{
  "captured_at": "2026-09-28T10:04:34.745560+00:00",
  "workload_version": "challenge-workload-v0.7",
  "leaderboard_url": "https://linux-slai.tail6d76d1.ts.net:8443/api/transformer/leaderboard",
  "entries": [
    {
      "rank": 1,
      "name": "方案1",
      "score": 86665.61952554528,
      "submission_id": "TRF-6A0A6DAACD214831",
      "created_at": "2026-09-28T01:29:01.267015+00:00",
      "score_source": "verified",
      "hardware": {
        "version": "challenge-hardware-v0.4",
        "sm_count": 16,
        "tc_count": 1,
        "vector_lanes": 16,
        "sfu_lanes": 16,
        "reduction_units": 0,
        "rf_ports": "8R4W",
        "shared_kib": 0,
        "dma_depth": 2,
        "dma_engines": 2,
        "sm_noc_bytes_per_cycle": 64,
        "multicast": true,
        "noc_bytes_per_cycle": 512,
        "hbm_channels": 8,
        "cache_mib": 0,
        "tc_array": "8x16",
        "tc_k_parallel": 2
      },
      "gates": {
        "area_gate_passed": true,
        "area_mm2": 23.28176530432,
        "eligible": true,
        "scenarios": {
          "M1_P1": {
            "cycles": 210399,
            "first_token_or_average_step_seconds": 0.000420798,
            "functional_passed": true,
            "latency_gate_passed": true,
            "peak_power_w": 19.217624187188445,
            "power_gate_passed": true,
            "throughput_ratio_to_baseline": 147.3248209354607,
            "throughput_tokens_per_second": 304183.9552469356
          },
          "M2_D1": {
            "cycles": 25539,
            "first_token_or_average_step_seconds": 6.38475e-06,
            "functional_passed": true,
            "latency_gate_passed": true,
            "peak_power_w": 19.28940560781613,
            "power_gate_passed": true,
            "throughput_ratio_to_baseline": 50.98210579897412,
            "throughput_tokens_per_second": 156623.20372763224
          }
        },
        "score": 86665.61952554528
      },
      "hashes": {
        "zip": "bdbe6f901433fcc2d157e3be899ff3fdecfcff21e19f94eafe9c4f70313c7779",
        "report": "071af5ed4dfe2c176fd202b45006f16ae8e90e619b1c15ea7175d825c8a1196a",
        "hardware": "fd3309e936be2fd2da7127c5d8ee15b875659b00f266dda674ce1d40bdfce64f",
        "canonical_hardware": "ca16ef9acbdde8ddea54f128413f701edf1ea4059d6e5f83a09b5d2576192ede"
      },
      "trace_files": [],
      "programs": {
        "M1_P1": {
          "sha256": "b3a9c6a5b865c90240ad70add4cf443f413ae67d30d3d04f324bfc172106d69a",
          "source_lines": 28242,
          "ops": {
            "WG.BEGIN": 80,
            "VEC": 20905,
            "ST": 1711,
            "LD": 13858,
            "MMA.ACC": 22848,
            "WAIT": 7921,
            "SFU": 780,
            "WG.END": 80,
            "STEP.COMMIT": 1,
            "REDUCE": 120
          },
          "mma_shapes": {
            "(1, 32, 32)": 576,
            "(1, 128, 8)": 96,
            "(16, 32, 32)": 12672,
            "(16, 16, 32)": 96,
            "(16, 16, 1)": 48,
            "(16, 32, 1)": 240,
            "(16, 48, 1)": 48,
            "(16, 64, 1)": 144,
            "(16, 1, 16)": 48,
            "(16, 32, 16)": 96,
            "(16, 1, 32)": 48,
            "(16, 1, 48)": 48,
            "(16, 1, 64)": 48,
            "(16, 16, 64)": 6144,
            "(32, 32, 1)": 96,
            "(2, 16, 64)": 1536,
            "(1, 1, 32)": 48,
            "(1, 32, 1)": 48,
            "(2, 64, 16)": 768
          },
          "groups": {
            "w0": 0,
            "w1": 1,
            "w2": 2,
            "w3": 3,
            "w4": 4,
            "w5": 5,
            "w6": 6,
            "w7": 7,
            "w8": 8,
            "w9": 9,
            "w10": 10,
            "w11": 11,
            "w12": 12,
            "w13": 13,
            "w14": 14,
            "w15": 15,
            "q0": 0,
            "t0": 0,
            "f0": 0,
            "g0": 0,
            "q1": 1,
            "t1": 1,
            "f1": 1,
            "g1": 1,
            "q2": 2,
            "t2": 2,
            "f2": 2,
            "g2": 2,
            "q3": 3,
            "t3": 3,
            "f3": 3,
            "g3": 3,
            "q4": 4,
            "t4": 4,
            "f4": 4,
            "g4": 4,
            "q5": 5,
            "t5": 5,
            "f5": 5,
            "g5": 5,
            "q6": 6,
            "t6": 6,
            "f6": 6,
            "g6": 6,
            "q7": 7,
            "t7": 7,
            "f7": 7,
            "g7": 7,
            "q8": 8,
            "t8": 8,
            "f8": 8,
            "g8": 8,
            "q9": 9,
            "t9": 9,
            "f9": 9,
            "g9": 9,
            "q10": 10,
            "t10": 10,
            "f10": 10,
            "g10": 10,
            "q11": 11,
            "t11": 11,
            "f11": 11,
            "g11": 11,
            "q12": 12,
            "t12": 12,
            "f12": 12,
            "g12": 12,
            "q13": 13,
            "t13": 13,
            "f13": 13,
            "g13": 13,
            "q14": 14,
            "t14": 14,
            "f14": 14,
            "g14": 14,
            "q15": 15,
            "t15": 15,
            "f15": 15,
            "g15": 15
          },
          "segments_by_commit": [
            {
              "ST:scratch": 4376064,
              "LD:input/prompt": 131072,
              "LD:layer0/ln1_g": 2048,
              "LD:layer0/ln1_b": 2048,
              "LD:scratch": 11870208,
              "LD:layer0/wqkv": 1572864,
              "ST:layer0/new_k": 131072,
              "LD:layer0/wo": 524288,
              "ST:layer0/new_v": 131072,
              "LD:layer0/ln2_g": 2048,
              "LD:layer0/ln2_b": 2048,
              "LD:layer0/b1": 8192,
              "LD:layer0/w1": 2097152,
              "LD:layer0/w2": 2097152,
              "LD:layer0/b2": 2048,
              "LD:layer1/ln1_g": 2048,
              "LD:layer1/ln1_b": 2048,
              "LD:layer1/wqkv": 1572864,
              "ST:layer1/new_k": 131072,
              "LD:layer1/wo": 524288,
              "ST:layer1/new_v": 131072,
              "LD:layer1/ln2_g": 2048,
              "LD:layer1/ln2_b": 2048,
              "LD:layer1/b1": 8192,
              "LD:layer1/w1": 2097152,
              "LD:layer1/w2": 2097152,
              "LD:layer1/b2": 2048,
              "LD:layer2/ln1_g": 2048,
              "LD:layer2/ln1_b": 2048,
              "LD:layer2/wqkv": 1572864,
              "ST:layer2/new_k": 131072,
              "LD:layer2/wo": 524288,
              "ST:layer2/new_v": 131072,
              "LD:layer2/ln2_g": 2048,
              "LD:layer2/ln2_b": 2048,
              "LD:layer2/b1": 8192,
              "LD:layer2/w1": 2097152,
              "LD:layer2/w2": 2097152,
              "LD:layer2/b2": 2048,
              "ST:output/hidden": 131072
            },
            {
              "LD:layer0/new_k": 131072,
              "LD:layer0/b1": 4096,
              "LD:input/step0": 4096,
              "LD:layer0/ln1_g": 1024,
              "LD:layer0/ln1_b": 1024,
              "LD:layer0/b2": 1024,
              "LD:layer0/new_v": 131072,
              "ST:scratch": 73728,
              "LD:scratch": 729088,
              "LD:layer0/w1": 1048576,
              "LD:layer0/wqkv": 786432,
              "LD:layer0/wo": 262144,
              "ST:layer0/new_k": 2048,
              "ST:layer0/new_v": 2048,
              "LD:layer0/w2": 1048576,
              "LD:layer0/ln2_g": 1024,
              "LD:layer0/ln2_b": 1024,
              "LD:layer1/new_k": 131072,
              "LD:layer1/b1": 4096,
              "LD:layer1/new_v": 131072,
              "LD:layer1/w1": 1048576,
              "LD:layer1/b2": 1024,
              "LD:layer1/ln1_g": 1024,
              "LD:layer1/ln1_b": 1024,
              "LD:layer1/wqkv": 786432,
              "LD:layer1/wo": 262144,
              "ST:layer1/new_k": 2048,
              "ST:layer1/new_v": 2048,
              "LD:layer1/w2": 1048576,
              "LD:layer1/ln2_g": 1024,
              "LD:layer1/ln2_b": 1024,
              "LD:layer2/new_k": 131072,
              "LD:layer2/b1": 4096,
              "LD:layer2/new_v": 131072,
              "LD:layer2/w1": 1048576,
              "LD:layer2/b2": 1024,
              "LD:layer2/ln1_g": 1024,
              "LD:layer2/ln1_b": 1024,
              "LD:layer2/wqkv": 786432,
              "LD:layer2/wo": 262144,
              "ST:layer2/new_k": 2048,
              "ST:layer2/new_v": 2048,
              "LD:layer2/w2": 1048576,
              "LD:layer2/ln2_g": 1024,
              "LD:layer2/ln2_b": 1024,
              "ST:output/hidden": 2048
            }
          ],
          "timing": {
            "area_mm2": 23.28176530432,
            "average_power_w": 12.908411814431396,
            "cache_hits": 0,
            "cache_misses": 0,
            "cycles": 210399,
            "dynamic_energy_pj": 5186910867.775921,
            "hbm_read_bytes": 25334528,
            "hbm_write_bytes": 5381632,
            "instruction_count": 68304,
            "peak_power_upper_bound_w": 23.265158106264018,
            "peak_window_power_w": 19.217624187188445,
            "total_energy_pj": 5431833874.689102
          },
          "mean_sm_utilization": {
            "tc": 0.4117034776781259,
            "rf_read": 0.7324868464203728,
            "rf_write": 0.45730344488329316,
            "vec": 0.10364622693073636,
            "sfu": 0.0400025665521224
          }
        },
        "M2_D1": {
          "sha256": "64290db4628b0f2024cef4db9d25f55810306d4f8b45562a92bfbee8297a33d2",
          "source_lines": 27300,
          "ops": {
            "WG.BEGIN": 62,
            "LD": 2840,
            "VEC": 9542,
            "MMA.ACC": 6656,
            "ST": 3264,
            "WAIT": 2370,
            "REDUCE": 1472,
            "SFU": 1024,
            "STEP.COMMIT": 8,
            "WG.END": 62
          },
          "mma_shapes": {
            "(1, 2, 16)": 256,
            "(1, 16, 64)": 2240,
            "(16, 16, 16)": 192,
            "(1, 16, 32)": 3584,
            "(1, 1, 32)": 8,
            "(1, 32, 1)": 8,
            "(1, 1, 4)": 64,
            "(1, 32, 5)": 72,
            "(1, 16, 17)": 128,
            "(1, 2, 32)": 8,
            "(1, 32, 2)": 8,
            "(1, 3, 32)": 8,
            "(1, 32, 3)": 8,
            "(1, 4, 32)": 8,
            "(1, 32, 4)": 8,
            "(1, 5, 32)": 8,
            "(1, 6, 32)": 8,
            "(1, 32, 6)": 8,
            "(1, 7, 32)": 8,
            "(1, 32, 7)": 8,
            "(1, 8, 32)": 8,
            "(1, 32, 8)": 8
          },
          "groups": {
            "a0_0": 0,
            "a0_1": 1,
            "a0_2": 2,
            "a0_4": 4,
            "a0_5": 5,
            "a0_6": 6,
            "a0_8": 8,
            "a0_9": 9,
            "a0_10": 10,
            "a0_12": 12,
            "a0_13": 13,
            "a0_14": 14,
            "f0_0": 0,
            "f0_1": 1,
            "f0_2": 2,
            "f0_3": 3,
            "f0_4": 4,
            "f0_5": 5,
            "f0_6": 6,
            "f0_7": 7,
            "f0_8": 8,
            "f0_9": 9,
            "f0_10": 10,
            "f0_11": 11,
            "f0_12": 12,
            "f0_13": 13,
            "f0_14": 14,
            "f0_15": 15,
            "a1_0": 0,
            "a1_1": 1,
            "a1_2": 2,
            "a1_4": 4,
            "a1_5": 5,
            "a1_6": 6,
            "a1_8": 8,
            "a1_9": 9,
            "a1_10": 10,
            "a1_12": 12,
            "a1_13": 13,
            "a1_14": 14,
            "f1_0": 0,
            "f1_1": 1,
            "f1_2": 2,
            "f1_3": 3,
            "f1_4": 4,
            "f1_5": 5,
            "f1_6": 6,
            "f1_7": 7,
            "f1_8": 8,
            "f1_9": 9,
            "f1_10": 10,
            "f1_11": 11,
            "f1_12": 12,
            "f1_13": 13,
            "f1_14": 14,
            "f1_15": 15,
            "p_3": 3,
            "p_7": 7,
            "p_11": 11,
            "p_15": 15,
            "pace0": 11,
            "pace1": 15
          },
          "segments_by_commit": [
            {
              "LD:layer0/ln1_g": 6656,
              "LD:layer0/ln1_b": 6144,
              "LD:layer0/ln2_g": 8192,
              "LD:layer0/ln2_b": 8192,
              "LD:layer0/b1": 2048,
              "LD:layer0/b2": 512,
              "LD:layer1/ln1_g": 6656,
              "LD:layer1/ln2_g": 8704,
              "LD:layer1/ln1_b": 6144,
              "LD:layer1/ln2_b": 8192,
              "LD:layer1/b1": 2048,
              "LD:layer1/b2": 512,
              "LD:input/step0": 6656,
              "ST:scratch": 24000,
              "LD:layer0/wqkv": 196608,
              "LD:layer0/history_k": 65536,
              "LD:layer0/history_v": 65536,
              "LD:layer0/w1": 262144,
              "LD:layer0/wo": 65536,
              "LD:layer0/w2": 262144,
              "LD:scratch": 57536,
              "ST:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "LD:layer0/new_k": 512,
              "LD:layer0/new_v": 512,
              "LD:layer1/wqkv": 196608,
              "LD:layer1/history_k": 65536,
              "LD:layer1/history_v": 65536,
              "LD:layer1/wo": 65536,
              "LD:layer1/w1": 262144,
              "LD:layer1/w2": 262144,
              "ST:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "LD:layer1/new_k": 512,
              "LD:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step1": 6656,
              "ST:scratch": 24000,
              "LD:scratch": 57536,
              "ST:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "LD:layer0/new_k": 512,
              "LD:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "LD:layer1/new_k": 512,
              "LD:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step2": 6656,
              "ST:scratch": 24000,
              "LD:scratch": 57536,
              "ST:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "LD:layer0/new_k": 512,
              "LD:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "LD:layer1/new_k": 512,
              "LD:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step3": 6656,
              "ST:scratch": 24000,
              "LD:scratch": 57536,
              "ST:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "LD:layer0/new_k": 512,
              "LD:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "LD:layer1/new_k": 512,
              "LD:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step4": 6656,
              "ST:scratch": 24000,
              "LD:scratch": 57536,
              "ST:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "LD:layer0/new_k": 512,
              "LD:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "LD:layer1/new_k": 512,
              "LD:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step5": 6656,
              "ST:scratch": 24000,
              "LD:scratch": 57536,
              "ST:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "LD:layer0/new_k": 512,
              "LD:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "LD:layer1/new_k": 512,
              "LD:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step6": 6656,
              "ST:scratch": 24000,
              "LD:scratch": 57536,
              "ST:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "LD:layer0/new_k": 512,
              "LD:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "LD:layer1/new_k": 512,
              "LD:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step7": 6656,
              "ST:scratch": 24000,
              "LD:scratch": 57536,
              "ST:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "LD:layer0/new_k": 512,
              "LD:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "LD:layer1/new_k": 512,
              "LD:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {}
          ],
          "timing": {
            "area_mm2": 23.28176530432,
            "average_power_w": 8.408793397341354,
            "cache_hits": 0,
            "cache_misses": 0,
            "cycles": 25539,
            "dynamic_energy_pj": 399774698.9440502,
            "hbm_read_bytes": 2106880,
            "hbm_write_bytes": 212480,
            "instruction_count": 27300,
            "peak_power_upper_bound_w": 22.591079360678524,
            "peak_window_power_w": 19.28940560781613,
            "total_energy_pj": 429504349.1494016
          },
          "mean_sm_utilization": {
            "tc": 0.3845099651513372,
            "rf_read": 0.4082579584165394,
            "rf_write": 0.16309859430674653,
            "vec": 0.1115010376287247,
            "sfu": 0.020674262892047456
          }
        }
      }
    },
    {
      "rank": 2,
      "name": "方案2",
      "score": 73915.76594187708,
      "submission_id": "TRF-E0690E958CB24B05",
      "created_at": "2026-09-27T14:27:09.789696+00:00",
      "score_source": "verified",
      "hardware": {
        "version": "challenge-hardware-v0.4",
        "sm_count": 16,
        "tc_count": 1,
        "tc_array": "8x16",
        "tc_k_parallel": 2,
        "vector_lanes": 16,
        "sfu_lanes": 14,
        "reduction_units": 0,
        "rf_ports": "8R4W",
        "shared_kib": 0,
        "dma_depth": 1,
        "dma_engines": 4,
        "sm_noc_bytes_per_cycle": 128,
        "multicast": true,
        "noc_bytes_per_cycle": 512,
        "hbm_channels": 8,
        "cache_mib": 0
      },
      "gates": {
        "area_gate_passed": true,
        "area_mm2": 23.99856530432,
        "eligible": true,
        "scenarios": {
          "M1_P1": {
            "cycles": 282431,
            "first_token_or_average_step_seconds": 0.000564862,
            "functional_passed": true,
            "latency_gate_passed": true,
            "peak_power_w": 19.440274715137782,
            "power_gate_passed": true,
            "throughput_ratio_to_baseline": 109.75068246757615,
            "throughput_tokens_per_second": 226604.02009694403
          },
          "M2_D1": {
            "cycles": 26155,
            "first_token_or_average_step_seconds": 6.53875e-06,
            "functional_passed": true,
            "latency_gate_passed": true,
            "peak_power_w": 19.591432724606783,
            "power_gate_passed": true,
            "throughput_ratio_to_baseline": 49.78138023322501,
            "throughput_tokens_per_second": 152934.42936341043
          }
        },
        "score": 73915.76594187708
      },
      "hashes": {
        "zip": "ae794be6f9dbe4a5a10fa9a21dc02d7dbe2b2bd8017926dfac4d5e363298cb37",
        "report": "cf871f9757f817093aeb73e1904f52f48abd18dce56421968c8a4786b5fa342d",
        "hardware": "f5f2a5f64b625ca83b678e852adbafaf812cbbf29e610b1f5d7d29ca5ed7c881",
        "canonical_hardware": "97b046273b8d21f31c313d8f478b942c0a3762aeec611f2cb5d7e74ab50f5d78"
      },
      "trace_files": [],
      "programs": {
        "M1_P1": {
          "sha256": "d6ef98892eb9ada7dbd9105ee7032b321410c6d534b78ce530a7c33843cf4033",
          "source_lines": 10161,
          "ops": {
            "WG.BEGIN": 64,
            "VEC": 35536,
            "LD": 16640,
            "MMA.ACC": 24048,
            "ST": 1376,
            "WAIT": 9032,
            "SFU": 40368,
            "REDUCE": 3552,
            "STEP.COMMIT": 1,
            "WG.END": 64
          },
          "mma_shapes": {
            "(64, 1, 16)": 384,
            "(64, 16, 16)": 4608,
            "(16, 16, 16)": 576,
            "(16, 16, 1)": 48,
            "(16, 1, 16)": 48,
            "(16, 32, 16)": 96,
            "(16, 32, 1)": 48,
            "(16, 1, 32)": 48,
            "(16, 16, 32)": 96,
            "(16, 48, 16)": 96,
            "(16, 48, 1)": 48,
            "(16, 1, 48)": 48,
            "(16, 16, 48)": 96,
            "(16, 64, 16)": 96,
            "(16, 64, 1)": 48,
            "(16, 1, 64)": 48,
            "(16, 16, 64)": 96,
            "(32, 16, 32)": 13824,
            "(64, 16, 1)": 480,
            "(4, 48, 16)": 768,
            "(1, 32, 32)": 192,
            "(1, 1, 32)": 48,
            "(1, 32, 1)": 48,
            "(2, 16, 64)": 528,
            "(4, 64, 16)": 768,
            "(2, 16, 32)": 864
          },
          "groups": {
            "p00": 0,
            "x000": 0,
            "x100": 0,
            "x200": 0,
            "p01": 1,
            "x001": 1,
            "x101": 1,
            "x201": 1,
            "p02": 2,
            "x002": 2,
            "x102": 2,
            "x202": 2,
            "p03": 3,
            "x003": 3,
            "x103": 3,
            "x203": 3,
            "p04": 4,
            "x004": 4,
            "x104": 4,
            "x204": 4,
            "p05": 5,
            "x005": 5,
            "x105": 5,
            "x205": 5,
            "p06": 6,
            "x006": 6,
            "x106": 6,
            "x206": 6,
            "p07": 7,
            "x007": 7,
            "x107": 7,
            "x207": 7,
            "p10": 8,
            "x010": 8,
            "x110": 8,
            "x210": 8,
            "p11": 9,
            "x011": 9,
            "x111": 9,
            "x211": 9,
            "p12": 10,
            "x012": 10,
            "x112": 10,
            "x212": 10,
            "p13": 11,
            "x013": 11,
            "x113": 11,
            "x213": 11,
            "p14": 12,
            "x014": 12,
            "x114": 12,
            "x214": 12,
            "p15": 13,
            "x015": 13,
            "x115": 13,
            "x215": 13,
            "p16": 14,
            "x016": 14,
            "x116": 14,
            "x216": 14,
            "p17": 15,
            "x017": 15,
            "x117": 15,
            "x217": 15
          },
          "segments_by_commit": [
            {
              "LD:layer0/ln1_g": 2048,
              "LD:layer0/ln1_b": 2048,
              "LD:layer0/ln2_g": 2048,
              "LD:layer0/ln2_b": 2048,
              "LD:layer0/b1": 8192,
              "LD:layer0/b2": 2048,
              "LD:layer1/ln1_g": 2048,
              "LD:layer1/ln1_b": 2048,
              "LD:layer1/ln2_g": 2048,
              "LD:layer1/ln2_b": 2048,
              "LD:layer1/b1": 8192,
              "LD:layer1/b2": 2048,
              "LD:layer2/ln1_g": 2048,
              "LD:layer2/ln1_b": 2048,
              "LD:layer2/ln2_g": 2048,
              "LD:layer2/ln2_b": 2048,
              "LD:layer2/b1": 8192,
              "LD:layer2/b2": 2048,
              "LD:input/prompt": 131072,
              "ST:scratch": 2801664,
              "LD:scratch": 25559040,
              "LD:layer0/w2": 2686976,
              "LD:layer0/wqkv": 1572864,
              "ST:layer0/new_k": 131072,
              "ST:layer0/new_v": 131072,
              "LD:layer0/wo": 524288,
              "LD:layer0/new_k": 131072,
              "LD:layer0/new_v": 131072,
              "LD:layer0/w1": 2097152,
              "LD:layer1/w2": 2686976,
              "LD:layer1/wqkv": 1572864,
              "ST:layer1/new_k": 131072,
              "ST:layer1/new_v": 131072,
              "LD:layer1/wo": 524288,
              "LD:layer1/new_k": 131072,
              "LD:layer1/new_v": 131072,
              "LD:layer1/w1": 2097152,
              "LD:layer2/w2": 2686976,
              "LD:layer2/wqkv": 1572864,
              "ST:layer2/new_k": 131072,
              "ST:layer2/new_v": 131072,
              "LD:layer2/wo": 524288,
              "LD:layer2/new_k": 131072,
              "LD:layer2/new_v": 131072,
              "LD:layer2/w1": 2097152,
              "ST:output/hidden": 131072
            },
            {
              "LD:layer0/ln1_g": 16384,
              "LD:layer0/ln1_b": 16384,
              "LD:layer0/ln2_g": 16384,
              "LD:layer0/ln2_b": 16384,
              "LD:layer1/ln1_g": 16384,
              "LD:layer1/ln1_b": 16384,
              "LD:layer1/ln2_g": 16384,
              "LD:layer1/ln2_b": 16384,
              "LD:layer2/ln1_g": 16384,
              "LD:layer2/ln1_b": 16384,
              "LD:layer2/ln2_g": 16384,
              "LD:layer2/ln2_b": 16384,
              "LD:input/step0": 32768,
              "LD:layer0/wqkv": 786432,
              "LD:layer0/wo": 262144,
              "LD:layer0/w1": 1048576,
              "ST:scratch": 65536,
              "LD:scratch": 679936,
              "ST:layer0/new_k": 2048,
              "ST:layer0/new_v": 2048,
              "LD:layer0/b1": 4096,
              "LD:layer0/w2": 458752,
              "LD:layer0/b2": 1024,
              "LD:layer1/wqkv": 786432,
              "LD:layer1/wo": 262144,
              "LD:layer1/w1": 1048576,
              "ST:layer1/new_k": 2048,
              "ST:layer1/new_v": 2048,
              "LD:layer1/b1": 4096,
              "LD:layer1/w2": 458752,
              "LD:layer1/b2": 1024,
              "LD:layer2/wqkv": 786432,
              "LD:layer2/wo": 262144,
              "LD:layer2/w1": 1048576,
              "ST:layer2/new_k": 2048,
              "ST:layer2/new_v": 2048,
              "LD:layer2/b1": 4096,
              "LD:layer2/w2": 458752,
              "LD:layer2/b2": 1024,
              "ST:output/hidden": 2048
            }
          ],
          "timing": {
            "area_mm2": 23.99856530432,
            "average_power_w": 11.691842620608055,
            "cache_hits": 0,
            "cache_misses": 0,
            "cycles": 282431,
            "dynamic_energy_pj": 6265380666.4886875,
            "hbm_read_bytes": 33855488,
            "hbm_write_bytes": 3799040,
            "instruction_count": 130681,
            "peak_power_upper_bound_w": 22.922377452672244,
            "peak_window_power_w": 19.440274715137782,
            "total_energy_pj": 6604277606.361908
          },
          "mean_sm_utilization": {
            "tc": 0.31123708091533864,
            "rf_read": 0.5872655622081145,
            "rf_write": 0.4585651008564924,
            "vec": 0.09004323179820913,
            "sfu": 0.08699469959034242
          }
        },
        "M2_D1": {
          "sha256": "b755b364bc7188d8f45edf33777ce28619b7d793c455adcf0a87a95c28c54034",
          "source_lines": 14790,
          "ops": {
            "WG.BEGIN": 64,
            "LD": 5536,
            "VEC": 21984,
            "WAIT": 1932,
            "MMA.ACC": 6592,
            "REDUCE": 1664,
            "SFU": 1088,
            "ST": 3200,
            "STEP.COMMIT": 8,
            "WG.END": 64
          },
          "mma_shapes": {
            "(2, 24, 32)": 128,
            "(2, 32, 32)": 128,
            "(1, 24, 32)": 1024,
            "(1, 16, 32)": 3072,
            "(1, 1, 32)": 8,
            "(1, 32, 1)": 8,
            "(1, 32, 4)": 72,
            "(1, 8, 16)": 256,
            "(1, 8, 56)": 512,
            "(1, 32, 32)": 1024,
            "(1, 8, 18)": 256,
            "(1, 2, 32)": 8,
            "(1, 32, 2)": 8,
            "(1, 3, 32)": 8,
            "(1, 32, 3)": 8,
            "(1, 4, 32)": 8,
            "(1, 5, 32)": 8,
            "(1, 32, 5)": 8,
            "(1, 6, 32)": 8,
            "(1, 32, 6)": 8,
            "(1, 7, 32)": 8,
            "(1, 32, 7)": 8,
            "(1, 8, 32)": 8,
            "(1, 32, 8)": 8
          },
          "groups": {
            "g00": 0,
            "g01": 1,
            "g02": 2,
            "g03": 3,
            "g04": 4,
            "g05": 5,
            "g06": 6,
            "g07": 7,
            "g08": 8,
            "g09": 9,
            "g010": 10,
            "g011": 11,
            "g012": 12,
            "g013": 13,
            "g014": 14,
            "g015": 15,
            "a000": 0,
            "a010": 4,
            "a020": 8,
            "a030": 12,
            "a001": 1,
            "a011": 5,
            "a021": 9,
            "a031": 13,
            "a002": 2,
            "a012": 6,
            "a022": 10,
            "a032": 14,
            "a003": 3,
            "a013": 7,
            "a023": 11,
            "a033": 15,
            "g10": 0,
            "g11": 1,
            "g12": 2,
            "g13": 3,
            "g14": 4,
            "g15": 5,
            "g16": 6,
            "g17": 7,
            "g18": 8,
            "g19": 9,
            "g110": 10,
            "g111": 11,
            "g112": 12,
            "g113": 13,
            "g114": 14,
            "g115": 15,
            "a100": 0,
            "a110": 4,
            "a120": 8,
            "a130": 12,
            "a101": 1,
            "a111": 5,
            "a121": 9,
            "a131": 13,
            "a102": 2,
            "a112": 6,
            "a122": 10,
            "a132": 14,
            "a103": 3,
            "a113": 7,
            "a123": 11,
            "a133": 15
          },
          "segments_by_commit": [
            {
              "LD:layer0/ln1_g": 8192,
              "LD:layer0/ln1_b": 8192,
              "LD:layer0/ln2_g": 8192,
              "LD:layer0/ln2_b": 8192,
              "LD:layer0/b1": 2048,
              "LD:layer0/b2": 512,
              "LD:layer0/wqkv": 196608,
              "LD:layer0/history_k": 65536,
              "LD:layer0/history_v": 65536,
              "LD:layer0/wo": 65536,
              "LD:layer0/w1": 262144,
              "LD:layer0/w2": 262144,
              "LD:layer1/ln1_g": 8192,
              "LD:layer1/ln1_b": 8192,
              "LD:layer1/ln2_g": 8192,
              "LD:layer1/ln2_b": 8192,
              "LD:layer1/b1": 2048,
              "LD:layer1/b2": 512,
              "LD:layer1/wqkv": 196608,
              "LD:layer1/history_k": 65536,
              "LD:layer1/history_v": 65536,
              "LD:layer1/wo": 65536,
              "LD:layer1/w1": 262144,
              "LD:layer1/w2": 262144,
              "LD:input/step0": 8192,
              "ST:scratch": 25280,
              "LD:scratch": 66752,
              "ST:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step1": 8192,
              "ST:scratch": 25280,
              "LD:scratch": 66752,
              "ST:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step2": 8192,
              "ST:scratch": 25280,
              "LD:scratch": 66752,
              "ST:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step3": 8192,
              "ST:scratch": 25280,
              "LD:scratch": 66752,
              "ST:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step4": 8192,
              "ST:scratch": 25280,
              "LD:scratch": 66752,
              "ST:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step5": 8192,
              "ST:scratch": 25280,
              "LD:scratch": 66752,
              "ST:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step6": 8192,
              "ST:scratch": 25280,
              "LD:scratch": 66752,
              "ST:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step7": 8192,
              "ST:scratch": 25280,
              "LD:scratch": 66752,
              "ST:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {}
          ],
          "timing": {
            "area_mm2": 23.99856530432,
            "average_power_w": 8.404782656029676,
            "cache_hits": 0,
            "cache_misses": 0,
            "cycles": 26155,
            "dynamic_energy_pj": 408270056.9601879,
            "hbm_read_bytes": 2127936,
            "hbm_write_bytes": 222720,
            "instruction_count": 42132,
            "peak_power_upper_bound_w": 22.58370383008622,
            "peak_window_power_w": 19.591432724606783,
            "total_energy_pj": 439654180.73691237
          },
          "mean_sm_utilization": {
            "tc": 0.389371057159243,
            "rf_read": 0.4281399350028675,
            "rf_write": 0.2287134391129803,
            "vec": 0.2642706939399732,
            "sfu": 0.025387115274326133
          }
        }
      }
    },
    {
      "rank": 3,
      "name": "方案3",
      "score": 72237.04865290383,
      "submission_id": "TRF-1A499DC4E9184E6E",
      "created_at": "2026-09-27T21:45:34.201233+00:00",
      "score_source": "verified",
      "hardware": {
        "version": "challenge-hardware-v0.4",
        "sm_count": 16,
        "tc_count": 1,
        "vector_lanes": 32,
        "sfu_lanes": 18,
        "reduction_units": 0,
        "rf_ports": "8R4W",
        "shared_kib": 0,
        "dma_depth": 2,
        "dma_engines": 2,
        "sm_noc_bytes_per_cycle": 128,
        "multicast": true,
        "noc_bytes_per_cycle": 512,
        "hbm_channels": 8,
        "cache_mib": 0,
        "tc_array": "8x16",
        "tc_k_parallel": 2
      },
      "gates": {
        "area_gate_passed": true,
        "area_mm2": 23.99856530432,
        "eligible": true,
        "scenarios": {
          "M1_P1": {
            "cycles": 254686,
            "first_token_or_average_step_seconds": 0.000509372,
            "functional_passed": true,
            "latency_gate_passed": true,
            "peak_power_w": 19.69401699661881,
            "power_gate_passed": true,
            "throughput_ratio_to_baseline": 121.70670943828871,
            "throughput_tokens_per_second": 251289.82354742702
          },
          "M2_D1": {
            "cycles": 30368,
            "first_token_or_average_step_seconds": 7.592e-06,
            "functional_passed": true,
            "latency_gate_passed": true,
            "peak_power_w": 18.8882096366068,
            "power_gate_passed": true,
            "throughput_ratio_to_baseline": 42.87513171759747,
            "throughput_tokens_per_second": 131717.59747102213
          }
        },
        "score": 72237.04865290383
      },
      "hashes": {
        "zip": "bd51cf8a694c4218ecd3a4b8d390ad8cd4af54e603409c6754f7ec125e5a2d2f",
        "report": "fddc5abfafda4b254690450bba883532d299b8d9f8d195c092d0f932236ec26f",
        "hardware": "06c4f6410f5c3334303912ffa06e2d6e10a3c949ca9d9943dbd5f580f87619c4",
        "canonical_hardware": "22cb8a8e5a1515a7e89eb53386a5c16b21c328308f37ed6a6dff164027d0de98"
      },
      "trace_files": [
        "agent-trace/rollout-2026-09-27T00-59-52-01a0dea8-b101-7561-be15-573f071b1903.jsonl",
        "agent-trace/rollout-2026-09-27T01-32-17-01a0dec6-6029-7b70-812f-2a3277fb7bc3.jsonl",
        "agent-trace/rollout-2026-09-27T01-38-51-01a0decc-630c-79f2-8461-9ef2bc4d04fc.jsonl",
        "agent-trace/rollout-2026-09-27T09-40-29-01a0decc-630c-79f2-8461-9ef2bc4d04fc_01a0e085-5300-71d2-8c7f-82212d776789.jsonl",
        "agent-trace/rollout-2026-09-27T12-05-14-01a0e109-d9aa-7150-95b1-dd8d251f3f94.jsonl"
      ],
      "programs": {
        "M1_P1": {
          "sha256": "a906784a06d61fce050fc4483e26e1c296ab641f94c1369d310d06a5c26e4f4f",
          "source_lines": 19269,
          "ops": {
            "WG.BEGIN": 17,
            "BARRIER": 216,
            "LD": 28486,
            "REDUCE": 7800,
            "VEC": 20388,
            "SFU": 4044,
            "ST": 5122,
            "MMA.ACC": 45264,
            "STEP.COMMIT": 1,
            "WG.END": 17
          },
          "mma_shapes": {
            "(16, 16, 32)": 37104,
            "(16, 16, 16)": 96,
            "(8, 16, 32)": 1248,
            "(8, 16, 16)": 192,
            "(32, 16, 1)": 768,
            "(16, 16, 1)": 384,
            "(2, 16, 32)": 4608,
            "(1, 8, 32)": 384,
            "(1, 1, 32)": 48,
            "(1, 32, 8)": 384,
            "(1, 32, 1)": 48
          },
          "groups": {
            "c": 0,
            "w0": 0,
            "w1": 1,
            "w2": 2,
            "w3": 3,
            "w4": 4,
            "w5": 5,
            "w6": 6,
            "w7": 7,
            "w8": 8,
            "w9": 9,
            "w10": 10,
            "w11": 11,
            "w12": 12,
            "w13": 13,
            "w14": 14,
            "w15": 15
          },
          "segments_by_commit": [
            {
              "LD:layer0/ln1_g": 16384,
              "LD:layer0/ln1_b": 16384,
              "LD:input/prompt": 262144,
              "ST:scratch": 3932160,
              "LD:scratch": 23855104,
              "LD:layer0/wqkv": 1572864,
              "ST:layer0/new_k": 131072,
              "ST:layer0/new_v": 131072,
              "LD:layer0/new_k": 229376,
              "LD:layer0/new_v": 229376,
              "LD:layer0/wo": 524288,
              "LD:layer0/ln2_g": 16384,
              "LD:layer0/ln2_b": 16384,
              "LD:layer0/w1": 2097152,
              "LD:layer0/b1": 8192,
              "LD:layer0/w2": 2097152,
              "LD:layer0/b2": 8192,
              "LD:layer1/ln1_g": 16384,
              "LD:layer1/ln1_b": 16384,
              "LD:layer1/wqkv": 1572864,
              "ST:layer1/new_k": 131072,
              "ST:layer1/new_v": 131072,
              "LD:layer1/new_k": 229376,
              "LD:layer1/new_v": 229376,
              "LD:layer1/wo": 524288,
              "LD:layer1/ln2_g": 16384,
              "LD:layer1/ln2_b": 16384,
              "LD:layer1/w1": 2097152,
              "LD:layer1/b1": 8192,
              "LD:layer1/w2": 2097152,
              "LD:layer1/b2": 8192,
              "LD:layer2/ln1_g": 16384,
              "LD:layer2/ln1_b": 16384,
              "LD:layer2/wqkv": 1572864,
              "ST:layer2/new_k": 131072,
              "ST:layer2/new_v": 131072,
              "LD:layer2/new_k": 229376,
              "LD:layer2/new_v": 229376,
              "LD:layer2/wo": 524288,
              "LD:layer2/ln2_g": 16384,
              "LD:layer2/ln2_b": 16384,
              "LD:layer2/w1": 2097152,
              "LD:layer2/b1": 8192,
              "LD:layer2/w2": 2097152,
              "LD:layer2/b2": 8192,
              "ST:output/hidden": 131072
            },
            {
              "LD:layer0/ln1_g": 2048,
              "LD:layer0/ln1_b": 2048,
              "LD:input/step0": 4096,
              "ST:scratch": 61440,
              "LD:scratch": 716800,
              "LD:layer0/wqkv": 786432,
              "ST:layer0/new_k": 2048,
              "ST:layer0/new_v": 2048,
              "LD:layer0/new_k": 133120,
              "LD:layer0/new_v": 133120,
              "LD:layer0/wo": 262144,
              "LD:layer0/ln2_g": 2048,
              "LD:layer0/ln2_b": 2048,
              "LD:layer0/w1": 1048576,
              "LD:layer0/b1": 4096,
              "LD:layer0/w2": 1048576,
              "LD:layer0/b2": 1024,
              "LD:layer1/ln1_g": 2048,
              "LD:layer1/ln1_b": 2048,
              "LD:layer1/wqkv": 786432,
              "ST:layer1/new_k": 2048,
              "ST:layer1/new_v": 2048,
              "LD:layer1/new_k": 133120,
              "LD:layer1/new_v": 133120,
              "LD:layer1/wo": 262144,
              "LD:layer1/ln2_g": 2048,
              "LD:layer1/ln2_b": 2048,
              "LD:layer1/w1": 1048576,
              "LD:layer1/b1": 4096,
              "LD:layer1/w2": 1048576,
              "LD:layer1/b2": 1024,
              "LD:layer2/ln1_g": 2048,
              "LD:layer2/ln1_b": 2048,
              "LD:layer2/wqkv": 786432,
              "ST:layer2/new_k": 2048,
              "ST:layer2/new_v": 2048,
              "LD:layer2/new_k": 133120,
              "LD:layer2/new_v": 133120,
              "LD:layer2/wo": 262144,
              "LD:layer2/ln2_g": 2048,
              "LD:layer2/ln2_b": 2048,
              "LD:layer2/w1": 1048576,
              "LD:layer2/b1": 4096,
              "LD:layer2/w2": 1048576,
              "LD:layer2/b2": 1024,
              "ST:output/hidden": 2048
            }
          ],
          "timing": {
            "area_mm2": 23.99856530432,
            "average_power_w": 10.757594113425936,
            "cache_hits": 0,
            "cache_misses": 0,
            "cycles": 254686,
            "dynamic_energy_pj": 5174012298.589194,
            "hbm_read_bytes": 25547776,
            "hbm_write_bytes": 4925440,
            "instruction_count": 111355,
            "peak_power_upper_bound_w": 22.821056704036405,
            "peak_window_power_w": 19.69401699661881,
            "total_energy_pj": 5479617228.743997
          },
          "mean_sm_utilization": {
            "tc": 0.3518567962118059,
            "rf_read": 0.6108364810001335,
            "rf_write": 0.4505292791908468,
            "vec": 0.06465903112067409,
            "sfu": 0.031491719215033416
          }
        },
        "M2_D1": {
          "sha256": "43789102433e882cefaf33436bbb8230fb744b88a56156639a264004b40a4327",
          "source_lines": 28939,
          "ops": {
            "WG.BEGIN": 49,
            "BARRIER": 123,
            "LD": 3648,
            "REDUCE": 1280,
            "VEC": 11920,
            "ST": 1888,
            "WAIT": 1686,
            "SFU": 928,
            "MMA.ACC": 7360,
            "STEP.COMMIT": 8,
            "WG.END": 49
          },
          "mma_shapes": {
            "(3, 16, 32)": 448,
            "(1, 16, 32)": 6720,
            "(1, 1, 32)": 8,
            "(1, 32, 1)": 8,
            "(1, 33, 5)": 64,
            "(1, 2, 32)": 8,
            "(1, 32, 2)": 8,
            "(1, 3, 32)": 8,
            "(1, 32, 3)": 8,
            "(1, 4, 32)": 8,
            "(1, 32, 4)": 8,
            "(1, 5, 32)": 8,
            "(1, 32, 5)": 8,
            "(1, 6, 32)": 8,
            "(1, 32, 6)": 8,
            "(1, 7, 32)": 8,
            "(1, 32, 7)": 8,
            "(1, 8, 32)": 8,
            "(1, 32, 8)": 8
          },
          "groups": {
            "c": 0,
            "p0": 0,
            "p1": 0,
            "p2": 1,
            "p3": 1,
            "p4": 2,
            "p5": 2,
            "p6": 3,
            "p7": 3,
            "p8": 4,
            "p9": 4,
            "p10": 5,
            "p11": 5,
            "p12": 6,
            "p13": 6,
            "p14": 7,
            "p15": 7,
            "p16": 8,
            "p17": 8,
            "p18": 9,
            "p19": 9,
            "p20": 10,
            "p21": 10,
            "p22": 11,
            "p23": 11,
            "p24": 12,
            "p25": 12,
            "p26": 13,
            "p27": 13,
            "p28": 14,
            "p29": 14,
            "p30": 15,
            "p31": 15,
            "a0": 0,
            "a1": 1,
            "a2": 2,
            "a3": 3,
            "a4": 4,
            "a5": 5,
            "a6": 6,
            "a7": 7,
            "a8": 8,
            "a9": 9,
            "a10": 10,
            "a11": 11,
            "a12": 12,
            "a13": 13,
            "a14": 14,
            "a15": 15
          },
          "segments_by_commit": [
            {
              "LD:layer0/wqkv": 196608,
              "LD:layer0/w2": 262144,
              "LD:layer0/wo": 65536,
              "LD:layer1/w2": 262144,
              "LD:layer1/wqkv": 196608,
              "LD:layer1/wo": 65536,
              "LD:layer0/w1": 262144,
              "LD:layer1/w1": 262144,
              "LD:layer0/b1": 2048,
              "LD:layer0/b2": 1024,
              "LD:layer1/b1": 2048,
              "LD:layer1/b2": 1024,
              "LD:layer0/ln2_g": 8192,
              "LD:layer0/ln2_b": 8192,
              "LD:layer0/ln1_g": 8192,
              "LD:layer0/ln1_b": 8192,
              "LD:layer1/ln2_g": 8192,
              "LD:layer1/ln2_b": 8192,
              "LD:layer1/ln1_g": 8192,
              "LD:layer1/ln1_b": 8192,
              "LD:layer0/history_k": 65536,
              "LD:layer0/history_v": 65536,
              "LD:layer1/history_k": 65536,
              "LD:layer1/history_v": 65536,
              "LD:input/step0": 9216,
              "ST:scratch": 14432,
              "LD:scratch": 80960,
              "ST:layer0/new_k": 512,
              "LD:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "LD:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step1": 9216,
              "ST:scratch": 14432,
              "LD:scratch": 80960,
              "ST:layer0/new_k": 512,
              "LD:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "LD:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step2": 9216,
              "ST:scratch": 14432,
              "LD:scratch": 80960,
              "ST:layer0/new_k": 512,
              "LD:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "LD:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step3": 9216,
              "ST:scratch": 14432,
              "LD:scratch": 80960,
              "ST:layer0/new_k": 512,
              "LD:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "LD:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step4": 9216,
              "ST:scratch": 14432,
              "LD:scratch": 80960,
              "ST:layer0/new_k": 512,
              "LD:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "LD:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step5": 9216,
              "ST:scratch": 14432,
              "LD:scratch": 80960,
              "ST:layer0/new_k": 512,
              "LD:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "LD:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step6": 9216,
              "ST:scratch": 14432,
              "LD:scratch": 80960,
              "ST:layer0/new_k": 512,
              "LD:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "LD:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step7": 9216,
              "ST:scratch": 14432,
              "LD:scratch": 80960,
              "ST:layer0/new_k": 512,
              "LD:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "LD:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {}
          ],
          "timing": {
            "area_mm2": 23.99856530432,
            "average_power_w": 6.788839475075046,
            "cache_hits": 0,
            "cache_misses": 0,
            "cycles": 30368,
            "dynamic_energy_pj": 375887532.8000785,
            "hbm_read_bytes": 2064448,
            "hbm_write_bytes": 135936,
            "instruction_count": 28939,
            "peak_power_upper_bound_w": 20.420508132607804,
            "peak_window_power_w": 18.8882096366068,
            "total_energy_pj": 412326954.358158
          },
          "mean_sm_utilization": {
            "tc": 0.28543203371970494,
            "rf_read": 0.3122036354056902,
            "rf_write": 0.14643703898840885,
            "vec": 0.09724051633298209,
            "sfu": 0.01356691253951528
          }
        }
      }
    },
    {
      "rank": 4,
      "name": "方案4",
      "score": 57655.77325101823,
      "submission_id": "TRF-685CB3E940A34A42",
      "created_at": "2026-09-27T17:41:26.223635+00:00",
      "score_source": "verified",
      "hardware": {
        "cache_mib": 0,
        "dma_depth": 1,
        "dma_engines": 4,
        "hbm_channels": 8,
        "multicast": true,
        "noc_bytes_per_cycle": 512,
        "reduction_units": 1,
        "rf_ports": "8R4W",
        "sfu_lanes": 8,
        "shared_kib": 0,
        "sm_count": 16,
        "sm_noc_bytes_per_cycle": 128,
        "tc_array": "8x16",
        "tc_count": 1,
        "tc_k_parallel": 2,
        "vector_lanes": 8,
        "version": "challenge-hardware-v0.4"
      },
      "gates": {
        "area_gate_passed": true,
        "area_mm2": 23.902565304319996,
        "eligible": true,
        "scenarios": {
          "M1_P1": {
            "cycles": 361868,
            "first_token_or_average_step_seconds": 0.000723736,
            "functional_passed": true,
            "latency_gate_passed": true,
            "peak_power_w": 19.593172068616152,
            "power_gate_passed": true,
            "throughput_ratio_to_baseline": 85.65829252655665,
            "throughput_tokens_per_second": 176860.07052295312
          },
          "M2_D1": {
            "cycles": 33551,
            "first_token_or_average_step_seconds": 8.38775e-06,
            "functional_passed": true,
            "latency_gate_passed": true,
            "peak_power_w": 18.224176325712495,
            "power_gate_passed": true,
            "throughput_ratio_to_baseline": 38.80754671991893,
            "throughput_tokens_per_second": 119221.48371136478
          }
        },
        "score": 57655.77325101823
      },
      "hashes": {
        "zip": "e8aad4649538473665cf3e9f3010196b70b12b9da8797b6c4a4cfb9c1b3e1850",
        "report": "5916bd23f1f8e6ff86869d993906ebc017610dae14b0679dfd39605b96eed02d",
        "hardware": "af89a4a111b225e4cc7e7f1663a247a2696a04e142a679f3b87e00c003eca9a7",
        "canonical_hardware": "a4623dab145a154fd50c77db2641945c6e2fae2d6eb7be5b2db2a65d099c36eb"
      },
      "trace_files": [],
      "programs": {
        "M1_P1": {
          "sha256": "2844afc5d33c4060f74d893fdddc31c59503d5513285e4f726b7d043a45cbfc9",
          "source_lines": 32721,
          "ops": {
            "WG.BEGIN": 49,
            "BARRIER": 231,
            "LD": 17442,
            "VEC": 17322,
            "MMA.ACC": 9900,
            "SFU": 726,
            "ST": 1902,
            "WAIT": 1224,
            "REDUCE": 3168,
            "STEP.COMMIT": 1,
            "WG.END": 49
          },
          "mma_shapes": {
            "(8, 1, 256)": 192,
            "(32, 32, 64)": 3120,
            "(32, 32, 32)": 96,
            "(32, 1, 32)": 48,
            "(32, 64, 32)": 48,
            "(32, 1, 64)": 48,
            "(32, 32, 1)": 384,
            "(32, 16, 64)": 3072,
            "(32, 16, 1)": 192,
            "(2, 1, 256)": 12,
            "(2, 16, 64)": 1536,
            "(1, 64, 32)": 48,
            "(1, 1, 32)": 48,
            "(1, 32, 64)": 48,
            "(1, 32, 1)": 48,
            "(2, 32, 1)": 96,
            "(2, 8, 128)": 768,
            "(2, 8, 1)": 96
          },
          "groups": {
            "g": 0,
            "pool_aux_0": 0,
            "pool_aux_1": 1,
            "pool_aux_2": 2,
            "pool_aux_3": 3,
            "pool_aux_4": 4,
            "pool_aux_5": 5,
            "pool_aux_6": 6,
            "pool_aux_7": 7,
            "pool_aux_8": 8,
            "pool_aux_9": 9,
            "pool_aux_10": 10,
            "pool_aux_11": 11,
            "pool_aux_12": 12,
            "pool_aux_13": 13,
            "pool_aux_14": 14,
            "pool_aux_15": 15,
            "rfgemm_0": 0,
            "rfgemm_1": 1,
            "rfgemm_2": 2,
            "rfgemm_3": 3,
            "rfgemm_4": 4,
            "rfgemm_5": 5,
            "rfgemm_6": 6,
            "rfgemm_7": 7,
            "rfgemm_8": 8,
            "rfgemm_9": 9,
            "rfgemm_10": 10,
            "rfgemm_11": 11,
            "rfgemm_12": 12,
            "rfgemm_13": 13,
            "rfgemm_14": 14,
            "rfgemm_15": 15,
            "rfgemm_16": 0,
            "rfgemm_17": 1,
            "rfgemm_18": 2,
            "rfgemm_19": 3,
            "rfgemm_20": 4,
            "rfgemm_21": 5,
            "rfgemm_22": 6,
            "rfgemm_23": 7,
            "rfgemm_24": 8,
            "rfgemm_25": 9,
            "rfgemm_26": 10,
            "rfgemm_27": 11,
            "rfgemm_28": 12,
            "rfgemm_29": 13,
            "rfgemm_30": 14,
            "rfgemm_31": 15
          },
          "segments_by_commit": [
            {
              "LD:layer0/ln1_g": 16384,
              "LD:layer0/ln1_b": 16384,
              "LD:input/prompt": 262144,
              "ST:scratch": 5373952,
              "LD:scratch": 44171264,
              "LD:layer0/wqkv": 3145728,
              "ST:layer0/new_k": 131072,
              "ST:layer0/new_v": 131072,
              "LD:layer0/wo": 1048576,
              "LD:layer0/ln2_g": 16384,
              "LD:layer0/ln2_b": 16384,
              "LD:layer0/w1": 1048576,
              "LD:layer0/b1": 16384,
              "LD:layer0/w2": 1048576,
              "LD:layer0/b2": 4096,
              "LD:layer1/ln1_g": 16384,
              "LD:layer1/ln1_b": 16384,
              "LD:layer1/wqkv": 3145728,
              "ST:layer1/new_k": 131072,
              "ST:layer1/new_v": 131072,
              "LD:layer1/wo": 1048576,
              "LD:layer1/ln2_g": 16384,
              "LD:layer1/ln2_b": 16384,
              "LD:layer1/w1": 1048576,
              "LD:layer1/b1": 16384,
              "LD:layer1/w2": 1048576,
              "LD:layer1/b2": 4096,
              "LD:layer2/ln1_g": 16384,
              "LD:layer2/ln1_b": 16384,
              "LD:layer2/wqkv": 3145728,
              "ST:layer2/new_k": 131072,
              "ST:layer2/new_v": 131072,
              "LD:layer2/wo": 1048576,
              "LD:layer2/ln2_g": 16384,
              "LD:layer2/ln2_b": 16384,
              "LD:layer2/w1": 1048576,
              "LD:layer2/b1": 16384,
              "LD:layer2/w2": 1048576,
              "LD:layer2/b2": 4096,
              "ST:output/hidden": 131072
            },
            {
              "LD:layer0/ln1_g": 1024,
              "LD:layer0/ln1_b": 1024,
              "LD:input/step0": 4096,
              "ST:scratch": 71680,
              "LD:layer0/wqkv": 786432,
              "LD:scratch": 1267712,
              "ST:layer0/new_k": 2048,
              "ST:layer0/new_v": 2048,
              "LD:layer0/new_k": 133120,
              "LD:layer0/new_v": 133120,
              "LD:layer0/wo": 262144,
              "LD:layer0/ln2_g": 1024,
              "LD:layer0/ln2_b": 1024,
              "LD:layer0/w1": 1048576,
              "LD:layer0/b1": 4096,
              "LD:layer0/w2": 1048576,
              "LD:layer0/b2": 1024,
              "LD:layer1/ln1_g": 1024,
              "LD:layer1/ln1_b": 1024,
              "LD:layer1/wqkv": 786432,
              "ST:layer1/new_k": 2048,
              "ST:layer1/new_v": 2048,
              "LD:layer1/new_k": 133120,
              "LD:layer1/new_v": 133120,
              "LD:layer1/wo": 262144,
              "LD:layer1/ln2_g": 1024,
              "LD:layer1/ln2_b": 1024,
              "LD:layer1/w1": 1048576,
              "LD:layer1/b1": 4096,
              "LD:layer1/w2": 1048576,
              "LD:layer1/b2": 1024,
              "LD:layer2/ln1_g": 1024,
              "LD:layer2/ln1_b": 1024,
              "LD:layer2/wqkv": 786432,
              "ST:layer2/new_k": 2048,
              "ST:layer2/new_v": 2048,
              "LD:layer2/new_k": 133120,
              "LD:layer2/new_v": 133120,
              "LD:layer2/wo": 262144,
              "LD:layer2/ln2_g": 1024,
              "LD:layer2/ln2_b": 1024,
              "LD:layer2/w1": 1048576,
              "LD:layer2/b1": 4096,
              "LD:layer2/w2": 1048576,
              "LD:layer2/b2": 1024,
              "ST:output/hidden": 2048
            }
          ],
          "timing": {
            "area_mm2": 23.902565304319996,
            "average_power_w": 9.747885088667733,
            "cache_hits": 0,
            "cache_misses": 0,
            "cycles": 361868,
            "dynamic_energy_pj": 6622416687.454847,
            "hbm_read_bytes": 33288704,
            "hbm_write_bytes": 6377472,
            "instruction_count": 52014,
            "peak_power_upper_bound_w": 22.940041723341576,
            "peak_window_power_w": 19.593172068616152,
            "total_energy_pj": 7054895362.532031
          },
          "mean_sm_utilization": {
            "tc": 0.24024298915626693,
            "rf_read": 0.4085218919606044,
            "rf_write": 0.3213554389998563,
            "vec": 0.11009643019001404,
            "sfu": 0.0478413399361093
          }
        },
        "M2_D1": {
          "sha256": "c797af645e600fa03031cdb5b3dc1bb69ee22fd520555c170c4e78f9c2f0703e",
          "source_lines": 20465,
          "ops": {
            "WG.BEGIN": 64,
            "BARRIER": 25,
            "LD": 2672,
            "VEC": 9568,
            "MMA.ACC": 2624,
            "REDUCE": 1280,
            "SFU": 1024,
            "ST": 1472,
            "WAIT": 1664,
            "STEP.COMMIT": 8,
            "WG.END": 64
          },
          "mma_shapes": {
            "(2, 24, 64)": 64,
            "(2, 32, 64)": 64,
            "(1, 24, 64)": 512,
            "(1, 32, 32)": 512,
            "(1, 1, 32)": 8,
            "(1, 32, 1)": 8,
            "(1, 33, 4)": 64,
            "(1, 16, 64)": 256,
            "(1, 32, 64)": 512,
            "(1, 16, 128)": 512,
            "(1, 2, 32)": 8,
            "(1, 32, 2)": 8,
            "(1, 3, 32)": 8,
            "(1, 32, 3)": 8,
            "(1, 4, 32)": 8,
            "(1, 32, 4)": 8,
            "(1, 5, 32)": 8,
            "(1, 32, 5)": 8,
            "(1, 6, 32)": 8,
            "(1, 32, 6)": 8,
            "(1, 7, 32)": 8,
            "(1, 32, 7)": 8,
            "(1, 8, 32)": 8,
            "(1, 32, 8)": 8
          },
          "groups": {
            "g": 0,
            "rf_l0w1": 1,
            "rf_l0w2": 2,
            "rf_l0w3": 3,
            "rf_l0w4": 4,
            "rf_l0w5": 5,
            "rf_l0w6": 6,
            "rf_l0w7": 7,
            "rf_l0w8": 8,
            "rf_l0w9": 9,
            "rf_l0w10": 10,
            "rf_l0w11": 11,
            "rf_l0w12": 12,
            "rf_l0w13": 13,
            "rf_l0w14": 14,
            "rf_l0w15": 15,
            "rf_l1w0": 0,
            "rf_l1w1": 1,
            "rf_l1w2": 2,
            "rf_l1w3": 3,
            "rf_l1w4": 4,
            "rf_l1w5": 5,
            "rf_l1w6": 6,
            "rf_l1w7": 7,
            "rf_l1w8": 8,
            "rf_l1w9": 9,
            "rf_l1w10": 10,
            "rf_l1w11": 11,
            "rf_l1w12": 12,
            "rf_l1w13": 13,
            "rf_l1w14": 14,
            "rf_l1w15": 15,
            "rf_att_l0h0p0": 0,
            "rf_att_l0h0p1": 1,
            "rf_att_l0h0p2": 2,
            "rf_att_l0h0p3": 3,
            "rf_att_l0h1p0": 4,
            "rf_att_l0h1p1": 5,
            "rf_att_l0h1p2": 6,
            "rf_att_l0h1p3": 7,
            "rf_att_l0h2p0": 8,
            "rf_att_l0h2p1": 9,
            "rf_att_l0h2p2": 10,
            "rf_att_l0h2p3": 11,
            "rf_att_l0h3p0": 12,
            "rf_att_l0h3p1": 13,
            "rf_att_l0h3p2": 14,
            "rf_att_l0h3p3": 15,
            "rf_att_l1h0p0": 0,
            "rf_att_l1h0p1": 1,
            "rf_att_l1h0p2": 2,
            "rf_att_l1h0p3": 3,
            "rf_att_l1h1p0": 4,
            "rf_att_l1h1p1": 5,
            "rf_att_l1h1p2": 6,
            "rf_att_l1h1p3": 7,
            "rf_att_l1h2p0": 8,
            "rf_att_l1h2p1": 9,
            "rf_att_l1h2p2": 10,
            "rf_att_l1h2p3": 11,
            "rf_att_l1h3p0": 12,
            "rf_att_l1h3p1": 13,
            "rf_att_l1h3p2": 14,
            "rf_att_l1h3p3": 15
          },
          "segments_by_commit": [
            {
              "LD:layer0/ln1_g": 8192,
              "LD:layer1/ln1_g": 8192,
              "LD:layer0/ln1_b": 8192,
              "LD:layer1/ln1_b": 8192,
              "LD:layer0/ln2_g": 8192,
              "LD:layer1/ln2_g": 8192,
              "LD:layer0/ln2_b": 8192,
              "LD:layer1/ln2_b": 8192,
              "LD:layer0/b1": 2048,
              "LD:layer1/b1": 2048,
              "LD:layer0/b2": 512,
              "LD:layer1/b2": 512,
              "LD:layer0/wqkv": 196608,
              "LD:layer1/wqkv": 196608,
              "LD:layer0/wo": 65536,
              "LD:layer1/wo": 65536,
              "LD:layer0/w1": 262144,
              "LD:layer1/w1": 262144,
              "LD:layer0/w2": 262144,
              "LD:layer1/w2": 262144,
              "LD:layer0/history_k": 65536,
              "LD:layer1/history_k": 65536,
              "LD:layer0/history_v": 65536,
              "LD:layer1/history_v": 65536,
              "LD:input/step0": 8704,
              "ST:scratch": 16000,
              "LD:scratch": 79488,
              "ST:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step1": 8704,
              "ST:scratch": 16000,
              "LD:scratch": 79488,
              "ST:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step2": 8704,
              "ST:scratch": 16000,
              "LD:scratch": 79488,
              "ST:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step3": 8704,
              "ST:scratch": 16000,
              "LD:scratch": 79488,
              "ST:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step4": 8704,
              "ST:scratch": 16000,
              "LD:scratch": 79488,
              "ST:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step5": 8704,
              "ST:scratch": 16000,
              "LD:scratch": 79488,
              "ST:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step6": 8704,
              "ST:scratch": 16000,
              "LD:scratch": 79488,
              "ST:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step7": 8704,
              "ST:scratch": 16000,
              "LD:scratch": 79488,
              "ST:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {}
          ],
          "timing": {
            "area_mm2": 23.902565304319996,
            "average_power_w": 6.160594471853109,
            "cache_hits": 0,
            "cache_misses": 0,
            "cycles": 33551,
            "dynamic_energy_pj": 373290461.8240253,
            "hbm_read_bytes": 1998848,
            "hbm_write_bytes": 148480,
            "instruction_count": 20465,
            "peak_power_upper_bound_w": 22.873142203632316,
            "peak_window_power_w": 18.224176325712495,
            "total_energy_pj": 413388210.2502873
          },
          "mean_sm_utilization": {
            "tc": 0.26204882119757983,
            "rf_read": 0.29769604482727785,
            "rf_write": 0.13284253822538822,
            "vec": 0.11385651694435338,
            "sfu": 0.023367410807427498
          }
        }
      }
    },
    {
      "rank": 5,
      "name": "方案5",
      "score": 52508.65987540518,
      "submission_id": "TRF-46FFD60D9C754BBE",
      "created_at": "2026-09-28T08:43:41.821944+00:00",
      "score_source": "verified",
      "hardware": {
        "cache_mib": 0,
        "dma_depth": 2,
        "dma_engines": 1,
        "hbm_channels": 8,
        "multicast": true,
        "noc_bytes_per_cycle": 512,
        "reduction_units": 0,
        "rf_ports": "8R4W",
        "sfu_lanes": 8,
        "shared_kib": 0,
        "sm_count": 24,
        "sm_noc_bytes_per_cycle": 64,
        "tc_array": "8x16",
        "tc_count": 1,
        "tc_k_parallel": 1,
        "vector_lanes": 16,
        "version": "challenge-hardware-v0.4"
      },
      "gates": {
        "area_gate_passed": true,
        "area_mm2": 23.64464795648,
        "eligible": true,
        "scenarios": {
          "M1_P1": {
            "cycles": 352305,
            "first_token_or_average_step_seconds": 0.00070461,
            "functional_passed": true,
            "latency_gate_passed": true,
            "peak_power_w": 16.957613829934928,
            "power_gate_passed": true,
            "throughput_ratio_to_baseline": 87.98340926186117,
            "throughput_tokens_per_second": 181660.77688366614
          },
          "M2_D1": {
            "cycles": 41549,
            "first_token_or_average_step_seconds": 1.038725e-05,
            "functional_passed": true,
            "latency_gate_passed": true,
            "peak_power_w": 19.35723071733613,
            "power_gate_passed": true,
            "throughput_ratio_to_baseline": 31.337264434763775,
            "throughput_tokens_per_second": 96271.8717658668
          }
        },
        "score": 52508.65987540518
      },
      "hashes": {
        "zip": "b39168be598adfa67e7d184ba720c4b35548a9cb6ecc86131731e452bd82c491",
        "report": "3515ed0ed701c85c5c3f1ba3e3cf668188d25f1481a6006052d66f07c3d64d5f",
        "hardware": "3ac0ebd1f06c8c89b5c40954859f0b20ad8761796b3699766d05c269f64254fb",
        "canonical_hardware": "ebdc196649d569050013abc3e5b16b498fbf7159be0c2093e4860ef4b8cfabe9"
      },
      "trace_files": [],
      "programs": {
        "M1_P1": {
          "sha256": "7c96cfd20a9cd880cb5e6659098a19aab5f49a9fb6b172fa16cd31904a540aa7",
          "source_lines": 39342,
          "ops": {
            "WG.BEGIN": 24,
            "VEC": 20124,
            "LD": 14594,
            "WAIT": 8300,
            "REDUCE": 1656,
            "SFU": 1044,
            "ST": 2472,
            "MMA.ACC": 23760,
            "STEP.COMMIT": 1,
            "WG.END": 24
          },
          "mma_shapes": {
            "(32, 32, 16)": 18432,
            "(16, 16, 32)": 96,
            "(16, 16, 1)": 48,
            "(16, 32, 16)": 288,
            "(16, 32, 32)": 384,
            "(16, 32, 1)": 48,
            "(16, 48, 1)": 48,
            "(16, 64, 1)": 48,
            "(32, 32, 1)": 480,
            "(2, 32, 32)": 1536,
            "(1, 32, 32)": 192,
            "(1, 1, 32)": 48,
            "(1, 32, 1)": 48,
            "(2, 32, 1)": 96,
            "(2, 16, 1)": 48,
            "(2, 16, 16)": 768,
            "(2, 32, 16)": 1152
          },
          "groups": {
            "p0": 0,
            "p1": 1,
            "p2": 2,
            "p3": 3,
            "p4": 4,
            "p5": 5,
            "p6": 6,
            "p7": 7,
            "p8": 8,
            "p9": 9,
            "p10": 10,
            "p11": 11,
            "p12": 12,
            "p13": 13,
            "p14": 14,
            "p15": 15,
            "p16": 16,
            "p17": 17,
            "p18": 18,
            "p19": 19,
            "p20": 20,
            "p21": 21,
            "p22": 22,
            "p23": 23
          },
          "segments_by_commit": [
            {
              "LD:layer0/ln1_g": 24576,
              "LD:layer0/ln1_b": 24576,
              "LD:input/prompt": 262144,
              "ST:scratch": 6160384,
              "LD:scratch": 22282240,
              "LD:layer0/wqkv": 1572864,
              "ST:layer0/new_k": 131072,
              "ST:layer0/new_v": 131072,
              "LD:layer0/new_k": 131072,
              "LD:layer0/new_v": 131072,
              "LD:layer0/wo": 524288,
              "LD:layer0/ln2_g": 24576,
              "LD:layer0/ln2_b": 24576,
              "LD:layer0/b1": 8192,
              "LD:layer0/w1": 2097152,
              "LD:layer0/b2": 2048,
              "LD:layer0/w2": 2097152,
              "LD:layer1/ln1_g": 24576,
              "LD:layer1/ln1_b": 24576,
              "LD:layer1/wqkv": 1572864,
              "ST:layer1/new_k": 131072,
              "ST:layer1/new_v": 131072,
              "LD:layer1/new_k": 131072,
              "LD:layer1/new_v": 131072,
              "LD:layer1/wo": 524288,
              "LD:layer1/ln2_g": 24576,
              "LD:layer1/ln2_b": 24576,
              "LD:layer1/b1": 8192,
              "LD:layer1/w1": 2097152,
              "LD:layer1/b2": 2048,
              "LD:layer1/w2": 2097152,
              "LD:layer2/ln1_g": 24576,
              "LD:layer2/ln1_b": 24576,
              "LD:layer2/wqkv": 1572864,
              "ST:layer2/new_k": 131072,
              "ST:layer2/new_v": 131072,
              "LD:layer2/new_k": 131072,
              "LD:layer2/new_v": 131072,
              "LD:layer2/wo": 524288,
              "LD:layer2/ln2_g": 24576,
              "LD:layer2/ln2_b": 24576,
              "LD:layer2/b1": 8192,
              "LD:layer2/w1": 2097152,
              "LD:layer2/b2": 2048,
              "LD:layer2/w2": 2097152,
              "ST:output/hidden": 131072
            },
            {
              "LD:layer0/ln1_g": 2048,
              "LD:layer0/ln1_b": 2048,
              "LD:input/step0": 4096,
              "ST:scratch": 96256,
              "LD:scratch": 593920,
              "LD:layer0/wqkv": 786432,
              "ST:layer0/new_k": 2048,
              "ST:layer0/new_v": 2048,
              "LD:layer0/new_k": 133120,
              "LD:layer0/new_v": 133120,
              "LD:layer0/wo": 262144,
              "LD:layer0/ln2_g": 2048,
              "LD:layer0/ln2_b": 2048,
              "LD:layer0/b1": 4096,
              "LD:layer0/w1": 1048576,
              "LD:layer0/b2": 1024,
              "LD:layer0/w2": 1048576,
              "LD:layer1/ln1_g": 2048,
              "LD:layer1/ln1_b": 2048,
              "LD:layer1/wqkv": 786432,
              "ST:layer1/new_k": 2048,
              "ST:layer1/new_v": 2048,
              "LD:layer1/new_k": 133120,
              "LD:layer1/new_v": 133120,
              "LD:layer1/wo": 262144,
              "LD:layer1/ln2_g": 2048,
              "LD:layer1/ln2_b": 2048,
              "LD:layer1/b1": 4096,
              "LD:layer1/w1": 1048576,
              "LD:layer1/b2": 1024,
              "LD:layer1/w2": 1048576,
              "LD:layer2/ln1_g": 2048,
              "LD:layer2/ln1_b": 2048,
              "LD:layer2/wqkv": 786432,
              "ST:layer2/new_k": 2048,
              "ST:layer2/new_v": 2048,
              "LD:layer2/new_k": 133120,
              "LD:layer2/new_v": 133120,
              "LD:layer2/wo": 262144,
              "LD:layer2/ln2_g": 2048,
              "LD:layer2/ln2_b": 2048,
              "LD:layer2/b1": 4096,
              "LD:layer2/w1": 1048576,
              "LD:layer2/b2": 1024,
              "LD:layer2/w2": 1048576,
              "ST:output/hidden": 2048
            }
          ],
          "timing": {
            "area_mm2": 23.64464795648,
            "average_power_w": 8.79950566832929,
            "cache_hits": 0,
            "cache_misses": 0,
            "cycles": 352305,
            "dynamic_energy_pj": 5783713304.046119,
            "hbm_read_bytes": 27695360,
            "hbm_write_bytes": 7188480,
            "instruction_count": 71999,
            "peak_power_upper_bound_w": 22.192829429681368,
            "peak_window_power_w": 16.957613829934928,
            "total_energy_pj": 6200219688.961503
          },
          "mean_sm_utilization": {
            "tc": 0.31305261066405526,
            "rf_read": 0.33681185336569164,
            "rf_write": 0.29101890880534387,
            "vec": 0.0495713941045401,
            "sfu": 0.03147556804473397
          }
        },
        "M2_D1": {
          "sha256": "3269111204509cfa19f0046cdf79753927c8797bb2070d4ce6ca66b5ccd1dba3",
          "source_lines": 10765,
          "ops": {
            "WG.BEGIN": 48,
            "LD": 2944,
            "VEC": 19400,
            "MMA.ACC": 5184,
            "WAIT": 1646,
            "REDUCE": 3328,
            "SFU": 2048,
            "ST": 1944,
            "STEP.COMMIT": 8,
            "WG.END": 48
          },
          "mma_shapes": {
            "(1, 16, 64)": 3008,
            "(1, 22, 32)": 136,
            "(1, 32, 22)": 136,
            "(1, 21, 32)": 192,
            "(1, 32, 21)": 192,
            "(1, 32, 6)": 512,
            "(1, 16, 32)": 640,
            "(1, 16, 48)": 256,
            "(1, 23, 32)": 8,
            "(1, 32, 23)": 8,
            "(1, 24, 32)": 8,
            "(1, 32, 24)": 8,
            "(1, 25, 32)": 8,
            "(1, 32, 25)": 8,
            "(1, 26, 32)": 8,
            "(1, 32, 26)": 8,
            "(1, 27, 32)": 8,
            "(1, 32, 27)": 8,
            "(1, 28, 32)": 8,
            "(1, 32, 28)": 8,
            "(1, 29, 32)": 8,
            "(1, 32, 29)": 8
          },
          "groups": {
            "L0w0": 0,
            "L0w1": 1,
            "L0w2": 2,
            "L0w3": 3,
            "L0w4": 4,
            "L0w5": 5,
            "L0w6": 6,
            "L0w7": 7,
            "L0w8": 8,
            "L0w9": 9,
            "L0w10": 10,
            "L0w11": 11,
            "L0w12": 12,
            "L0w13": 13,
            "L0w14": 14,
            "L0w15": 15,
            "L0w16": 16,
            "L0w17": 17,
            "L0w18": 18,
            "L0w19": 19,
            "L0w20": 20,
            "L0w21": 21,
            "L0w22": 22,
            "L0w23": 23,
            "L1w0": 0,
            "L1w1": 1,
            "L1w2": 2,
            "L1w3": 3,
            "L1w4": 4,
            "L1w5": 5,
            "L1w6": 6,
            "L1w7": 7,
            "L1w8": 8,
            "L1w9": 9,
            "L1w10": 10,
            "L1w11": 11,
            "L1w12": 12,
            "L1w13": 13,
            "L1w14": 14,
            "L1w15": 15,
            "L1w16": 16,
            "L1w17": 17,
            "L1w18": 18,
            "L1w19": 19,
            "L1w20": 20,
            "L1w21": 21,
            "L1w22": 22,
            "L1w23": 23
          },
          "segments_by_commit": [
            {
              "LD:layer0/ln1_g": 12288,
              "LD:layer0/ln1_b": 12288,
              "LD:layer0/ln2_g": 12288,
              "LD:layer0/ln2_b": 12288,
              "LD:layer0/wqkv": 196608,
              "LD:layer0/history_k": 65536,
              "LD:layer0/history_v": 65536,
              "LD:layer0/wo": 65536,
              "LD:layer0/w1": 262144,
              "LD:layer0/b1": 3584,
              "LD:layer0/w2": 262144,
              "LD:layer0/b2": 512,
              "LD:input/step0": 12288,
              "ST:scratch": 17792,
              "ST:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "LD:scratch": 191488,
              "LD:layer0/new_k": 512,
              "LD:layer0/new_v": 512,
              "LD:layer1/ln1_g": 12288,
              "LD:layer1/ln1_b": 12288,
              "LD:layer1/ln2_g": 12288,
              "LD:layer1/ln2_b": 12288,
              "LD:layer1/wqkv": 196608,
              "LD:layer1/history_k": 65536,
              "LD:layer1/history_v": 65536,
              "LD:layer1/wo": 65536,
              "LD:layer1/w1": 262144,
              "LD:layer1/b1": 3584,
              "LD:layer1/w2": 262144,
              "LD:layer1/b2": 512,
              "ST:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "LD:layer1/new_k": 512,
              "LD:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step1": 12288,
              "ST:scratch": 17792,
              "ST:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "LD:scratch": 191488,
              "LD:layer0/new_k": 512,
              "LD:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "LD:layer1/new_k": 512,
              "LD:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step2": 12288,
              "ST:scratch": 17792,
              "ST:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "LD:scratch": 191488,
              "LD:layer0/new_k": 512,
              "LD:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "LD:layer1/new_k": 512,
              "LD:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step3": 12288,
              "ST:scratch": 17792,
              "ST:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "LD:scratch": 191488,
              "LD:layer0/new_k": 512,
              "LD:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "LD:layer1/new_k": 512,
              "LD:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step4": 12288,
              "ST:scratch": 17792,
              "ST:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "LD:scratch": 191488,
              "LD:layer0/new_k": 512,
              "LD:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "LD:layer1/new_k": 512,
              "LD:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step5": 12288,
              "ST:scratch": 17792,
              "ST:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "LD:scratch": 191488,
              "LD:layer0/new_k": 512,
              "LD:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "LD:layer1/new_k": 512,
              "LD:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step6": 12288,
              "ST:scratch": 17792,
              "ST:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "LD:scratch": 191488,
              "LD:layer0/new_k": 512,
              "LD:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "LD:layer1/new_k": 512,
              "LD:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {
              "LD:input/step7": 12288,
              "ST:scratch": 17792,
              "ST:layer0/new_k": 512,
              "ST:layer0/new_v": 512,
              "LD:scratch": 191488,
              "LD:layer0/new_k": 512,
              "LD:layer0/new_v": 512,
              "ST:layer1/new_k": 512,
              "ST:layer1/new_v": 512,
              "LD:layer1/new_k": 512,
              "LD:layer1/new_v": 512,
              "ST:output/hidden": 512
            },
            {}
          ],
          "timing": {
            "area_mm2": 23.64464795648,
            "average_power_w": 5.3023408468217195,
            "cache_hits": 0,
            "cache_misses": 0,
            "cycles": 41549,
            "dynamic_energy_pj": 391493345.79200184,
            "hbm_read_bytes": 2124032,
            "hbm_write_bytes": 162816,
            "instruction_count": 36598,
            "peak_power_upper_bound_w": 20.758130224885573,
            "peak_window_power_w": 19.35723071733613,
            "total_energy_pj": 440613919.6891912
          },
          "mean_sm_utilization": {
            "tc": 0.2841303842049949,
            "rf_read": 0.19750174492767578,
            "rf_write": 0.11140260094506887,
            "vec": 0.1208532896900848,
            "sfu": 0.018131202515904915
          }
        }
      }
    }
  ],
  "validation": {
    "exact_program_hashes": true,
    "canonical_hardware_hashes": true,
    "score_matches_board": true,
    "expanded_instruction_counts_match_reports": true,
    "d1_static_weights_and_history_loaded_only_before_first_commit": true,
    "reran_grading": false,
    "revalidated_hmac_signature": false
  },
  "limitations": [
    "Server verified status and saved official reports; no private HMAC key read.",
    "Historical trace claims are not all independently regraded.",
    "Static logical reads differ from physical HBM transactions."
  ]
}