gpu": "NVIDIA H800 PCIe"
  },
  "rows": [
    {
      "shape": {
        "batch": 4,
        "length": 2048,
        "width": 2048
      },
      "architecture": "rglru",
      "backend": "fused_grouped_prefix32_hierarchical_chunk16",
      "controller": null,
      "forward": {
        "samples_ms": [
          0.9279027581214905,
          0.9122826159000397,
          0.9021908044815063,
          0.9017027914524078,
          0.9072422981262207,
          0.9075775742530823,
          0.8990168571472168,
          0.9066574275493622,
          0.8933767676353455,
          0.890243798494339
        ],
        "median_ms": 0.9044241160154343,
        "mean_ms": 0.9048193693161011,
        "std_ms": 0.009995356827149377,
        "minimum_ms": 0.890243798494339,
        "maximum_ms": 0.9279027581214905
      },
      "backward": {
        "samples_ms": [
          1.5889070928096771,
          1.5861056745052338,
          1.5913508832454681,
          1.590002328157425,
          1.5919022262096405,
          1.5796571969985962,
          1.5892498195171356,
          1.5820302069187164,
          1.579616218805313,
          1.583687961101532
        ],
        "median_ms": 1.5875063836574554,
        "mean_ms": 1.5862509608268738,
        "std_ms": 0.004471792916509476,
        "minimum_ms": 1.579616218805313,
        "maximum_ms": 1.5919022262096405
      },
      "forward_backward": {
        "samples_ms": [
          2.552807331085205,
          2.567969262599945,
          2.5577284395694733,
          2.5533176958560944,
          2.5453120470046997,
          2.5247856974601746,
          2.542320638895035,
          2.55802646279335,
          2.535700798034668,
          2.5539137423038483
        ],
        "median_ms": 2.5530625134706497,
        "mean_ms": 2.5491882115602493,
        "std_ms": 0.011814569771883056,
        "minimum_ms": 2.5247856974601746,
        "maximum_ms": 2.567969262599945
      },
      "rounds": [
        {
          "round": 0,
          "order_index": 0,
          "forward_ms": [
            0.9279027581214905,
            0.9122826159000397,
            0.9021908044815063,
            0.9017027914524078,
            0.9072422981262207
          ],
          "backward_ms": [
            1.5889070928096771,
            1.5861056745052338,
            1.5913508832454681,
            1.590002328157425,
            1.5919022262096405
          ],
          "forward_backward_ms": [
            2.552807331085205,
            2.567969262599945,
            2.5577284395694733,
            2.5533176958560944,
            2.5453120470046997
          ]
        },
        {
          "round": 1,
          "order_index": 1,
          "forward_ms": [
            0.9075775742530823,
            0.8990168571472168,
            0.9066574275493622,
            0.8933767676353455,
            0.890243798494339
          ],
          "backward_ms": [
            1.5796571969985962,
            1.5892498195171356,
            1.5820302069187164,
            1.579616218805313,
            1.583687961101532
          ],
          "forward_backward_ms": [
            2.5247856974601746,
            2.542320638895035,
            2.55802646279335,
            2.535700798034668,
            2.5539137423038483
          ]
        }
      ],
      "peak_allocated_delta_bytes": 387392512,
      "peak_reserved_delta_bytes": 478150656,
      "peak_allocated_bytes": 523784704,
      "peak_reserved_bytes": 650117120,
      "baseline_allocated_bytes": 136392192,
      "baseline_reserved_bytes": 171966464,
      "parameters": 530432,
      "parameter_bytes": 2121728,
      "dynamic_control_values_per_token": 4096,
      "materialized_state_sized_transition": false
    },
    {
      "shape": {
        "batch": 4,
        "length": 2048,
        "width": 2048
      },
      "architecture": "samu",
      "backend": "fused_output_fused_write_shared_sfu_chunk32",
      "controller": "fp32_fused_coords",
      "forward": {
        "samples_ms": [
          0.7736720144748688,
          0.7423795759677887,
          0.7291920483112335,
          0.7657632231712341,
          0.7558241486549377,
          0.7706284523010254,
          0.7439553737640381,
          0.7278844714164734,
          0.7242672145366669,
          0.7194466888904572
        ],
        "median_ms": 0.7431674748659134,
        "mean_ms": 0.7453013211488724,
        "std_ms": 0.01913080837473192,
        "minimum_ms": 0.7194466888904572,
        "maximum_ms": 0.7736720144748688
      },
      "backward": {
        "samples_ms": [
          1.0979995131492615,
          1.1016465723514557,
          1.100648194551468,
          1.09945610165596,
          1.0985955595970154,
          1.1018961668014526,
          1.0969676077365875,
          1.0961219668388367,
          1.0954923927783966,
          1.0974667966365814
        ],
        "median_ms": 1.0982975363731384,
        "mean_ms": 1.0986290872097015,
        "std_ms": 0.0021253987511796043,
        "minimum_ms": 1.0954923927783966,
        "maximum_ms": 1.1018961668014526
      },
      "forward_backward": {
        "samples_ms": [
          1.9353777170181274,
          1.9312873482704163,
          1.9532255828380585,
          1.9252672791481018,
          1.920856535434723,
          1.9226931035518646,
          1.912292093038559,
          1.9087344408035278,
          1.9416026771068573,
          1.914583146572113
        ],
        "median_ms": 1.9239801913499832,
        "mean_ms": 1.9265919923782349,
        "std_ms": 0.013233848396225198,
        "minimum_ms": 1.9087344408035278,
        "maximum_ms": 1.9532255828380585
      },
      "rounds": [
        {
          "round": 0,
          "order_index": 1,
          "forward_ms": [
            0.7736720144748688,
            0.7423795759677887,
            0.7291920483112335,
            0.7657632231712341,
            0.7558241486549377
          ],
          "backward_ms": [
            1.0979995131492615,
            1.1016465723514557,
            1.100648194551468,
            1.09945610165596,
            1.0985955595970154
          ],
          "forward_backward_ms": [
            1.9353777170181274,
            1.9312873482704163,
            1.9532255828380585,
            1.9252672791481018,
            1.920856535434723
          ]
        },
        {
          "round": 1,
          "order_index": 0,
          "forward_ms": [
            0.7706284523010254,
            0.7439553737640381,
            0.7278844714164734,
            0.7242672145366669,
            0.7194466888904572
          ],
          "backward_ms": [
            1.1018961668014526,
            1.0969676077365875,
            1.0961219668388367,
            1.0954923927783966,
            1.0974667966365814
          ],
          "forward_backward_ms": [
            1.9226931035518646,
            1.912292093038559,
            1.9087344408035278,
            1.9416026771068573,
            1.914583146572113
          ]
        }
      ],
      "peak_allocated_delta_bytes": 302249472,
      "peak_reserved_delta_bytes": 274726912,
      "peak_allocated_bytes": 440737280,
      "peak_reserved_bytes": 448790528,
      "baseline_allocated_bytes": 138487808,
      "baseline_reserved_bytes": 174063616,
      "parameters": 6148,
      "parameter_bytes": 24592,
      "dynamic_control_values_per_token": 2,
      "materialized_state_sized_transition": false
    },
    {
      "shape": {
        "batch": 1,
        "length": 8192,
        "width": 2560
      },
      "architecture": "rglru",
      "backend": "fused_grouped_prefix32_hierarchical_chunk16",
      "controller": null,
      "forward": {
        "samples_ms": [
          1.0851770639419556,
          1.0833553969860077,
          1.0850951075553894,
          1.087583601474762,
          1.083914190530777,
          1.085188239812851,
          1.0785721242427826,
          1.0809451341629028,
          1.0889172554016113,
          1.0823681950569153
        ],
        "median_ms": 1.0845046490430832,
        "mean_ms": 1.0841116309165955,
        "std_ms": 0.002877035692075149,
        "minimum_ms": 1.0785721242427826,
        "maximum_ms": 1.0889172554016113
      },
      "backward": {
        "samples_ms": [
          1.9751079380512238,
          1.9753128290176392,
          1.9731149077415466,
          1.977190375328064,
          1.9784271717071533,
          1.978319138288498,
          1.9810721278190613,
          1.9775480031967163,
          1.9794441759586334,
          1.9757896661758423
        ],
        "median_ms": 1.9773691892623901,
        "mean_ms": 1.9771326333284378,
        "std_ms": 0.0022243896788278427,
        "minimum_ms": 1.9731149077415466,
        "maximum_ms": 1.9810721278190613
      },
      "forward_backward": {
        "samples_ms": [
          3.1344592571258545,
          3.1301677227020264,
          3.127768635749817,
          3.1264349818229675,
          3.1397975981235504,
          3.118518739938736,
          3.1206682324409485,
          3.1388700008392334,
          3.1435824930667877,
          3.1290873885154724
        ],
        "median_ms": 3.1296275556087494,
        "mean_ms": 3.1309355050325394,
        "std_ms": 0.007795753903631978,
        "minimum_ms": 3.118518739938736,
        "maximum_ms": 3.1435824930667877
      },
      "rounds": [
        {
          "round": 0,
          "order_index": 1,
          "forward_ms": [
            1.0851770639419556,
            1.0833553969860077,
            1.0850951075553894,
            1.087583601474762,
            1.083914190530777
          ],
          "backward_ms": [
            1.9751079380512238,
            1.9753128290176392,
            1.9731149077415466,
            1.977190375328064,
            1.9784271717071533
          ],
          "forward_backward_ms": [
            3.1344592571258545,
            3.1301677227020264,
            3.127768635749817,
            3.1264349818229675,
            3.1397975981235504
          ]
        },
        {
          "round": 1,
          "order_index": 0,
          "forward_ms": [
            1.085188239812851,
            1.0785721242427826,
            1.0809451341629028,
            1.0889172554016113,
            1.0823681950569153
          ],
          "backward_ms": [
            1.978319138288498,
            1.9810721278190613,
            1.9775480031967163,
            1.9794441759586334,
            1.9757896661758423
          ],
          "forward_backward_ms": [
            3.118518739938736,
            3.1206682324409485,
            3.1388700008392334,
            3.1435824930667877,
            3.1290873885154724
          ]
        }
      ],
      "peak_allocated_delta_bytes": 484506624,
      "peak_reserved_delta_bytes": 585105408,
      "peak_allocated_bytes": 638874112,
      "peak_reserved_bytes": 771751936,
      "baseline_allocated_bytes": 154367488,
      "baseline_reserved_bytes": 186646528,
      "parameters": 826880,
      "parameter_bytes": 3307520,
      "dynamic_control_values_per_token": 5120,
      "materialized_state_sized_transition": false
    },
    {
      "shape": {
        "batch": 1,
        "length": 8192,
        "width": 2560
      },
      "architecture": "samu",
      "backend": "fused_output_fused_write_shared_sfu_grouped_prefix64_hierarchical_chunk32",
      "controller": "fp32_fused_coords",
      "forward": {
        "samples_ms": [
          0.9420216083526611,
          0.9067356586456299,
          0.8699931204319,
          0.8872412145137787,
          0.9084157645702362,
          0.8951611816883087,
          0.8825547993183136,
          0.8826702833175659,
          0.8766800165176392,
          0.8629001677036285
        ],
        "median_ms": 0.8849557489156723,
        "mean_ms": 0.8914373815059662,
        "std_ms": 0.02181813990421382,
        "minimum_ms": 0.8629001677036285,
        "maximum_ms": 0.9420216083526611
      },
      "backward": {
        "samples_ms": [
          1.2913234531879425,
          1.2894012033939362,
          1.2915432453155518,
          1.3994313776493073,
          1.2904070317745209,
          1.2955181300640106,
          1.2936852872371674,
          1.2956075370311737,
          1.294407993555069,
          1.296989619731903
        ],
        "median_ms": 1.2940466403961182,
        "mean_ms": 1.3038314878940582,
        "std_ms": 0.03195394180916331,
        "minimum_ms": 1.2894012033939362,
        "maximum_ms": 1.3994313776493073
      },
      "forward_backward": {
        "samples_ms": [
          2.2789016366004944,
          2.2646114230155945,
          2.244453877210617,
          2.2643916308879852,
          2.2643283009529114,
          2.277109771966934,
          2.2587627172470093,
          2.296488732099533,
          2.2661015391349792,
          2.252452075481415
        ],
        "median_ms": 2.26450152695179,
        "mean_ms": 2.2667601704597473,
        "std_ms": 0.01386053125494159,
        "minimum_ms": 2.244453877210617,
        "maximum_ms": 2.296488732099533
      },
      "rounds": [
        {
          "round": 0,
          "order_index": 0,
          "forward_ms": [
            0.9420216083526611,
            0.9067356586456299,
            0.8699931204319,
            0.8872412145137787,
            0.9084157645702362
          ],
          "backward_ms": [
            1.2913234531879425,
            1.2894012033939362,
            1.2915432453155518,
            1.3994313776493073,
            1.2904070317745209
          ],
          "forward_backward_ms": [
            2.2789016366004944,
            2.2646114230155945,
            2.244453877210617,
            2.2643916308879852,
            2.2643283009529114
          ]
        },
        {
          "round": 1,
          "order_index": 1,
          "forward_ms": [
            0.8951611816883087,
            0.8825547993183136,
            0.8826702833175659,
            0.8766800165176392,
            0.8629001677036285
          ],
          "backward_ms": [
            1.2955181300640106,
            1.2936852872371674,
            1.2956075370311737,
            1.294407993555069,
            1.296989619731903
          ],
          "forward_backward_ms": [
            2.277109771966934,
            2.2587627172470093,
            2.296488732099533,
            2.2661015391349792,
            2.252452075481415
          ]
        }
      ],
      "peak_allocated_delta_bytes": 377754112,
      "peak_reserved_delta_bytes": 377487360,
      "peak_allocated_bytes": 535396864,
      "peak_reserved_bytes": 564133888,
      "baseline_allocated_bytes": 157642752,
      "baseline_reserved_bytes": 186646528,
      "parameters": 7684,
      "parameter_bytes": 30736,
      "dynamic_control_values_per_token": 2,
      "materialized_state_sized_transition": false
    },
    {
      "shape": {
        "batch": 1,
        "length": 32768,
        "width": 1024
      },
      "architecture": "rglru",
      "backend": "fused_grouped_prefix32_hierarchical_chunk16",
      "controller": null,
      "forward": {
        "samples_ms": [
          1.713864505290985,
          1.703694462776184,
          1.6937106847763062,
          1.6922466456890106,
          1.6987062990665436,
          1.7120614647865295,
          1.700911670923233,
          1.7167478799819946,
          1.6942545771598816,
          1.6917027533054352
        ],
        "median_ms": 1.6998089849948883,
        "mean_ms": 1.7017900943756104,
        "std_ms": 0.0089656487856422,
        "minimum_ms": 1.6917027533054352,
        "maximum_ms": 1.7167478799819946
      },
      "backward": {
        "samples_ms": [
          3.093324601650238,
          3.08971107006073,
          3.083445131778717,
          3.087073564529419,
          3.098614513874054,
          3.0861087143421173,
          3.091622143983841,
          3.091219812631607,
          3.092147409915924,
          3.085799515247345
        ],
        "median_ms": 3.0904654413461685,
        "mean_ms": 3.0899066478013992,
        "std_ms": 0.0042226333146147535,
        "minimum_ms": 3.083445131778717,
        "maximum_ms": 3.098614513874054
      },
      "forward_backward": {
        "samples_ms": [
          4.891674965620041,
          4.84820082783699,
          4.852220416069031,
          4.853308200836182,
          4.835762083530426,
          4.896078258752823,
          4.856150597333908,
          4.842191934585571,
          4.85934317111969,
          4.844155162572861
        ],
        "median_ms": 4.852764308452606,
        "mean_ms": 4.857908561825752,
        "std_ms": 0.019174560276487278,
        "minimum_ms": 4.835762083530426,
        "maximum_ms": 4.896078258752823
      },
      "rounds": [
        {
          "round": 0,
          "order_index": 0,
          "forward_ms": [
            1.713864505290985,
            1.703694462776184,
            1.6937106847763062,
            1.6922466456890106,
            1.6987062990665436
          ],
          "backward_ms": [
            3.093324601650238,
            3.08971107006073,
            3.083445131778717,
            3.087073564529419,
            3.098614513874054
          ],
          "forward_backward_ms": [
            4.891674965620041,
            4.84820082783699,
            4.852220416069031,
            4.853308200836182,
            4.835762083530426
          ]
        },
        {
          "round": 1,
          "order_index": 1,
          "forward_ms": [
            1.7120614647865295,
            1.700911670923233,
            1.7167478799819946,
            1.6942545771598816,
            1.6917027533054352
          ],
          "backward_ms": [
            3.0861087143421173,
            3.091622143983841,
            3.091219812631607,
            3.092147409915924,
            3.085799515247345
          ],
          "forward_backward_ms": [
            4.896078258752823,
            4.856150597333908,
            4.842191934585571,
            4.85934317111969,
            4.844155162572861
          ]
        }
      ],
      "peak_allocated_delta_bytes": 789591552,
      "peak_reserved_delta_bytes": 754974720,
      "peak_allocated_bytes": 991482880,
      "peak_reserved_bytes": 991952896,
      "baseline_allocated_bytes": 201891328,
      "baseline_reserved_bytes": 236978176,
      "parameters": 134144,
      "parameter_bytes": 536576,
      "dynamic_control_values_per_token": 2048,
      "materialized_state_sized_transition": false
    },
    {
      "shape": {
        "batch": 1,
        "length": 32768,
        "width": 1024
      },
      "architecture": "samu",
      "backend": "fused_output_fused_write_shared_sfu_grouped_prefix64_hierarchical_chunk32",
      "controller": "fp32_fused_coords",
      "forward": {
        "samples_ms": [
          0.892098993062973,
          0.9111687541007996,
          0.8835271000862122,
          0.8731558918952942,
          0.8769519627094269,
          0.8863545954227448,
          0.8594878017902374,
          0.8979551494121552,
          0.8840784430503845,
          0.8880086243152618
        ],
        "median_ms": 0.8852165192365646,
        "mean_ms": 0.885278731584549,
        "std_ms": 0.013311986650921108,
        "minimum_ms": 0.8594878017902374,
        "maximum_ms": 0.9111687541007996
      },
      "backward": {
        "samples_ms": [
          2.059612423181534,
          2.0613446831703186,
          2.049162983894348,
          2.0650364458560944,
          2.065274864435196,
          2.051331102848053,
          2.058371901512146,
          2.0644627511501312,
          2.077147364616394,
          2.0544640719890594
        ],
        "median_ms": 2.060478553175926,
        "mean_ms": 2.0606208592653275,
        "std_ms": 0.007688497048614947,
        "minimum_ms": 2.049162983894348,
        "maximum_ms": 2.077147364616394
      },
      "forward_backward": {
        "samples_ms": [
          3.0279532074928284,
          3.063168376684189,
          3.0513517558574677,
          3.023579716682434,
          3.0220039188861847,
          3.0708834528923035,
          3.047708421945572,
          3.0349381268024445,
          3.025650978088379,
          3.049716353416443
        ],
        "median_ms": 3.041323274374008,
        "mean_ms": 3.0416954308748245,
        "std_ms": 0.01647614328462086,
        "minimum_ms": 3.0220039188861847,
        "maximum_ms": 3.0708834528923035
      },
      "rounds": [
        {
          "round": 0,
          "order_index": 1,
          "forward_ms": [
            0.892098993062973,
            0.9111687541007996,
            0.8835271000862122,
            0.8731558918952942,
            0.8769519627094269
          ],
          "backward_ms": [
            2.059612423181534,
            2.0613446831703186,
            2.049162983894348,
            2.0650364458560944,
            2.065274864435196
          ],
          "forward_backward_ms": [
            3.0279532074928284,
            3.063168376684189,
            3.0513517558574677,
            3.023579716682434,
            3.0220039188861847
          ]
        },
        {
          "round": 1,
          "order_index": 0,
          "forward_ms": [
            0.8863545954227448,
            0.8594878017902374,
            0.8979551494121552,
            0.8840784430503845,
            0.8880086243152618
          ],
          "backward_ms": [
            2.051331102848053,
            2.058371901512146,
            2.0644627511501312,
            2.077147364616394,
            2.0544640719890594
          ],
          "forward_backward_ms": [
            3.0708834528923035,
            3.047708421945572,
            3.0349381268024445,
            3.025650978088379,
            3.049716353416443
          ]
        }
      ],
      "peak_allocated_delta_bytes": 604913152,
      "peak_reserved_delta_bytes": 606076928,
      "peak_allocated_bytes": 807327232,
      "peak_reserved_bytes": 843055104,
      "baseline_allocated_bytes": 202414080,
      "baseline_reserved_bytes": 236978176,
      "parameters": 3076,
      "parameter_bytes": 12304,
      "dynamic_control_values_per_token": 2,
      "materialized_state_sized_transition": false
    },
    {
      "shape": {
        "batch": 1,
        "length": 32768,
        "width": 2048
      },
      "architecture": "rglru",
      "backend": "fused_grouped_prefix32_hierarchical_chunk16",
      "controller": null,
      "forward": {
        "samples_ms": [
          2.9669590294361115,
          2.9472261667251587,
          2.951335161924362,
          2.944283187389374,
          2.942904829978943,
          2.958834171295166,
          2.950340509414673,
          2.956252545118332,
          2.9529817402362823,
          2.948623150587082
        ],
        "median_ms": 2.9508378356695175,
        "mean_ms": 2.9519740492105484,
        "std_ms": 0.006849667931695254,
        "minimum_ms": 2.942904829978943,
        "maximum_ms": 2.9669590294361115
      },
      "backward": {
        "samples_ms": [
          5.733184516429901,
          5.73517382144928,
          5.732398480176926,
          5.738086998462677,
          5.725208669900894,
          5.739562213420868,
          5.731556564569473,
          5.7291872799396515,
          5.723584443330765,
          5.725633352994919
        ],
        "median_ms": 5.7319775223731995,
        "mean_ms": 5.731357634067535,
        "std_ms": 0.005170428146545174,
        "minimum_ms": 5.723584443330765,
        "maximum_ms": 5.739562213420868
      },
      "forward_backward": {
        "samples_ms": [
          8.826836943626404,
          8.748695254325867,
          8.77295807003975,
          8.740019053220749,
          8.741144090890884,
          8.782442659139633,
          8.769046515226364,
          8.79686325788498,
          8.755963295698166,
          8.740145713090897
        ],
        "median_ms": 8.762504905462265,
        "mean_ms": 8.76741148531437,
        "std_ms": 0.027061537545544,
        "minimum_ms": 8.740019053220749,
        "maximum_ms": 8.826836943626404
      },
      "rounds": [
        {
          "round": 0,
          "order_index": 1,
          "forward_ms": [
            2.9669590294361115,
            2.9472261667251587,
            2.951335161924362,
            2.944283187389374,
            2.942904829978943
          ],
          "backward_ms": [
            5.733184516429901,
            5.73517382144928,
            5.732398480176926,
            5.738086998462677,
            5.725208669900894
          ],
          "forward_backward_ms": [
            8.826836943626404,
            8.748695254325867,
            8.77295807003975,
            8.740019053220749,
            8.741144090890884
          ]
        },
        {
          "round": 1,
          "order_index": 0,
          "forward_ms": [
            2.958834171295166,
            2.950340509414673,
            2.956252545118332,
            2.9529817402362823,
            2.948623150587082
          ],
          "backward_ms": [
            5.739562213420868,
            5.731556564569473,
            5.7291872799396515,
            5.723584443330765,
            5.725633352994919
          ],
          "forward_backward_ms": [
            8.782442659139633,
            8.769046515226364,
            8.79686325788498,
            8.755963295698166,
            8.740145713090897
          ]
        }
      ],
      "peak_allocated_delta_bytes": 1579705856,
      "peak_reserved_delta_bytes": 1545601024,
      "peak_allocated_bytes": 1917424640,
      "peak_reserved_bytes": 1918894080,
      "baseline_allocated_bytes": 337718784,
      "baseline_reserved_bytes": 373293056,
      "parameters": 530432,
      "parameter_bytes": 2121728,
      "dynamic_control_values_per_token": 4096,
      "materialized_state_sized_transition": false
    },
    {
      "shape": {
        "batch": 1,
        "length": 32768,
        "width": 2048
      },
      "architecture": "samu",
      "backend": "fused_output_fused_write_shared_sfu_grouped_prefix64_hierarchical_chunk32",
      "controller": "fp32_fused_coords",
      "forward": {
        "samples_ms": [
          1.356080174446106,
          1.3549365103244781,
          1.3494640588760376,
          1.4060735702514648,
          1.3238266110420227,
          1.4113448560237885,
          1.3967528939247131,
          1.3872310519218445,
          1.3216659426689148,
          1.3961829245090485
        ],
        "median_ms": 1.3716556131839752,
        "mean_ms": 1.3703558593988419,
        "std_ms": 0.03165148233365389,
        "minimum_ms": 1.3216659426689148,
        "maximum_ms": 1.4113448560237885
      },
      "backward": {
        "samples_ms": [
          3.8497857749462128,
          3.788471221923828,
          3.7913545966148376,
          3.8472898304462433,
          3.8478001952171326,
          3.7849731743335724,
          3.8395263254642487,
          3.801632672548294,
          3.8512349128723145,
          3.8362257182598114
        ],
        "median_ms": 3.83787602186203,
        "mean_ms": 3.8238294422626495,
        "std_ms": 0.026939122342785126,
        "minimum_ms": 3.7849731743335724,
        "maximum_ms": 3.8512349128723145
      },
      "forward_backward": {
        "samples_ms": [
          5.188338458538055,
          5.271311849355698,
          5.2545517683029175,
          5.235884338617325,
          5.257472395896912,
          5.594611167907715,
          5.22671639919281,
          5.2386559545993805,
          5.2258893847465515,
          5.237080156803131
        ],
        "median_ms": 5.237868055701256,
        "mean_ms": 5.2730511873960495,
        "std_ms": 0.10926778573833633,
        "minimum_ms": 5.188338458538055,
        "maximum_ms": 5.594611167907715
      },
      "rounds": [
        {
          "round": 0,
          "order_index": 0,
          "forward_ms": [
            1.356080174446106,
            1.3549365103244781,
            1.3494640588760376,
            1.4060735702514648,
            1.3238266110420227
          ],
          "backward_ms": [
            3.8497857749462128,
            3.788471221923828,
            3.7913545966148376,
            3.8472898304462433,
            3.8478001952171326
          ],
          "forward_backward_ms": [
            5.188338458538055,
            5.271311849355698,
            5.2545517683029175,
            5.235884338617325,
            5.257472395896912
          ]
        },
        {
          "round": 1,
          "order_index": 1,
          "forward_ms": [
            1.4113448560237885,
            1.3967528939247131,
            1.3872310519218445,
            1.3216659426689148,
            1.3961829245090485
          ],
          "backward_ms": [
            3.7849731743335724,
            3.8395263254642487,
            3.801632672548294,
            3.8512349128723145,
            3.8362257182598114
          ],
          "forward_backward_ms": [
            5.594611167907715,
            5.22671639919281,
            5.2386559545993805,
            5.2258893847465515,
            5.237080156803131
          ]
        }
      ],
      "peak_allocated_delta_bytes": 1208907264,
      "peak_reserved_delta_bytes": 1210056704,
      "peak_allocated_bytes": 1548721664,
      "peak_reserved_bytes": 1585446912,
      "baseline_allocated_bytes": 339814400,
      "baseline_reserved_bytes": 375390208,
      "parameters": 6148,
      "parameter_bytes": 24592,
      "dynamic_control_values_per_token": 2,
      "materialized_state_sized_transition": false
    }
  ]
}
