==PROF== Kernoscope 0.1.0 - provenance-honest profile (no performance counters required)
==PROF== Target: bash bench/shape_sweep_target.sh bench/artifacts/.build/ledger_probe
==PROF== DRAM 估算不可用：没有 NVML 采样点（--ks-no-nvml 打开，或采样器不可用）；缺少本机实测峰值带宽：先跑 tools/machine-profile.py，再加 --ks-calibration machine-profile.json（或 --ks-peak-gbps）
==PROF== Kernel time: 10.786 ms of 4174.821 ms wall (0.3%)
==PROF== Window: mode=auto 采样 0 点 / 参与积分 0.0 s / 裁掉 0.0 s / 进程 wall 4.2 s
==PROF== source[kernel activity]: nsys sqlite CUPTI_ACTIVITY_KIND_KERNEL (no counters needed)
==PROF== source[device counters]: nvidia-smi (energy/PCIe/mem-busy are management counters, available without admin)
==PROF== source[static]: not used
==PROF== source[peaks]: not provided

Kernel: copy_k(const float *, float *, long)
  nsys short name: copy_k
  1050 launch(es), 10.7858 ms total, 0.0103 ms mean
  ⚠ 3 组 launch 配置（grid1024x1x1 block256x1x1 smem0, grid256x1x1 block256x1x1 smem0, grid64x1x1 block256x1x1 smem0）：逐组读数见 LaunchConfigs

  Section: Duration                           [exact]
    gpu__time_duration.sum                                10.7858 ms       [exact]
    ks__launch_time_mean                                   0.0103 ms       [exact]  (**多组 launch 配置的混合均值**，不可当作某一组配置的读数)
    ks__launch_time_min                                    0.0095 ms       [exact]
    ks__launch_time_max                                    0.0119 ms       [exact]
    ks__launch_count                                         1050          [exact]
    ks__launch_config_count                                     3          [exact]  (逐组读数见 LaunchConfigs section)

  Section: LaunchStats                        [exact] (+2 n/a)
    launch__grid_size                                         n/a          [unavailable]  该 kernel 有 3 组不同 launch 配置（grid=64/256/1024）：单一数值会误导 2/3 的 launch，故报 n/a；逐组读数见 LaunchConfigs section
    launch__block_size                                        n/a          [unavailable]  该 kernel 有 3 组不同 launch 配置（block=256）：单一数值会误导，故报 n/a
    ks__grid_sizes_observed                           64|256|1024          [exact]  (取值集合口径；逐组时长与 occupancy 见 LaunchConfigs section)
    ks__block_sizes_observed                                  256          [exact]
    launch__registers_per_thread                               32          [exact]
    ks__registers_used_for_occupancy                           32          [exact]
    launch__shared_mem_per_block_static                         0 B        [exact]
    launch__shared_mem_per_block_dynamic                        0 B        [exact]
    launch__local_mem_per_thread                                0 B        [exact]
    ks__cuda_graph_capture                                    yes          [exact]
    ks__dominant_config                            grid1024x1x1_block256x1x1_smem0          [exact]  (350/1050 次 launch；本节未标注逐组的数值均取这组配置)

  Section: LaunchConfigs                      [exact]
    ks__cfg_grid1024x1x1_block256x1x1_smem0__launch_time_mean_ms         0.0098 ms       [exact]
    ks__cfg_grid1024x1x1_block256x1x1_smem0__launch_count            350          [exact]
    ks__cfg_grid1024x1x1_block256x1x1_smem0__theoretical_occupancy_pct            100 %        [exact]
    ks__cfg_grid256x1x1_block256x1x1_smem0__launch_time_mean_ms         0.0096 ms       [exact]
    ks__cfg_grid256x1x1_block256x1x1_smem0__launch_count            350          [exact]
    ks__cfg_grid256x1x1_block256x1x1_smem0__theoretical_occupancy_pct            100 %        [exact]
    ks__cfg_grid64x1x1_block256x1x1_smem0__launch_time_mean_ms         0.0114 ms       [exact]
    ks__cfg_grid64x1x1_block256x1x1_smem0__launch_count            350          [exact]
    ks__cfg_grid64x1x1_block256x1x1_smem0__theoretical_occupancy_pct            100 %        [exact]

  Section: Occupancy                          [exact] (+1 n/a)
    ks__theoretical_occupancy_pct                             100 %        [exact]
    ks__blocks_per_sm                                           6          [exact]
    ks__warps_per_sm                                           48          [exact]
    ks__occupancy_limiter                           threads+warps          [exact]  (并列的限制因素用 + 连接；本节按主导配置 grid1024x1x1_block256x1x1_smem0 计算，逐组见 LaunchConfigs)
    sm__maximum_warps_per_active_cycle_pct                    n/a          [unavailable]  requires GPU hardware performance counters, which are restricted here (RmProfilingAdminOnly=1 and no CAP_SYS_ADMIN/CAP_PERFMON in the container); CUPTI reports 

  Section: SpeedOfLight                       [unavailable] (+2 n/a)
    est__achieved_tflops                                      n/a          [unavailable]  workload not provided: pass --ks-gemm M,N,K or --ks-flops/--ks-bytes (plus --ks-calibration machine-profile.json for measured peaks)
    sm__throughput.avg.pct_of_peak_sustained_elapsed            n/a          [unavailable]  requires GPU hardware performance counters, which are restricted here (RmProfilingAdminOnly=1 and no CAP_SYS_ADMIN/CAP_PERFMON in the container); CUPTI reports 

----------------------------------------------------------------------------------------------------
Kernel: (device)
  0 launch(es), 0.0000 ms total, 0.0000 ms mean

  Section: DeviceStatistics                   [unavailable] (+1 n/a)
    meas__nvml_mem_busy_mean_pct                              n/a          [unavailable]  没有 NVML 采样点：采样被关闭（--ks-no-nvml）或 nvidia-smi/pynvml 不可用；没有它就无法给出显存忙碌代理

  Section: DramTraffic                        [derived] (+4 n/a)
    est__dram_bytes_total                                     n/a          [unavailable]  没有 NVML 采样点（--ks-no-nvml 打开，或采样器不可用）；缺少本机实测峰值带宽：先跑 tools/machine-profile.py，再加 --ks-calibration machine-profile.json（或 --ks-peak-gbps）
    est__dram_bytes_per_launch                                n/a          [unavailable]  没有 NVML 采样点（--ks-no-nvml 打开，或采样器不可用）；缺少本机实测峰值带宽：先跑 tools/machine-profile.py，再加 --ks-calibration machine-profile.json（或 --ks-peak-gbps）
    est__dram_throughput_pct_of_peak                          n/a          [unavailable]  没有 NVML 采样点（--ks-no-nvml 打开，或采样器不可用）；缺少本机实测峰值带宽：先跑 tools/machine-profile.py，再加 --ks-calibration machine-profile.json（或 --ks-peak-gbps）
    ks__sampler_window_coverage_pct                             0 %        [derived]  [resolution 1 个采样点]  (前提①「窗口被被测负载主导」：<80% 不给 per-launch 归因，<50% 连总量也不给)
    est__dram_bytes_per_launch_trace                          n/a          [unavailable]  未提供访存轨迹：加 --ks-memtrace <trace> 可换用轨迹口径，在 NVML 代理口径不成立时仍能完成账本对拍

----------------------------------------------------------------------------------------------------

Provenance summary: exact=28, derived=1, unavailable=10
Legend: exact=与 ncu 同源 | derived=本机实测标定/模型 | measured=我们的探针实测 | estimated=采样或代理(附误差) | unavailable=需要硬件计数器，只报 n/a
Output(json): bench/artifacts/20261008-225555/rep-shape.ksreport.json
Output(perfetto): bench/artifacts/20261008-225555/rep-shape.perfetto.json
Open the timeline at https://ui.perfetto.dev -> Load trace -> the file above
