The AI CUDA Engineer 👷 Archive

Kernel Leaderboard 🏆

Level	Task	Rank	Task & Kernel Name	Runtime (ms)	Speedup Native	Speedup Compile
2	23	🥇	23_Conv3d_GroupNorm_Mean fused_ops_strided_optimized_base	0.01	128.51	82.34
2	13	🥇	13_ConvTranspose3d_Mean_Add_Softmax_Tanh_Scaling 13_convtranspose3d_mean_add_softmax_tanh_scaling_optimized_edit_1	0.01	84.55	66.82
1	12	🥇	12_Matmul_with_diagonal_matrices_ stride_loop_diag_matmul_base	0.05	54.40	55.46
2	18	🥇	18_Matmul_Sum_Max_AvgPool_LogSumExp_LogSumExp vectorized_inner_loop_warp_shuffle_base	0.01	14.42	2.64
1	95	🥇	95_CrossEntropyLoss ce_loss_grid_stride_unroll_base	0.01	8.97	2.45
2	42	🥇	42_ConvTranspose2d_GlobalAvgPool_BiasAdd_LogSumExp_Sum_Multiply optimized_fused_conv_lse_min_divergence_edit_1	0.01	8.66	7.99
1	40	🥇	40_LayerNorm optimized_layernorm_streamed_base	0.94	8.60	0.70
1	97	🥇	97_CosineSimilarityLoss blocksize_tuning_cosine_loss_base	0.01	7.64	5.47
3	3	🥇	3_DeepNarrowMLP uniform_control_flow_optimization_edit_1	0.05	7.46	5.98
2	66	🥇	66_Matmul_Dropout_Mean_Softmax warp_divergence_optimized_matmul_base	0.01	6.47	11.33
2	45	🥇	45_Gemm_Sigmoid_Sum_LogSumExp fused_gemm_sigmoid_logsumexp_base	0.01	5.98	2.74
2	95	🥇	95_Matmul_Add_Swish_Tanh_GELU_Hardtanh warp_level_vec_ldg_opt_edit_1	0.01	5.91	8.85
1	88	🥇	88_MinGPTNewGelu optimized_gelu_combined_edit_1	0.02	5.72	2.99
3	49	🥇	49_Mamba2ReturnFinalState 49_mamba2returnfinalstate_shared_balanced_base	0.07	4.75	0.81
2	9	🥇	9_Matmul_Subtract_Multiply_ReLU tiled_grid_stride_base	0.01	4.05	2.63
2	40	🥇	40_Matmul_Scaling_ResidualAdd warp_level_matmul_base	0.01	3.92	3.10
1	99	🥇	99_TripletMarginLoss vectorized_warp_shfl_reduction_base	0.01	3.87	2.62
2	55	🥇	55_Matmul_MaxPool_Sum_Scale tiled_matmul_pool_sync_opt_base	0.01	3.73	2.23
2	65	🥇	65_Conv2d_AvgPool_Sigmoid_Sum block512_conv_pool_sigsum_base	0.01	3.67	3.78
2	29	🥇	29_Matmul_Mish_Mish optimized_tiled_matmul_mish_base	0.01	3.65	9.33
2	28	🥇	28_BMM_InstanceNorm_Sum_ResidualAdd_Multiply 28_bmm_instancenorm_ldg_aligned_base	0.01	3.50	2.39
2	56	🥇	56_Matmul_Sigmoid_Sum block_size_tuning_base_base	0.01	3.31	2.40
1	90	🥇	90_cumprod cumprod_min_sync_base_base	0.01	3.09	3.13
2	68	🥇	68_Matmul_Min_Subtract stride_loop_optimization_thread_base_base	0.01	2.95	1.92
3	34	🥇	34_VanillaRNNHidden stride_rnn_hidden_warp_base	0.01	2.91	6.68
1	98	🥇	98_KLDivLoss optimized_kl_div_cuda_base	0.01	2.83	3.20
1	36	🥇	36_RMSNorm_ 36_rmsnorm_even_workload_base	0.19	2.81	2.68
2	99	🥇	99_Matmul_GELU_Softmax warp_divergence_minimized_kernel_base	0.01	2.78	2.26
1	50	🥇	50_Product_reduction_over_a_dimension block_size_optimized_reduction_base	0.01	2.69	5.41
2	14	🥇	14_Gemm_Divide_Sum_Scaling balanced_workload_gemm_base	0.01	2.52	6.31
1	30	🥇	30_Softsign shared_mem_warp_opt_base_base	0.01	2.47	4.72
3	4	🥇	4_LeNet5 4_LeNet5_fused_even_edit_1	0.05	2.38	1.47
3	24	🥇	24_EfficientNetB2 efficientnetb2_warp_shuffle_optimization_base_base	0.63	2.37	1.63
2	20	🥇	20_ConvTranspose3d_Sum_ResidualAdd_Multiply_ResidualAdd coalesced_vectorized_fused_kernel_base	1.52	2.36	0.73
3	2	🥇	2_ShallowWideMLP min_sync_warp_base	0.04	2.29	4.36
2	97	🥇	97_Matmul_BatchNorm_BiasAdd_Divide_Swish block_experiment_fused_bn_swish_base	0.03	2.28	1.94
3	18	🥇	18_SqueezeNet 18_squeezenet_shared_memory_reduction_base	0.52	2.26	0.59
1	89	🥇	89_cumsum hybrid_aligned_cumsum_base	0.01	2.21	2.14
1	74	🥇	74_conv_transposed_1D_dilated shared_mem_tiled_74_conv_transposed_1D_dilated_edit_1	0.01	2.20	2.79
2	81	🥇	81_Gemm_Swish_Divide_Clamp_Tanh_Clamp fused_activation_base	0.02	2.19	1.90
2	22	🥇	22_Matmul_Scale_ResidualAdd_Clamp_LogSumExp_Mish 22_matmul_scale_residualadd_clamp_logsumexp_mish_syncthreads_optimized_base	0.03	2.19	1.49
1	53	🥇	53_Min_reduction_over_a_dimension min_reduce_fused_warp_base	0.01	2.19	3.09
1	38	🥇	38_L1Norm_ l1norm_shared_memory_optimization_edit_1	0.01	2.04	4.17
2	33	🥇	33_Gemm_Scale_BatchNorm fused_scale_bn_coalesced_base	0.03	2.02	1.00
2	91	🥇	91_ConvTranspose2d_Softmax_BiasAdd_Scaling_Sigmoid optimized_fused_ops_kernel_minimized_warp_divergence_edit_1	0.15	2.01	0.68
2	82	🥇	82_Conv2d_Tanh_Scaling_BiasAdd_Max fused_conv_pool_base	0.03	2.00	2.05
3	46	🥇	46_NetVladWithGhostClusters netvlad_warp_shfl_sync_optimized_base	0.10	1.99	0.78
1	46	🥇	46_Average_Pooling_3D avgpool3d_combo_edit_1	0.29	1.96	3.44
2	88	🥇	88_Gemm_GroupNorm_Swish_Multiply_Swish minimal_sync_88_gemm_groupnorm_swish_base	0.02	1.95	1.98
1	45	🥇	45_Average_Pooling_2D modular_avg_pool2d_base_base	0.11	1.94	3.03
2	35	🥇	35_Conv2d_Subtract_HardSwish_MaxPool_Mish manually_unrolled_kernel_base	0.03	1.93	2.16
3	43	🥇	43_MinGPTCausalAttention coalesced_causal_attention_base_base	6.68	1.93	1.51
2	59	🥇	59_Matmul_Swish_Scaling 59_matmul_swish_scaling_coalesced_base	0.02	1.91	1.94
1	43	🥇	43_Max_Pooling_3D maxpool3d_unrolled_base_base	0.25	1.91	3.59
2	58	🥇	58_ConvTranspose3d_LogSumExp_HardSwish_Subtract_Clamp_Max fused_optimized_kernel_base	4.49	1.90	0.83
1	39	🥇	39_L2Norm_ l2norm_strided_optimized_base_base	0.01	1.89	5.60
1	76	🥇	76_conv_standard_1D_dilated_strided__ conv1d_warp_uniform_base_base	0.01	1.88	8.37
3	42	🥇	42_GRUBidirectionalHidden 42_GRUBidirectionalHidden_grid_optimized_base	51.74	1.87	1.96
2	92	🥇	92_Conv2d_GroupNorm_Tanh_HardSwish_ResidualAdd_LogSumExp optimal_block_size_kernel_base	0.06	1.86	1.06
1	51	🥇	51_Argmax_over_a_dimension warp_argmax_nosm_edit_1	0.01	1.85	2.54
1	100	🥇	100_HingeLoss 100_HingeLoss	0.01	1.85	1.74
2	49	🥇	49_ConvTranspose3d_Softmax_Sigmoid adaptive_block_softmax_sigmoid_base_base	1.57	1.83	0.95
2	31	🥇	31_Conv2d_Min_Add_Multiply block_size_tuned_conv2d_base_base	0.03	1.83	2.42
2	3	🥇	3_ConvTranspose3d_Sum_LayerNorm_AvgPool_GELU 3_convtranspose3d_sum_layernorm_avgpool_gelu_opt_customnorm_base	25.20	1.81	0.40
1	35	🥇	35_GroupNorm_ pipelined_stream_groupnorm_optimized_base	0.37	1.81	1.41
2	53	🥇	53_Gemm_Scaling_Hardtanh_GELU modular_functions_base_edit_1	0.03	1.81	1.78
3	47	🥇	47_NetVladNoGhostClusters netvlad_fused_streams_edit_1	0.07	1.79	1.14
2	17	🥇	17_Conv2d_InstanceNorm_Divide unrolled_fused_conv_instnorm_base_base	0.04	1.79	1.62
2	21	🥇	21_Conv2d_Add_Scale_Sigmoid_GroupNorm shared_memory_coalesced_access_kernel_base	0.04	1.79	1.53
2	32	🥇	32_Conv2d_Scaling_Min warp_aligned_conv_scale_min_base	0.03	1.77	1.86
1	48	🥇	48_Mean_reduction_over_a_dimension evenly_distributed_mean_base	0.01	1.76	3.62
2	87	🥇	87_Conv2d_Subtract_Subtract_Mish 87_conv2d_subtract_subtract_mish_templated_base_base	0.03	1.75	2.05
2	25	🥇	25_Conv2d_Min_Tanh_Tanh conv_min_tanh_optimized_base	0.03	1.74	1.83
1	52	🥇	52_Argmin_over_a_dimension 52_argmin_tuned_blocks_base_base	0.01	1.73	2.40
1	83	🥇	83_conv_depthwise_2D_square_input_asymmetric_kernel hybrid_tiled_warp_depthwise_conv_edit_1	0.02	1.72	22.65
2	70	🥇	70_Gemm_Sigmoid_Scaling_ResidualAdd optimized_sigmoid_scaling_residual_add_base	0.03	1.71	1.74
2	4	🥇	4_Conv2d_Mish_Mish conv2d_mish_warp_uniform_base_base	0.03	1.71	2.34
2	52	🥇	52_Conv2d_Activation_BatchNorm ldg_alignment_fusion_opt_base	0.06	1.70	1.14
2	80	🥇	80_Gemm_Max_Subtract_GELU warp_aligned_gemm_base_edit_1	0.03	1.70	1.81
2	16	🥇	16_ConvTranspose2d_Mish_Add_Hardtanh_Scaling 16_ConvTranspose2d_Mish_Add_Hardtanh_Scaling_coalesced_memory_base_edit_1	0.13	1.69	1.06
1	37	🥇	37_FrobeniusNorm_ modular_frobenius_norm_edit_1	0.20	1.64	2.53
2	67	🥇	67_Conv2d_GELU_GlobalAvgPool unrolled_fused_conv_gelu_pool_base	0.03	1.63	2.28
2	2	🥇	2_ConvTranspose2d_BiasAdd_Clamp_Scaling_Clamp_Divide ldg_128bit_align_opt_base	0.18	1.63	0.77
2	57	🥇	57_Conv2d_ReLU_HardSwish balanced_workload_conv2d_base	0.04	1.63	1.46
1	96	🥇	96_HuberLoss sync_optimized_unrolled_reduction_edit_1	0.01	1.62	5.37
2	51	🥇	51_Gemm_Subtract_GlobalAvgPool_LogSumExp_GELU_ResidualAdd fused_forward_base	0.05	1.62	0.92
1	25	🥇	25_Swish 25_Swish	0.01	1.56	8.89
3	22	🥇	22_EfficientNetB0 22_EfficientNetB0	1.60	1.56	0.69
2	94	🥇	94_Gemm_BiasAdd_Hardtanh_Mish_GroupNorm fused_aligned_ldg_base	0.03	1.56	1.55
2	26	🥇	26_ConvTranspose3d_Add_HardSwish ldg_smem_vectorized_edit2_edit_1	3.32	1.53	1.00
1	84	🥇	84_conv_depthwise_2D_asymmetric_input_square_kernel 84_conv_dw2d_unroll_gridstride_shared_kernel_base	0.01	1.52	3.22
2	75	🥇	75_Gemm_GroupNorm_Min_BiasAdd fused_groupnorm_min_base	0.02	1.51	2.10
1	49	🥇	49_Max_reduction_over_a_dimension adaptive_max_reduce_base	0.02	1.50	2.04
2	37	🥇	37_Matmul_Swish_Sum_GroupNorm fused_swish_bias_groupnorm_aligned_edit_1	0.03	1.50	1.53
2	84	🥇	84_Gemm_BatchNorm_Scaling_Softmax modular_fused_gemm_bn_softmax_edit_1	0.04	1.50	0.81
2	93	🥇	93_ConvTranspose2d_Add_Min_GELU_Multiply warp_optimized_reduction_with_shared_memory_edit_1	0.17	1.49	0.80
3	19	🥇	19_MobileNetV1 19_MobileNetV1	1.13	1.49	0.79
1	47	🥇	47_Sum_reduction_over_a_dimension fully_unrolled_warp_sum_reduction_base	0.01	1.47	2.97
2	62	🥇	62_Matmul_GroupNorm_LeakyReLU_Sum warp_fused_gn_lrelu_sum_base	0.02	1.46	2.24
1	93	🥇	93_masked_cumsum shared_memory_optimized_masked_cumsum_base	0.02	1.45	0.93
1	85	🥇	85_conv_depthwise_2D_asymmetric_input_asymmetric_kernel combined_conv_vectorized_edit_1	0.02	1.44	3.46
1	42	🥇	42_Max_Pooling_2D warp_divergence_optimized_unroll_base	0.02	1.43	3.04
3	9	🥇	9_ResNet18 resnet18_aligned_memory_base	0.68	1.43	0.58
2	60	🥇	60_ConvTranspose3d_Swish_GroupNorm_HardSwish warp_optimized_fused_base_base	5.40	1.42	0.84
3	41	🥇	41_GRUBirectional 41_GRUBirectional	69.30	1.42	1.45
3	15	🥇	15_DenseNet121 optimized_dense_net_base	4.19	1.42	0.90
1	67	🥇	67_conv_standard_1D aligned_ldg_conv1d_base	0.01	1.41	2.75
2	7	🥇	7_Conv3d_ReLU_LeakyReLU_GELU_Sigmoid_BiasAdd coalesced_memory_activation_kernel_base_base	0.76	1.41	0.69
2	46	🥇	46_Conv2d_Subtract_Tanh_Subtract_AvgPool balanced_workload_conv2d_subtract_tanh_avgpool_base	0.04	1.39	1.48
3	50	🥇	50_ReLUSelfAttention shared_memory_bias_tiling_edit_1	3.71	1.37	0.73
3	1	🥇	1_MLP strided_relu_fusion_base_base	0.03	1.35	6.59
2	61	🥇	61_ConvTranspose3d_ReLU_GroupNorm fused_rg_atomic_opt_base_base	0.18	1.34	1.14
2	27	🥇	27_Conv3d_HardSwish_ReLU_Softmax_Mean ldg_aligned_fused_kernel_base	0.83	1.34	0.72
3	10	🥇	10_ResNet101 resnet101_modular_functions_base_base	23.20	1.33	1.33
3	7	🥇	7_GoogleNetInceptionV1 optimized_thread_block_indexing_edit_1	1.70	1.31	0.71
2	12	🥇	12_Gemm_Multiply_LeakyReLU 12_gemm_warp_primitives_base	0.03	1.30	2.32
2	48	🥇	48_Conv3d_Scaling_Tanh_Multiply_Sigmoid optimized_hybrid_conv3d_base	0.78	1.29	0.68
3	23	🥇	23_EfficientNetB1 23_EfficientNetB1	1.09	1.28	0.68
2	54	🥇	54_Conv2d_Multiply_LeakyReLU_GELU dynamic_block_size_54conv_base	0.04	1.28	1.44
2	63	🥇	63_Gemm_ReLU_Divide unrolled_tiled_gemm_base_base	0.03	1.28	1.34
2	64	🥇	64_Gemm_LogSumExp_LeakyReLU_LeakyReLU_GELU_GELU tiled_fused_optimized_kernel_edit_1	0.06	1.25	0.80
2	90	🥇	90_Conv3d_LeakyReLU_Sum_Clamp_GELU aligned_vectorized_ldg_90_conv3d_edit_1	0.79	1.25	0.66
2	85	🥇	85_Conv2d_GroupNorm_Scale_MaxPool_Clamp conv2d_gn_scale_pool_clamp_sync_opt_base	0.06	1.24	1.15
1	82	🥇	82_conv_depthwise_2D_square_input_square_kernel manual_unroll_depthwise_2d_kernel_edit_1	0.03	1.24	2.14
3	39	🥇	39_GRU 39_gru_constant_memory_edit_1	27.75	1.24	1.83
1	44	🥇	44_Average_Pooling_1D vectorized_4x_base	0.01	1.24	8.08
2	44	🥇	44_ConvTranspose2d_Multiply_GlobalAvgPool_GlobalAvgPool_Mean optimized_spatial_reduction_edit_1	0.18	1.21	0.72
2	74	🥇	74_ConvTranspose3d_LeakyReLU_Multiply_LeakyReLU_Max 74_ConvTranspose3d_LeakyReLU_Multiply_LeakyReLU_Max_fused_edit_1	1.21	1.21	0.70
2	96	🥇	96_ConvTranspose3d_Multiply_Max_GlobalAvgPool_Clamp conv_transpose3d_opt_stride_loops_edit_1	4.39	1.21	1.22
2	8	🥇	8_Conv3d_Divide_Max_GlobalAvgPool_BiasAdd_Sum fused_stride_loops_base	0.75	1.21	0.91
3	33	🥇	33_VanillaRNN fused_rnn_i2h_warp_base	0.02	1.21	2.67
2	69	🥇	69_Conv2d_HardSwish_ReLU fused_hardswish_relu_const_edit_1	0.04	1.19	1.58
2	11	🥇	11_ConvTranspose2d_BatchNorm_Tanh_MaxPool_GroupNorm 11_convtranspose_bn_fusedtanhm_pool_groupnorm_warp_optimized_base	0.73	1.19	0.49
1	41	🥇	41_Max_Pooling_1D modular_device_functions_edit_1_base	0.01	1.18	5.01
2	71	🥇	71_Conv2d_Divide_LeakyReLU aligned_memory_access_base	0.04	1.18	1.54
2	19	🥇	19_ConvTranspose2d_GELU_GroupNorm opt_convtrans_gelu_gn_even_distribution_base	0.55	1.18	0.61
3	12	🥇	12_VGG19 vgg19_cudnn_optimized_base	3.15	1.17	0.70
2	1	🥇	1_Conv2D_ReLU_BiasAdd block_size_optimized_base	0.04	1.17	1.49
1	22	🥇	22_Tanh combined_tanh_kernel_edit_1	0.01	1.17	4.88
1	29	🥇	29_Softplus warp_optimized_softplus_base	0.01	1.16	4.88
2	36	🥇	36_ConvTranspose2d_Min_Sum_GELU_Add warp_reduction_optimized_kernel_base_base	0.18	1.15	0.74
2	89	🥇	89_ConvTranspose3d_MaxPool_Softmax_Subtract_Swish_Max balanced_thread_block_distribution_base	5.03	1.14	0.99
1	31	🥇	31_ELU vec_shared_elu_base	0.01	1.14	4.80
1	20	🥇	20_LeakyReLU shared_leakyrelu_base	0.01	1.13	4.85
2	100	🥇	100_ConvTranspose3d_Clamp_Min_Divide memory_coalescing_optimization_base	0.57	1.13	0.84
2	6	🥇	6_Conv3d_Softmax_MaxPool_MaxPool strided_maxpool_base_base	0.95	1.13	0.90
1	26	🥇	26_GELU_ 26_gelu_ldg_vec_base	0.01	1.13	4.96
1	28	🥇	28_HardSigmoid hardsigmoid_shared_optimized_edit_1	0.01	1.12	4.96
1	21	🥇	21_Sigmoid optimized_sigmoid_hybrid_edit_1	0.01	1.11	4.82
1	27	🥇	27_SELU_ 27_selu_aligned_ldg_base	0.01	1.10	4.96
2	5	🥇	5_ConvTranspose2d_Subtract_Tanh optimized_shared_mem_tanh_base	0.08	1.09	0.91
2	47	🥇	47_Conv3d_Mish_Tanh shared_mem_mish_tanh_base_base	0.10	1.09	0.95
1	72	🥇	72_conv_transposed_3D_asymmetric_input_asymmetric_kernel___strided_padded_grouped_ minimize_warp_divergence_base	25.36	1.08	1.09
1	19	🥇	19_ReLU 19_ReLU	0.01	1.08	4.75
3	44	🥇	44_MiniGPTBlock block_size_optimized_transformer_base_base	28.16	1.08	0.82
1	24	🥇	24_LogSoftmax strided_logsoftmax_base_base	0.01	1.07	3.72
2	34	🥇	34_ConvTranspose3d_LayerNorm_GELU_Scaling balanced_load_kernel_base	42.49	1.07	0.21
1	23	🥇	23_Softmax reduced_sync_softmax_kernel_edit_1	0.01	1.07	3.43
1	32	🥇	32_HardTanh 32_hardtanh_aligned_128_opt_edit_1	0.01	1.06	5.07
2	38	🥇	38_ConvTranspose3d_AvgPool_Clamp_Softmax_Multiply modular_device_functions_refactor_base	0.64	1.06	0.95
2	43	🥇	43_Conv3d_Max_LogSumExp_ReLU block_tuned_fused_kernel_base_base	0.79	1.05	1.01
2	72	🥇	72_ConvTranspose3d_BatchNorm_AvgPool_AvgPool warp_uniform_control_flow_edit_1	23.59	1.05	1.06
2	78	🥇	78_ConvTranspose3d_Max_Max_Sum optimized_maxpool_kernel_base	0.58	1.05	1.21
2	98	🥇	98_Matmul_AvgPool_GELU_Scale_Max fused_pipeline_base	0.03	1.04	1.50
1	94	🥇	94_MSELoss mse_unrolled_optimized_edit_1	0.02	1.03	2.04
1	75	🥇	75_conv_transposed_2D_asymmetric_input_asymmetric_kernel_strided__grouped____padded____dilated__ conv_transposed_2d_tiled_shared_bias_base	6.44	1.03	1.04
2	73	🥇	73_Conv2d_BatchNorm_Scaling 73_Conv2d_BatchNorm_Scaling	0.12	1.02	1.43
1	69	🥇	69_conv_transposed_2D__asymmetric_input__asymmetric_kernel 69_conv_transposed_2D__asymmetric_input__asymmetric_kernel	0.03	1.02	1.74
1	64	🥇	64_conv_transposed_1D modular_64_conv_transposed_1D_base	0.02	1.02	2.24
3	11	🥇	11_VGG16 optimized_vgg16_with_custom_conv_base	3.19	1.02	0.60
3	16	🥇	16_DenseNet201 warp_optimized_densenet_op_base	8.04	1.01	1.03
2	41	🥇	41_Gemm_BatchNorm_GELU_GroupNorm_Mean_ReLU 41_Gemm_BatchNorm_GELU_GroupNorm_Mean_ReLU	0.06	1.01	0.66
1	91	🥇	91_cumsum_reverse reverse_cumsum_block_size_tuning_edit_1	0.04	1.01	0.96
2	77	🥇	77_ConvTranspose3d_Scale_BatchNorm_GlobalAvgPool ldg_memory_alignment_optimization_edit_1	0.78	1.01	0.72
1	57	🥇	57_conv_transposed_2D__square_input__square_kernel stride_loop_optimized_conv_transpose2d_base	0.15	1.01	1.18
1	58	🥇	58_conv_transposed_3D__asymmetric_input__asymmetric_kernel 58_conv_transposed_3D__asymmetric_input__asymmetric_kernel	2.30	1.01	1.03
1	78	🥇	78_conv_transposed_2D_asymmetric_input_asymmetric_kernel___padded__ conv_trans_tuned_blocks_base_base	0.33	1.00	1.14
3	27	🥇	27_RegNet 27_RegNet	2.24	1.00	0.44
3	45	🥇	45_UNetSoftmax 45_UNetSoftmax	4.94	1.00	0.88
1	62	🥇	62_conv_standard_2D__square_input__asymmetric_kernel 62_conv_standard_2D__square_input__asymmetric_kernel	0.28	1.00	1.52
2	24	🥇	24_Conv3d_Min_Softmax shared_memory_tiling_optimization_base	0.77	1.00	0.69
3	8	🥇	8_ResNetBasicBlock 8_ResNetBasicBlock	1.68	1.00	0.48
2	50	🥇	50_ConvTranspose3d_Scaling_AvgPool_BiasAdd_Scaling 50_ConvTranspose3d_Scaling_AvgPool_BiasAdd_Scaling	5.73	1.00	0.75
1	8	🥇	8_Matmul_with_irregular_shapes_ cublas_handle_reuse_optimized_edit_1	6.22	1.00	1.02
3	25	🥇	25_ShuffleNetUnit 25_ShuffleNetUnit	9.27	1.00	1.31
3	14	🥇	14_DenseNet121DenseBlock 14_DenseNet121DenseBlock	7.15	1.00	0.86
1	5	🥇	5_Matrix_scalar_multiplication vector_no_sync_scalar_mult_edit_1	0.18	1.00	2.21
1	71	🥇	71_conv_transposed_2D__asymmetric_input__square_kernel atomic_optimized_transpose_base	0.27	1.00	1.16
2	15	🥇	15_ConvTranspose3d_BatchNorm_Subtract 15_ConvTranspose3d_BatchNorm_Subtract	2.02	1.00	0.53
3	6	🥇	6_GoogleNetInceptionModule 6_GoogleNetInceptionModule	8.72	1.00	1.15
1	2	🥇	2_Standard_matrix_multiplication_ hybrid_matmul_base	0.43	1.00	1.08
3	17	🥇	17_SqueezeNetFireModule 17_SqueezeNetFireModule	0.84	1.00	0.93
1	63	🥇	63_conv_standard_2D__square_input__square_kernel adaptive_conv2d_cuda_base	0.23	1.00	1.68
2	39	🥇	39_Gemm_Scale_BatchNorm gemm_scale_batchnorm_warp_divergence_optimized_edit_1	0.05	1.00	0.51
1	55	🥇	55_conv_standard_2D__asymmetric_input__square_kernel adaptive_conv2d_blocksize_tuning_base	0.12	1.00	1.75
3	5	🥇	5_AlexNet 5_AlexNet	0.56	0.99	0.84
3	40	🥇	40_GRUHidden 40_GRUHidden	36.25	0.97	1.38
3	13	🥇	13_DenseNet121TransitionLayer 13_DenseNet121TransitionLayer	0.57	0.94	0.52
1	4	🥇	4_Matrix_vector_multiplication_ modular_device_functions_matvec_base	0.07	0.94	2.61
1	79	🥇	79_conv_transposed_1D_asymmetric_input_square_kernel___padded____strided____dilated__ conv_transpose1d_shared_tile_sync_base	0.02	0.93	2.61
2	76	🥇	76_Gemm_Add_ReLU combined_warp_tile_base	0.03	0.93	1.54
2	86	🥇	86_Matmul_Divide_GELU block_size_optimized_fused_kernel_base_base	0.03	0.92	1.63
2	30	🥇	30_Gemm_GroupNorm_Hardtanh warp_divergence_minimization_base	0.06	0.88	0.91
1	86	🥇	86_conv_depthwise_separable_2D conv_dw_separable_strided_loops_edit_1	0.31	0.81	1.18
1	9	🥇	9_Tall_skinny_matrix_multiplication_ unrolled_loop_matmul_base	0.68	0.78	0.59
3	36	🥇	36_LTSMHn optimized_lstm_base	36.44	0.76	1.59
1	14	🥇	14_Matmul_for_upper_triangular_matrices coalesced_memory_access_upper_triangular_matmul_base	3.90	0.72	0.74
1	7	🥇	7_Matmul_with_small_K_dimension_ modular_matmul_refactored_base	0.99	0.67	0.58
1	13	🥇	13_Matmul_for_symmetric_matrices vec_ldg_aligned_matmul_128_base_optimized_base	4.29	0.64	0.67
1	33	🥇	33_BatchNorm adaptive_blocksize_batchnorm_base	0.90	0.62	0.39
3	37	🥇	37_LTSMCn 37_ltsmcn_balanced_workload_base_base	52.09	0.52	1.10
3	35	🥇	35_LTSM 35_lstm_grid_stride_base_base	72.97	0.44	0.83
1	1	🥇	1_Square_matrix_multiplication_ regtile_2x2_optimized_sync_edit_1	1.01	0.42	0.44
1	80	🥇	80_conv_standard_2D_square_input_asymmetric_kernel___dilated____padded__ warp_divergence_minimized_conv2d_base	0.88	0.31	0.48
1	3	🥇	3_Batched_matrix_multiplication bmm_tiled_shared_memory_optimized_edit_1	0.51	0.25	0.35
1	87	🥇	87_conv_pointwise_2D stride_loop_conv2d_base	0.45	0.25	0.77
1	11	🥇	11_4D_tensor_matrix_multiplication ldg_optimized_shared_mem_tiled_4d_matrix_mult_base	94.27	0.22	0.22
1	10	🥇	10_3D_tensor_matrix_multiplication unrolled_tiled_kernel_base	5.52	0.21	0.23
1	18	🥇	18_Matmul_with_transposed_both optimized_matmul_transpose_base	1.87	0.19	0.22
1	81	🥇	81_conv_transposed_2D_asymmetric_input_square_kernel___dilated____padded____strided__ conv_transpose2d_thread_block_map_edit_1	10.93	0.16	0.17
1	16	🥇	16_Matmul_with_transposed_A tiled_double_output_base	2.29	0.15	0.17
1	60	🥇	60_conv_standard_3D__square_input__asymmetric_kernel pipelined_streams_conv3d_base	37.87	0.14	0.14
1	17	🥇	17_Matmul_with_transposed_B warp_matmul_optimized_v2_base	2.53	0.14	0.16
1	56	🥇	56_conv_standard_2D__asymmetric_input__asymmetric_kernel 56_conv_unrolled_2d_base	1.13	0.13	0.20
1	6	🥇	6_Matmul_with_large_K_dimension_ double_buffered_matmul_base	5.11	0.07	0.11
1	65	🥇	65_conv_transposed_2D__square_input__asymmetric_kernel atomic_operations_minimization_base	3.71	0.05	0.06