{
  "generated_at": "2026-09-10T00:19:01.199727+00:00",
  "source": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/analysis/nsys_worker_flush_v2_physical/kernel_invocations.jsonl.gz",
  "source_sha256": "b93ccf0d00c664d889b6c4561d862e1ffc9b74ccfc7ee21ee62da6d514f618cb",
  "scanned_calls": 1091749,
  "steady_calls": 176984,
  "calls_by_physical_gpu": {
    "0": 44324,
    "2": 44220,
    "1": 44220,
    "3": 44220
  },
  "physical_launch_groups": 970,
  "kernel_names": 102,
  "names_with_multiple_launch_signatures": 36,
  "windows_ns": [
    [
      260193278338,
      261040772109
    ],
    [
      261040772109,
      261893917185
    ],
    [
      261893917185,
      262747644781
    ],
    [
      262747644781,
      263603828424
    ]
  ],
  "csv": "/data/zhoutaichang/feature/Lingbot_world_realtime/runs/analysis_report_20260908/analysis/nsys_worker_flush_v2_physical/launch_resources_by_physical_gpu_steady.csv",
  "csv_sha256": "8affd589e3c7c1bb4faa2c1be91709aadd9478f2a952e7a1f1bdff0c7b2549a9",
  "scope": "Recorded steady intervals5-8. Duration clipped to windows; sums across GPUs are not wall time. Register/shared allocation is launch metadata, not NCU occupancy, bandwidth or stalls. Launch signatures do not establish tensor shapes.",
  "top10": [
    {
      "physical_gpu": 3,
      "name": "nvjet_sm90_tst_128x200_64x5_2x1_v_bz_coopA_bias_TNT",
      "gridX": 2,
      "gridY": 66,
      "gridZ": 1,
      "blockX": 384,
      "blockY": 1,
      "blockZ": 1,
      "registersPerThread": 168,
      "staticSharedMemory": 0,
      "dynamicSharedMemory": 226492,
      "calls": 7200,
      "first_event_id": 870934,
      "clipped_summed_gpu_ms": 862.418299
    },
    {
      "physical_gpu": 0,
      "name": "nvjet_sm90_tst_128x200_64x5_2x1_v_bz_coopA_bias_TNT",
      "gridX": 2,
      "gridY": 66,
      "gridZ": 1,
      "blockX": 384,
      "blockY": 1,
      "blockZ": 1,
      "registersPerThread": 168,
      "staticSharedMemory": 0,
      "dynamicSharedMemory": 226492,
      "calls": 7200,
      "first_event_id": 870931,
      "clipped_summed_gpu_ms": 860.113418
    },
    {
      "physical_gpu": 1,
      "name": "nvjet_sm90_tst_128x200_64x5_2x1_v_bz_coopA_bias_TNT",
      "gridX": 2,
      "gridY": 66,
      "gridZ": 1,
      "blockX": 384,
      "blockY": 1,
      "blockZ": 1,
      "registersPerThread": 168,
      "staticSharedMemory": 0,
      "dynamicSharedMemory": 226492,
      "calls": 7200,
      "first_event_id": 870929,
      "clipped_summed_gpu_ms": 857.673055
    },
    {
      "physical_gpu": 2,
      "name": "nvjet_sm90_tst_128x200_64x5_2x1_v_bz_coopA_bias_TNT",
      "gridX": 2,
      "gridY": 66,
      "gridZ": 1,
      "blockX": 384,
      "blockY": 1,
      "blockZ": 1,
      "registersPerThread": 168,
      "staticSharedMemory": 0,
      "dynamicSharedMemory": 226492,
      "calls": 7200,
      "first_event_id": 870935,
      "clipped_summed_gpu_ms": 852.013691
    },
    {
      "physical_gpu": 3,
      "name": "void cutlass::device_kernel<flash::enable_sm90_or_later<flash::FlashAttnFwdSm90<flash::CollectiveMainloopFwdSm90<(int)2, cute::tuple<cute::C<(int)1>, cute::C<(int)1>, cute::C<(int)1>>, cute::tuple<cute::C<(int)128>, cute::C<(int)160>, cute::C<(int)128>>, (int)128, cutlass::bfloat16_t, float, cutlass::arch::Sm90, (bool)0, (bool)0, (bool)0, (bool)1, (bool)0, (bool)0, (bool)0, (bool)1, (bool)1, (bool)1, (bool)0, (bool)0, cutlass::bfloat16_t, (int)1>, flash::CollectiveEpilogueFwd<cute::tuple<cute::C<(int)128>, cute::C<(int)128>, cute::C<(int)160>>, cute::tuple<cute::C<(int)1>, cute::C<(int)1>, cute::C<(int)1>>, cutlass::bfloat16_t, cutlass::arch::Sm90, (int)256, (bool)1, (bool)1, (bool)0, (bool)0, (int)1>, flash::VarlenDynamicPersistentTileScheduler<(int)128, (int)160, (int)256, (int)128, (bool)0, (bool)1, (bool)1, (bool)0, (bool)0, (bool)1>>>>(T1::Params)",
      "gridX": 132,
      "gridY": 1,
      "gridZ": 1,
      "blockX": 384,
      "blockY": 1,
      "blockZ": 1,
      "registersPerThread": 168,
      "staticSharedMemory": 0,
      "dynamicSharedMemory": 199680,
      "calls": 800,
      "first_event_id": 870925,
      "clipped_summed_gpu_ms": 800.775723
    },
    {
      "physical_gpu": 2,
      "name": "void cutlass::device_kernel<flash::enable_sm90_or_later<flash::FlashAttnFwdSm90<flash::CollectiveMainloopFwdSm90<(int)2, cute::tuple<cute::C<(int)1>, cute::C<(int)1>, cute::C<(int)1>>, cute::tuple<cute::C<(int)128>, cute::C<(int)160>, cute::C<(int)128>>, (int)128, cutlass::bfloat16_t, float, cutlass::arch::Sm90, (bool)0, (bool)0, (bool)0, (bool)1, (bool)0, (bool)0, (bool)0, (bool)1, (bool)1, (bool)1, (bool)0, (bool)0, cutlass::bfloat16_t, (int)1>, flash::CollectiveEpilogueFwd<cute::tuple<cute::C<(int)128>, cute::C<(int)128>, cute::C<(int)160>>, cute::tuple<cute::C<(int)1>, cute::C<(int)1>, cute::C<(int)1>>, cutlass::bfloat16_t, cutlass::arch::Sm90, (int)256, (bool)1, (bool)1, (bool)0, (bool)0, (int)1>, flash::VarlenDynamicPersistentTileScheduler<(int)128, (int)160, (int)256, (int)128, (bool)0, (bool)1, (bool)1, (bool)0, (bool)0, (bool)1>>>>(T1::Params)",
      "gridX": 132,
      "gridY": 1,
      "gridZ": 1,
      "blockX": 384,
      "blockY": 1,
      "blockZ": 1,
      "registersPerThread": 168,
      "staticSharedMemory": 0,
      "dynamicSharedMemory": 199680,
      "calls": 800,
      "first_event_id": 870927,
      "clipped_summed_gpu_ms": 792.182782
    },
    {
      "physical_gpu": 1,
      "name": "void cutlass::device_kernel<flash::enable_sm90_or_later<flash::FlashAttnFwdSm90<flash::CollectiveMainloopFwdSm90<(int)2, cute::tuple<cute::C<(int)1>, cute::C<(int)1>, cute::C<(int)1>>, cute::tuple<cute::C<(int)128>, cute::C<(int)160>, cute::C<(int)128>>, (int)128, cutlass::bfloat16_t, float, cutlass::arch::Sm90, (bool)0, (bool)0, (bool)0, (bool)1, (bool)0, (bool)0, (bool)0, (bool)1, (bool)1, (bool)1, (bool)0, (bool)0, cutlass::bfloat16_t, (int)1>, flash::CollectiveEpilogueFwd<cute::tuple<cute::C<(int)128>, cute::C<(int)128>, cute::C<(int)160>>, cute::tuple<cute::C<(int)1>, cute::C<(int)1>, cute::C<(int)1>>, cutlass::bfloat16_t, cutlass::arch::Sm90, (int)256, (bool)1, (bool)1, (bool)0, (bool)0, (int)1>, flash::VarlenDynamicPersistentTileScheduler<(int)128, (int)160, (int)256, (int)128, (bool)0, (bool)1, (bool)1, (bool)0, (bool)0, (bool)1>>>>(T1::Params)",
      "gridX": 132,
      "gridY": 1,
      "gridZ": 1,
      "blockX": 384,
      "blockY": 1,
      "blockZ": 1,
      "registersPerThread": 168,
      "staticSharedMemory": 0,
      "dynamicSharedMemory": 199680,
      "calls": 800,
      "first_event_id": 870917,
      "clipped_summed_gpu_ms": 779.694468
    },
    {
      "physical_gpu": 0,
      "name": "void cutlass::device_kernel<flash::enable_sm90_or_later<flash::FlashAttnFwdSm90<flash::CollectiveMainloopFwdSm90<(int)2, cute::tuple<cute::C<(int)1>, cute::C<(int)1>, cute::C<(int)1>>, cute::tuple<cute::C<(int)128>, cute::C<(int)160>, cute::C<(int)128>>, (int)128, cutlass::bfloat16_t, float, cutlass::arch::Sm90, (bool)0, (bool)0, (bool)0, (bool)1, (bool)0, (bool)0, (bool)0, (bool)1, (bool)1, (bool)1, (bool)0, (bool)0, cutlass::bfloat16_t, (int)1>, flash::CollectiveEpilogueFwd<cute::tuple<cute::C<(int)128>, cute::C<(int)128>, cute::C<(int)160>>, cute::tuple<cute::C<(int)1>, cute::C<(int)1>, cute::C<(int)1>>, cutlass::bfloat16_t, cutlass::arch::Sm90, (int)256, (bool)1, (bool)1, (bool)0, (bool)0, (int)1>, flash::VarlenDynamicPersistentTileScheduler<(int)128, (int)160, (int)256, (int)128, (bool)0, (bool)1, (bool)1, (bool)0, (bool)0, (bool)1>>>>(T1::Params)",
      "gridX": 132,
      "gridY": 1,
      "gridZ": 1,
      "blockX": 384,
      "blockY": 1,
      "blockZ": 1,
      "registersPerThread": 168,
      "staticSharedMemory": 0,
      "dynamicSharedMemory": 199680,
      "calls": 800,
      "first_event_id": 870919,
      "clipped_summed_gpu_ms": 777.108374
    },
    {
      "physical_gpu": 0,
      "name": "ncclDevKernel_SendRecv(ncclDevKernelArgsStorage<(unsigned long)4096>)",
      "gridX": 32,
      "gridY": 1,
      "gridZ": 1,
      "blockX": 640,
      "blockY": 1,
      "blockZ": 1,
      "registersPerThread": 96,
      "staticSharedMemory": 36800,
      "dynamicSharedMemory": 82240,
      "calls": 3200,
      "first_event_id": 870889,
      "clipped_summed_gpu_ms": 299.711619
    },
    {
      "physical_gpu": 1,
      "name": "ncclDevKernel_SendRecv(ncclDevKernelArgsStorage<(unsigned long)4096>)",
      "gridX": 32,
      "gridY": 1,
      "gridZ": 1,
      "blockX": 640,
      "blockY": 1,
      "blockZ": 1,
      "registersPerThread": 96,
      "staticSharedMemory": 36800,
      "dynamicSharedMemory": 82240,
      "calls": 3200,
      "first_event_id": 870891,
      "clipped_summed_gpu_ms": 294.402449
    }
  ],
  "signature_counts_by_name": {
    "void at::native::elementwise_kernel<(int)128, (int)2, void at::native::gpu_kernel_impl_nocast<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 3)]::operator ()() const::[lambda() (instance 7)]::operator ()() const::[lambda(float) (instance 1)]>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3)": 30,
    "void at::native::vectorized_elementwise_kernel<(int)4, at::native::FillFunctor<float>, std::array<char *, (unsigned long)1>>(int, T2, T3)": 18,
    "void cudnn::engines_precompiled::nchwToNhwcKernel<float, float, float, (bool)0, (bool)1, (cudnnKernelDataType_t)0>(cudnn::engines_precompiled::nchw2nhwc_params_t<T3>, const T1 *, T2 *)": 17,
    "void at::native::elementwise_kernel<(int)128, (int)2, void at::native::gpu_kernel_impl_nocast<at::native::CUDAFunctor_add<float>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3)": 11,
    "void cudnn::engines_precompiled::nhwcToNchwKernel<float, float, float, (bool)1, (bool)0, (cudnnKernelDataType_t)0>(cudnn::engines_precompiled::nhwc2nchw_params_t<T3>, const T1 *, T2 *)": 10,
    "void at::native::vectorized_elementwise_kernel<(int)4, at::native::AUnaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float>>, std::array<char *, (unsigned long)2>>(int, T2, T3)": 8,
    "void at::native::elementwise_kernel<(int)128, (int)2, void at::native::gpu_kernel_impl_nocast<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::DivFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3)": 7,
    "void at::native::vectorized_elementwise_kernel<(int)4, at::native::CUDAFunctorOnSelf_add<float>, std::array<char *, (unsigned long)2>>(int, T2, T3)": 7,
    "void at::native::elementwise_kernel<(int)128, (int)2, void at::native::gpu_kernel_impl_nocast<at::native::BinaryFunctor<float, float, float, at::native::binary_internal::MulFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3)": 7,
    "void at::native::vectorized_elementwise_kernel<(int)4, at::native::CUDAFunctor_add<float>, std::array<char *, (unsigned long)3>>(int, T2, T3)": 6,
    "void at::native::vectorized_elementwise_kernel<(int)4, at::native::<unnamed>::launch_clamp_scalar(at::TensorIteratorBase &, c10::Scalar, c10::Scalar, at::native::detail::ClampLimits)::[lambda() (instance 1)]::operator ()() const::[lambda() (instance 7)]::operator ()() const::[lambda(float) (instance 1)], std::array<char *, (unsigned long)2>>(int, T2, T3)": 6,
    "void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 3)]::operator ()() const::[lambda() (instance 12)]::operator ()() const::[lambda(c10::BFloat16) (instance 1)]>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3)": 5,
    "ncclDevKernel_SendRecv(ncclDevKernelArgsStorage<(unsigned long)4096>)": 5,
    "void at::native::reduce_kernel<(int)128, (int)4, at::native::ReduceOp<float, at::native::NormTwoOps<float, float, float, (bool)1>, unsigned int, float, (int)4, (int)4>>(T3)": 5,
    "void at::native::vectorized_elementwise_kernel<(int)4, at::native::<unnamed>::silu_kernel(at::TensorIteratorBase &)::[lambda() (instance 1)]::operator ()() const::[lambda() (instance 2)]::operator ()() const::[lambda(float) (instance 1)], std::array<char *, (unsigned long)2>>(int, T2, T3)": 5,
    "void cudnn::engines_precompiled::nchwToNhwcKernel<float, float, float, (bool)0, (bool)1, (cudnnKernelDataType_t)2>(cudnn::engines_precompiled::nchw2nhwc_params_t<T3>, const T1 *, T2 *)": 4,
    "void at::native::vectorized_elementwise_kernel<(int)8, at::native::bfloat16_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda(float) (instance 1)], std::array<char *, (unsigned long)2>>(int, T2, T3)": 3,
    "triton_red_fused__to_copy_div_mul_pow_split_with_sizes_sqrt_sum_view_3": 3,
    "triton_poi_fused__to_copy_add_mul_split_squeeze_view_14": 3,
    "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 3)]::operator ()() const::[lambda() (instance 7)]::operator ()() const::[lambda(float) (instance 1)], std::array<char *, (unsigned long)2>, (int)4, TrivialOffsetCalculator<(int)1, unsigned int>, TrivialOffsetCalculator<(int)1, unsigned int>, at::native::memory::LoadWithCast<(int)1>, at::native::memory::StoreWithCast<(int)1>>(int, T1, T2, T4, T5, T6, T7)": 3,
    "sm90_xmma_fprop_implicit_gemm_f32f32_tf32f32_f32_nhwckrsc_nhwc_tilesize128x128x32_warpgroupsize1x1x1_g1_execute_segment_k_off_kernel__5x_cudnn": 3,
    "void at::native::<unnamed>::upsample_nearest2d_out_frame<float, &at::native::nearest_neighbor_exact_compute_source_index>(const T1 *, T1 *, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, float, float)": 3,
    "sm80_xmma_gemm_f32f32_f32f32_f32_nn_n_tilesize32x32x8_stage3_warpsize1x2x1_ffma_aligna4_alignc4_execute_kernel__5x_cublas": 2,
    "void at::native::reduce_kernel<(int)512, (int)1, at::native::ReduceOp<float, at::native::NormTwoOps<float, float, float, (bool)1>, unsigned int, float, (int)4, (int)4>>(T3)": 2,
    "void <unnamed>::elementwise_kernel_with_index<int, at::native::arange_cuda_out(const c10::Scalar &, const c10::Scalar &, const c10::Scalar &, at::Tensor &)::[lambda() (instance 1)]::operator ()() const::[lambda() (instance 7)]::operator ()() const::[lambda(long) (instance 1)]>(T1, T2, function_traits<T2>::result_type *)": 2,
    "void at::native::<unnamed>::CatArrayBatchedCopy_alignedK_contig<at::native::<unnamed>::OpaqueType<(unsigned int)4>, unsigned int, (int)4, (int)128, (int)1, (int)16>(T1 *, at::native::<unnamed>::CatArrInputTensorMetadata<T1, T2, T4, T5>, at::native::<unnamed>::TensorSizeStride<T2, (unsigned int)4>, int, T2)": 2,
    "void at::native::<unnamed>::CatArrayBatchedCopy<at::native::<unnamed>::OpaqueType<(unsigned int)4>, unsigned int, (int)4, (int)64, (int)64>(T1 *, at::native::<unnamed>::CatArrInputTensorMetadata<T1, T2, T4, T5>, at::native::<unnamed>::TensorSizeStride<T2, (unsigned int)4>, int, T2)": 2,
    "void at::native::vectorized_elementwise_kernel<(int)8, at::native::<unnamed>::silu_kernel(at::TensorIteratorBase &)::[lambda() (instance 1)]::operator ()() const::[lambda() (instance 6)]::operator ()() const::[lambda(c10::BFloat16) (instance 1)], std::array<char *, (unsigned long)2>>(int, T2, T3)": 2,
    "triton_red_fused__to_copy_add_mul_native_layer_norm_split_squeeze_1": 2,
    "triton_red_fused__to_copy_div_mul_pow_split_with_sizes_sqrt_sum_view_2": 2,
    "triton_poi_fused_silu_view_6": 2,
    "triton_poi_fused_add_view_7": 2,
    "triton_red_fused__to_copy_add_mul_native_layer_norm_split_squeeze_view_9": 2,
    "ncclDevKernel_AllGather_RING_LL(ncclDevKernelArgsStorage<(unsigned long)4096>)": 2,
    "void cutlass::Kernel2<cutlass_80_tensorop_s1688gemm_128x64_16x6_nn_align4>(T1::Params)": 2,
    "sm90_xmma_fprop_implicit_gemm_f32f32_tf32f32_f32_nhwckrsc_nhwc_tilesize128x128x32_warpgroupsize1x1x1_g1_execute_segment_k_on_kernel__5x_cudnn": 2,
    "void at::native::vectorized_elementwise_kernel<(int)4, at::native::AbsFunctor<float>, std::array<char *, (unsigned long)2>>(int, T2, T3)": 1,
    "void at::native::vectorized_elementwise_kernel<(int)8, at::native::AUnaryFunctor<float, float, bool, at::native::<unnamed>::CompareEqFunctor<float>>, std::array<char *, (unsigned long)2>>(int, T2, T3)": 1,
    "void at::native::vectorized_elementwise_kernel<(int)8, at::native::BinaryFunctor<float, float, bool, at::native::<unnamed>::CompareEqFunctor<float>>, std::array<char *, (unsigned long)3>>(int, T2, T3)": 1,
    "void at::native::vectorized_elementwise_kernel<(int)8, at::native::BinaryFunctor<bool, bool, bool, at::native::binary_internal::MulFunctor<bool>>, std::array<char *, (unsigned long)3>>(int, T2, T3)": 1,
    "void at::native::reduce_kernel<(int)512, (int)1, at::native::ReduceOp<bool, at::native::func_wrapper_t<bool, at::native::and_kernel_cuda(at::TensorIterator &)::[lambda() (instance 1)]::operator ()() const::[lambda() (instance 12)]::operator ()() const::[lambda(bool, bool) (instance 1)]>, unsigned int, bool, (int)4, (int)4>>(T3)": 1,
    "void at::native::elementwise_kernel<(int)128, (int)2, void at::native::gpu_kernel_impl_nocast<at::native::FillFunctor<float>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3)": 1,
    "std::enable_if<!T7, void>::type internal::gemvx::kernel<int, int, float, float, float, float, (bool)0, (bool)0, (bool)0, (bool)0, (int)6, (bool)0, cublasGemvParamsEx<int, cublasGemvTensorStridedBatched<const float>, cublasGemvTensorStridedBatched<const float>, cublasGemvTensorStridedBatched<float>, float>>(T13)": 1,
    "void at::native::vectorized_elementwise_kernel<(int)4, at::native::neg_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 2)]::operator ()() const::[lambda() (instance 7)]::operator ()() const::[lambda(float) (instance 1)], std::array<char *, (unsigned long)2>>(int, T2, T3)": 1,
    "void gemvNSP_kernel<float, float, float, float, (int)11, (int)32, (int)4, (int)1024, (bool)0, cublasGemvParamsEx<int, cublasGemvTensorStridedBatched<const float>, cublasGemvTensorStridedBatched<const float>, cublasGemvTensorStridedBatched<float>, float>>(T10)": 1,
    "void at::native::reduce_kernel<(int)512, (int)1, at::native::ReduceOp<float, at::native::func_wrapper_t<float, at::native::MaxNanFunctor<float>>, unsigned int, float, (int)4, (int)4>>(T3)": 1,
    "void at::native::vectorized_elementwise_kernel<(int)8, void at::native::compare_scalar_kernel<float>(at::TensorIteratorBase &, at::native::<unnamed>::OpType, T1)::[lambda(float) (instance 1)], std::array<char *, (unsigned long)2>>(int, T2, T3)": 1,
    "void at::native::index_elementwise_kernel<(int)128, (int)4, void at::native::gpu_index_kernel<void at::native::index_kernel_impl<at::native::OpaqueType<(int)4>>(at::TensorIteratorBase &, c10::ArrayRef<long>, c10::ArrayRef<long>)::[lambda(char *, const char *, long) (instance 1)]>(at::TensorIteratorBase &, c10::ArrayRef<long>, c10::ArrayRef<long>, const T1 &, bool)::[lambda(int) (instance 1)]>(long, T3)": 1,
    "void at::native::index_elementwise_kernel<(int)128, (int)4, void at::native::gpu_index_kernel<void at::native::index_put_kernel_impl<at::native::OpaqueType<(int)4>>(at::TensorIterator &, c10::ArrayRef<long>, c10::ArrayRef<long>)::[lambda(char *, const char *, long) (instance 1)]>(at::TensorIteratorBase &, c10::ArrayRef<long>, c10::ArrayRef<long>, const T1 &, bool)::[lambda(int) (instance 1)]>(long, T3)": 1,
    "void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl_nocast<at::native::AUnaryFunctor<float, float, bool, at::native::<unnamed>::CompareEqFunctor<float>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3)": 1,
    "void at::native::reduce_kernel<(int)512, (int)1, at::native::ReduceOp<bool, at::native::func_wrapper_t<bool, at::native::or_kernel_cuda(at::TensorIterator &)::[lambda() (instance 1)]::operator ()() const::[lambda() (instance 12)]::operator ()() const::[lambda(bool, bool) (instance 1)]>, unsigned int, bool, (int)4, (int)4>>(T3)": 1,
    "sm80_xmma_gemm_f32f32_f32f32_f32_tn_n_tilesize128x32x8_stage3_warpsize2x2x1_ffma_aligna4_alignc4_execute_kernel__5x_cublas": 1,
    "void at::native::<unnamed>::distribution_elementwise_grid_stride_kernel<float, (int)4, void at::native::templates::cuda::normal_and_transform<float, float, at::CUDAGeneratorImpl *, void at::native::templates::cuda::normal_kernel<at::CUDAGeneratorImpl *>(const at::TensorBase &, double, double, T1)::[lambda() (instance 1)]::operator ()() const::[lambda() (instance 2)]::operator ()() const::[lambda(float) (instance 1)]>(at::TensorIteratorBase &, T3, T4)::[lambda(curandStatePhilox4_32_10 *) (instance 2)], void at::native::<unnamed>::distribution_nullary_kernel<float, float, float4, at::CUDAGeneratorImpl *, void at::native::templates::cuda::normal_and_transform<float, float, at::CUDAGeneratorImpl *, void at::native::templates::cuda::normal_kernel<at::CUDAGeneratorImpl *>(const at::TensorBase &, double, double, T1)::[lambda() (instance 1)]::operator ()() const::[lambda() (instance 2)]::operator ()() const::[lambda(float) (instance 1)]>(at::TensorIteratorBase &, T3, T4)::[lambda(curandStatePhilox4_32_10 *) (instance 2)], void at::native::templates::cuda::normal_kernel<at::CUDAGeneratorImpl *>(const at::TensorBase &, double, double, T1)::[lambda() (instance 1)]::operator ()() const::[lambda() (instance 2)]::operator ()() const::[lambda(float) (instance 1)]>(at::TensorIteratorBase &, T4, const T5 &, T6)::[lambda(int, float) (instance 1)]>(long, at::PhiloxCudaState, T3, T4)": 1,
    "nvjet_sm90_tst_256x144_64x4_2x1_v_bz_coopA_bias_TNT": 1,
    "void at::native::vectorized_elementwise_kernel<(int)8, at::native::CUDAFunctor_add<c10::BFloat16>, std::array<char *, (unsigned long)3>>(int, T2, T3)": 1,
    "void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 3)]::operator ()() const::[lambda() (instance 6)]::operator ()() const::[lambda(double) (instance 1)]>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3)": 1,
    "void <unnamed>::elementwise_kernel_with_index<int, at::native::arange_cuda_out(const c10::Scalar &, const c10::Scalar &, const c10::Scalar &, at::Tensor &)::[lambda() (instance 1)]::operator ()() const::[lambda() (instance 6)]::operator ()() const::[lambda(long) (instance 1)]>(T1, T2, function_traits<T2>::result_type *)": 1,
    "void at::native::vectorized_elementwise_kernel<(int)2, at::native::neg_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 2)]::operator ()() const::[lambda() (instance 6)]::operator ()() const::[lambda(double) (instance 1)], std::array<char *, (unsigned long)2>>(int, T2, T3)": 1,
    "void at::native::vectorized_elementwise_kernel<(int)2, at::native::BUnaryFunctor<double, double, double, at::native::binary_internal::MulFunctor<double>>, std::array<char *, (unsigned long)2>>(int, T2, T3)": 1,
    "void at::native::vectorized_elementwise_kernel<(int)2, at::native::FillFunctor<long>, std::array<char *, (unsigned long)1>>(int, T2, T3)": 1,
    "void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl<at::native::<unnamed>::pow_tensor_tensor_kernel(at::TensorIteratorBase &)::[lambda() (instance 1)]::operator ()() const::[lambda() (instance 6)]::operator ()() const::[lambda(double, double) (instance 1)]>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3)": 1,
    "void at::native::elementwise_kernel<(int)128, (int)2, void at::native::gpu_kernel_impl_nocast<at::native::BinaryFunctor<double, double, double, at::native::binary_internal::MulFunctor<double>>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3)": 1,
    "void at::native::vectorized_elementwise_kernel<(int)2, at::native::cos_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 2)]::operator ()() const::[lambda() (instance 1)]::operator ()() const::[lambda(double) (instance 1)], std::array<char *, (unsigned long)2>>(int, T2, T3)": 1,
    "void at::native::vectorized_elementwise_kernel<(int)2, at::native::sin_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 2)]::operator ()() const::[lambda() (instance 1)]::operator ()() const::[lambda(double) (instance 1)], std::array<char *, (unsigned long)2>>(int, T2, T3)": 1,
    "void at::native::<unnamed>::CatArrayBatchedCopy_vectorized<at::native::<unnamed>::OpaqueType<(unsigned int)8>, unsigned int, (int)2, (int)128, (int)1, (int)16, (int)2>(char *, at::native::<unnamed>::CatArrInputTensorMetadata<T1, T2, T4, T5>, at::native::<unnamed>::TensorSizeStride<T2, (unsigned int)4>, int, T2)": 1,
    "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 3)]::operator ()() const::[lambda() (instance 12)]::operator ()() const::[lambda(c10::BFloat16) (instance 1)], std::array<char *, (unsigned long)2>, (int)4, TrivialOffsetCalculator<(int)1, unsigned int>, TrivialOffsetCalculator<(int)1, unsigned int>, at::native::memory::LoadWithCast<(int)1>, at::native::memory::StoreWithCast<(int)1>>(int, T1, T2, T4, T5, T6, T7)": 1,
    "nvjet_sm90_tst_64x8_64x16_4x1_v_bz_bias_TNT": 1,
    "nvjet_sm90_tst_64x8_64x16_2x1_v_bz_bias_TNT": 1,
    "nvjet_sm90_tst_256x152_64x4_1x2_h_bz_coopA_bias_TNT": 1,
    "rotary_kernel": 1,
    "triton_poi_fused_all_to_all_single_clone_split_with_sizes_stack_transpose_view_4": 1,
    "void at::native::index_elementwise_kernel<(int)128, (int)4, void at::native::gpu_index_kernel<void at::native::index_put_kernel_impl<at::native::OpaqueType<(int)2>>(at::TensorIterator &, c10::ArrayRef<long>, c10::ArrayRef<long>)::[lambda(char *, const char *, long) (instance 1)]>(at::TensorIteratorBase &, c10::ArrayRef<long>, c10::ArrayRef<long>, const T1 &, bool)::[lambda(int) (instance 1)]>(long, T3)": 1,
    "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 3)]::operator ()() const::[lambda() (instance 4)]::operator ()() const::[lambda(long) (instance 1)], std::array<char *, (unsigned long)2>, (int)4, TrivialOffsetCalculator<(int)1, unsigned int>, TrivialOffsetCalculator<(int)1, unsigned int>, at::native::memory::LoadWithCast<(int)1>, at::native::memory::StoreWithCast<(int)1>>(int, T1, T2, T4, T5, T6, T7)": 1,
    "void at::native::vectorized_gather_kernel<(int)16, long>(char *, char *, T2 *, int, long, long, long, long, bool)": 1,
    "void at::native::vectorized_elementwise_kernel<(int)4, at::native::FillFunctor<int>, std::array<char *, (unsigned long)1>>(int, T2, T3)": 1,
    "void at::native::<unnamed>::CatArrayBatchedCopy_alignedK_contig<at::native::<unnamed>::OpaqueType<(unsigned int)4>, unsigned int, (int)1, (int)128, (int)1, (int)16>(T1 *, at::native::<unnamed>::CatArrInputTensorMetadata<T1, T2, T4, T5>, at::native::<unnamed>::TensorSizeStride<T2, (unsigned int)4>, int, T2)": 1,
    "void flash::prepare_varlen_num_blocks_kernel<(int)1, (bool)0>(int, int, int, const int *, const int *, const int *, const int *, const int *, const int *, int, int, int, int, int, cutlass::FastDivmod, cutlass::FastDivmod, int *, int *, int *, int *, int *, bool, bool, bool, int)": 1,
    "void cutlass::device_kernel<flash::enable_sm90_or_later<flash::FlashAttnFwdSm90<flash::CollectiveMainloopFwdSm90<(int)2, cute::tuple<cute::C<(int)1>, cute::C<(int)1>, cute::C<(int)1>>, cute::tuple<cute::C<(int)128>, cute::C<(int)160>, cute::C<(int)128>>, (int)128, cutlass::bfloat16_t, float, cutlass::arch::Sm90, (bool)0, (bool)0, (bool)0, (bool)1, (bool)0, (bool)0, (bool)0, (bool)1, (bool)1, (bool)1, (bool)0, (bool)0, cutlass::bfloat16_t, (int)1>, flash::CollectiveEpilogueFwd<cute::tuple<cute::C<(int)128>, cute::C<(int)128>, cute::C<(int)160>>, cute::tuple<cute::C<(int)1>, cute::C<(int)1>, cute::C<(int)1>>, cutlass::bfloat16_t, cutlass::arch::Sm90, (int)256, (bool)1, (bool)1, (bool)0, (bool)0, (int)1>, flash::VarlenDynamicPersistentTileScheduler<(int)128, (int)160, (int)256, (int)128, (bool)0, (bool)1, (bool)1, (bool)0, (bool)0, (bool)1>>>>(T1::Params)": 1,
    "triton_poi_fused_all_to_all_single_clone_transpose_unsqueeze_view_5": 1,
    "nvjet_sm90_tst_128x200_64x5_2x1_v_bz_coopA_bias_TNT": 1,
    "triton_poi_fused_clone_transpose_view_8": 1,
    "triton_red_fused__to_copy_pow_sum_view_10": 1,
    "triton_poi_fused__to_copy_all_to_all_single_clone_div_mul_sqrt_transpose_view_11": 1,
    "void cutlass::device_kernel<flash::enable_sm90_or_later<flash::FlashAttnFwdSm90<flash::CollectiveMainloopFwdSm90<(int)2, cute::tuple<cute::C<(int)1>, cute::C<(int)1>, cute::C<(int)1>>, cute::tuple<cute::C<(int)128>, cute::C<(int)176>, cute::C<(int)128>>, (int)128, cutlass::bfloat16_t, float, cutlass::arch::Sm90, (bool)0, (bool)0, (bool)0, (bool)0, (bool)0, (bool)0, (bool)0, (bool)1, (bool)1, (bool)0, (bool)0, (bool)0>, flash::CollectiveEpilogueFwd<cute::tuple<cute::C<(int)128>, cute::C<(int)128>, cute::C<(int)176>>, cute::tuple<cute::C<(int)1>, cute::C<(int)1>, cute::C<(int)1>>, cutlass::bfloat16_t, cutlass::arch::Sm90, (int)256, (bool)0, (bool)0, (bool)0, (bool)0>, flash::StaticPersistentTileScheduler<(bool)0>>>>(T1::Params)": 1,
    "triton_red_fused__to_copy_add_mul_native_layer_norm_split_squeeze_view_12": 1,
    "triton_poi_fused_gelu_view_13": 1,
    "void at::native::elementwise_kernel<(int)128, (int)4, void at::native::gpu_kernel_impl<at::native::CUDAFunctor_add<float>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3)": 1,
    "void at::native::<unnamed>::vectorized_layer_norm_kernel<float, float, (bool)0>(int, T2, const T1 *, const T1 *, const T1 *, T2 *, T2 *, T1 *)": 1,
    "void at::native::elementwise_kernel<(int)128, (int)2, void at::native::gpu_kernel_impl_nocast<at::native::CUDAFunctorOnSelf_add<float>>(at::TensorIteratorBase &, const T1 &)::[lambda(int) (instance 1)]>(int, T3)": 1,
    "nvjet_sm90_tst_64x48_64x15_1x4_h_bz_bias_TNT": 1,
    "ncclDevKernel_AllReduce_Sum_f32_RING_LL(ncclDevKernelArgsStorage<(unsigned long)4096>)": 1,
    "sm80_xmma_fprop_implicit_gemm_tf32f32_tf32f32_f32_nhwckrsc_nchw_tilesize128x64x32_stage5_warpsize2x2x1_g1_tensor16x8x8_execute_kernel__5x_cudnn": 1,
    "sm90_xmma_fprop_implicit_gemm_f32f32_tf32f32_f32_nhwckrsc_nhwc_tilesize64x64x32_warpgroupsize1x1x1_g1_execute_segment_k_off_kernel__5x_cudnn": 1,
    "void cutlass::Kernel2<cutlass_80_tensorop_s1688gemm_128x128_32x3_nn_align4>(T1::Params)": 1,
    "fmha_cutlassF_f32_aligned_32x128_gmem_sm80(PyTorchMemEffAttention::AttentionKernel<float, cutlass::arch::Sm80, (bool)1, (int)32, (int)128, (int)65536, (bool)1, (bool)1>::Params)": 1,
    "void cutlass::Kernel2<cutlass_80_tensorop_s1688gemm_64x128_32x3_nn_align4>(T1::Params)": 1,
    "void cask_plugin__5x_cudnn::xmma__5x_cudnn::init_device_workspace_kernel<xmma__5x_cudnn::implicit_gemm::fprop::Warp_specialized_params_non_template<xmma__5x_cudnn::Grid_constant_params>>(T1, bool)": 1,
    "sm90_xmma_fprop_implicit_gemm_f32f32_tf32f32_f32_nhwckrsc_nhwc_tilesize64x192x32_warpgroupsize1x1x1_g1_execute_segment_k_off_kernel__5x_cudnn": 1,
    "sm90_xmma_fprop_implicit_gemm_f32f32_tf32f32_f32_nhwckrsc_nhwc_tilesize256x32x32_warpgroupsize1x1x1_g1_execute_segment_k_off_kernel__5x_cudnn": 1,
    "sm80_xmma_fprop_implicit_gemm_indexed_wo_smem_tf32f32_tf32f32_f32_nhwckrsc_nhwc_tilesize128x16x64_stage1_warpsize4x1x1_g1_tensor16x8x8_alignc4_execute_kernel__5x_cudnn": 1,
    "void at::native::vectorized_elementwise_kernel<(int)4, at::native::round_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 1)]::operator ()() const::[lambda() (instance 2)]::operator ()() const::[lambda(float) (instance 1)], std::array<char *, (unsigned long)2>>(int, T2, T3)": 1,
    "void at::native::unrolled_elementwise_kernel<at::native::direct_copy_kernel_cuda(at::TensorIteratorBase &)::[lambda() (instance 3)]::operator ()() const::[lambda() (instance 1)]::operator ()() const::[lambda(unsigned char) (instance 1)], std::array<char *, (unsigned long)2>, (int)4, TrivialOffsetCalculator<(int)1, unsigned int>, TrivialOffsetCalculator<(int)1, unsigned int>, at::native::memory::LoadWithCast<(int)1>, at::native::memory::StoreWithCast<(int)1>>(int, T1, T2, T4, T5, T6, T7)": 1
  }
}