framework,version,device,op_name,kernel_source,phase,batch_size,seq_len,num_tokens,d_model,d_state,d_conv,nheads,head_dim,n_groups,chunk_size,model_name,latency
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,1,1,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.017286400496959686
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,1,1,1,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.005008000135421753
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,1,1,1,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.00517439991235733
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,1,1,1,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.0065311998128890995
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,1,1,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.02502079904079437
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,2,1,2,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.007395199686288834
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,1,1,1,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.007599999755620956
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,2,1,2,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.007910399883985519
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,2,1,2,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.006569600105285645
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,2,1,2,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.01055999994277954
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,4,1,4,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.007791999727487564
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,4,1,4,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.011052799969911575
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,4,1,4,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.008556800335645676
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,4,1,4,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.015497599542140961
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,8,1,8,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.009609600156545639
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,8,1,8,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.01525759994983673
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,8,1,8,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.008374399691820144
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,8,1,8,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.025648000836372375
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,16,1,16,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.006694400310516357
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,16,1,16,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.008886399865150451
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,16,1,16,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.02502720057964325
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,16,1,16,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.0458624005317688
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,32,1,32,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.00761599987745285
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,32,1,32,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.046096000075340274
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,32,1,32,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.009974399954080582
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,32,1,32,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.08714240193367004
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,64,1,64,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.009312000125646591
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,64,1,64,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.08693119883537292
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,64,1,64,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.011865600198507308
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,64,1,64,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.16911360025405883
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,128,1,128,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.012969599664211273
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,128,1,128,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.1688480019569397
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,128,1,128,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.01746560037136078
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,128,1,128,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.3314624071121216
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,256,1,256,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.019465599954128266
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,256,1,256,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.3309920072555542
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,256,1,256,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.028019198775291444
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,256,1,256,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.656441593170166
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,512,1,512,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.031839999556541446
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,512,1,512,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.6558527946472168
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,512,1,512,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.05039039850234985
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,512,1,512,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,1.3049407958984376
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,1024,1,1024,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.05846080183982849
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,1024,1,1024,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,1.3047712326049805
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,1024,1,1024,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.09330880045890808
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,1024,1,1024,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,2.6032127380371093
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,1,1,1,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.005084799975156784
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,1,1,1,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.00647680014371872
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,2,1,2,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.006812799721956253
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,2,1,2,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.007612799853086471
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,1,1,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.017452800273895265
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,4,1,4,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.007078400254249573
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,4,1,4,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.011046399921178817
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,8,1,8,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.006617599725723266
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,8,1,8,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.015116800367832185
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,16,1,16,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.006876800209283829
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,16,1,16,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.025356799364089966
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,32,1,32,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.007382400333881378
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,32,1,32,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.04593920111656189
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,64,1,64,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.009107200056314468
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,64,1,64,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.08686400055885315
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,128,1,128,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.013184000551700593
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,128,1,128,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.16908799409866332
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,256,1,256,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.01976960003376007
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,256,1,256,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.3309727907180786
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,512,1,512,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.031548801064491275
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,512,1,512,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.6559296131134034
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,1024,1,1024,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.058361601829528806
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,1024,1,1024,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,1.3047807693481446
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,1,1,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.02502399981021881
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,1,1,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.021129600703716278
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,2,2,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.01709119975566864
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,2,2,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.0249439999461174
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,2,2,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.027062401175498962
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,2,2,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.02311359941959381
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,4,4,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.02559039890766144
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,4,4,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.018057599663734436
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,4,4,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.02741119861602783
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,4,4,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.02306240051984787
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,8,8,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.015030400454998016
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,8,8,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.02107200026512146
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,8,8,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.023043200373649597
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,8,8,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.025414401292800905
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,16,16,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.02134079933166504
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,16,16,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.014979200065135955
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,1,1,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.023177599906921385
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,16,16,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.025158399343490602
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,16,16,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.02305919975042343
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,2,2,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.017769600450992584
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,32,32,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.021772800385951994
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,32,32,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.027606400847434997
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,32,32,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.015440000593662262
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,32,32,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.023769600689411162
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,64,64,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.02133760005235672
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,64,64,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.02643199861049652
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,64,64,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.015388800203800202
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,64,64,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.025497600436210632
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,128,128,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.021359999477863312
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,128,128,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.029686400294303895
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,128,128,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.015782399475574492
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,128,128,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.029820799827575684
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,256,256,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.022969600558280946
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,256,256,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.017020800709724428
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,256,256,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.045686399936676024
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,256,256,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.03131519854068756
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,512,512,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.01676799952983856
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,512,512,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.023244799673557283
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,512,512,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.03743039965629578
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,512,512,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.07651200294494628
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,2,2,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.02534080147743225
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,1024,1024,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.0175135999917984
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,1024,1024,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.056252801418304445
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,1024,1024,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.02447039932012558
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,1024,1024,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.13046720027923583
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,4,4,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.018163199722766876
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,4,4,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.02595840096473694
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,2048,2048,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.026454401016235352
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,2048,2048,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.038115200400352475
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,2048,2048,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.09374399781227112
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,2048,2048,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.26276159286499023
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,8,8,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.015740799903869628
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,8,8,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.02627840042114258
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,4096,4096,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.04410560131072998
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,4096,4096,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.1816704034805298
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,4096,4096,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.06577600240707397
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,4096,4096,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.5040224075317383
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,16,16,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.015647999942302704
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,8192,8192,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.08395839929580688
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,8192,8192,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.35005760192871094
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,8192,8192,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.12526079416275024
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,8192,8192,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.9873344421386718
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,16384,16384,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.1574687957763672
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,16384,16384,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.6749216079711914
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,16384,16384,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.23818559646606446
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,16384,16384,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,1.9467199325561524
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,32768,32768,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.30589759349823
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,32768,32768,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,1.3205120086669921
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,32768,32768,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.46592001914978026
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,32768,32768,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,3.875436782836914
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,1,2,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.031760001182556154
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,1,2,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.04796159863471985
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,16,16,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.02560960054397583
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,32,32,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.015638400614261628
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,32,32,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.025839999318122864
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,64,64,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.015667200088500977
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,64,64,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.025567999482154845
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,128,128,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.015523199737071992
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,128,128,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.029811200499534608
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,256,256,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.016672000288963318
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,256,256,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.03806079924106598
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,512,512,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.017148800194263458
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,512,512,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.05802559852600098
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,1,2,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.03739199936389923
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,1,2,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.025091201066970825
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,1024,1024,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.0177824005484581
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,1024,1024,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.08749759793281556
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,2,4,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.047593599557876586
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,2,4,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.03160319924354553
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,2048,2048,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.026111999154090883
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,2048,2048,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.14904320240020752
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,1,1,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.009036800265312195
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,4096,4096,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.0438400000333786
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,4096,4096,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.2838912010192871
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,8192,8192,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.08239039778709412
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,8192,8192,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.5446784019470214
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,16384,16384,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.1560320019721985
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,16384,16384,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,1.0571743965148925
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,32768,32768,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.30604801177978513
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,32768,32768,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,2.078790473937988
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,1,2,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.03127999901771546
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,2,4,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.037308800220489505
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,2,4,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.0230335995554924
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,4,8,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.04605759978294373
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,4,8,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.037462401390075686
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,4,8,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.02990399897098541
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,4,8,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.023475199937820435
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,8,16,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.035606399178504944
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,8,16,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.037990400195121767
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,8,16,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.023423999547958374
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,8,16,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.025203201174736022
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,16,32,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.03590080142021179
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,16,32,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.02495039999485016
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,1,2,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.03531840145587921
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,2,4,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.030902400612831116
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,16,32,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.03795520067214966
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,16,32,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.025036799907684325
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,32,64,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.03695679903030395
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,32,64,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.03794879913330078
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,32,64,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.025203201174736022
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,32,64,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.02531520128250122
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,64,128,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.037084800004959104
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,64,128,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.03991039991378784
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,64,128,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.025497600436210632
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,64,128,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.027913600206375122
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,128,256,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.037913599610328676
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,128,256,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.046275201439857486
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,128,256,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.025536000728607178
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,128,256,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.032067200541496275
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,2,4,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.03526079952716828
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,256,512,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.037539198994636536
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,256,512,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.07454720139503479
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,256,512,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.02492160052061081
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,256,512,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.03532800078392029
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,4,8,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.029894399642944335
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,4,8,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.03583999872207642
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,512,1024,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.03897919952869415
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,512,1024,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.026739200949668883
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,512,1024,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.12844159603118896
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,512,1024,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.056006401777267456
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,8,16,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.0234047994017601
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,8,16,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.035913598537445066
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,1024,2048,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.041280001401901245
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,1024,2048,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.02794559895992279
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,16,32,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.0247871994972229
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,1024,2048,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.09461439847946167
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,1024,2048,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.2548255920410156
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,2048,4096,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.04385280013084412
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,2048,4096,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.06843199729919433
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,2048,4096,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.1754528045654297
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,2048,4096,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.4842048168182373
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,4096,8192,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.07893760204315185
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,4096,8192,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.12325119972229004
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,4096,8192,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.3367455959320068
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,4096,8192,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.9463839530944824
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,8192,16384,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.1466655969619751
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,8192,16384,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.22961599826812745
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,8192,16384,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.6457856178283692
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,8192,16384,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,1.8663103103637695
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,16384,32768,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.2811199903488159
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,16384,32768,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,1.267420768737793
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,16384,32768,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.4435264110565186
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,16384,32768,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,3.716352081298828
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,32768,65536,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.550380802154541
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,32768,65536,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,2.543526458740234
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,16,32,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.035334399342536925
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,32768,65536,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.8732383728027344
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,32768,65536,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,7.482685089111328
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,1,4,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.05591999888420105
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,32,64,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.024937599897384644
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,1,4,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.027744001150131224
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,32,64,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.03597759902477264
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,1,4,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.08789119720458985
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,1,4,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.05844159722328186
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,2,8,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.05559679865837097
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,2,8,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.029686400294303895
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,64,128,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.02542400062084198
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,64,128,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.03598720133304596
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,2,8,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.0875935971736908
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,2,8,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.06077119708061218
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,4,16,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.054611200094223024
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,4,16,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.029657599329948426
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,128,256,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.025676798820495606
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,128,256,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.04598079919815064
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,4,16,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.08729919791221619
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,4,16,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.06086080074310303
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,8,32,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.04205119907855988
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,8,32,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.02943040132522583
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,256,512,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.025567999482154845
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,256,512,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.05880640149116516
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,8,32,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.06628159880638122
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,8,32,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.060031998157501223
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,16,64,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.0419871985912323
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,16,64,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.029631999135017396
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,512,1024,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.02659519910812378
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,512,1024,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.08930559754371643
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,16,64,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.0661728024482727
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,32,128,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.04312320053577423
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,32,128,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.02943040132522583
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,16,64,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.060224002599716185
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,1024,2048,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.027952000498771667
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,1024,2048,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.15458240509033203
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,64,256,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.043536001443862916
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,32,128,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.06745920181274415
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,64,256,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.03162240087985992
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,32,128,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.06040319800376892
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,2048,4096,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.04426240026950836
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,2048,4096,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.2881407976150513
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,128,512,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.04362240135669708
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,64,256,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.06744319796562195
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,128,512,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.036025598645210266
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,64,256,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.06223040223121643
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,4096,8192,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.07898880243301391
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,4096,8192,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.553865623474121
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,256,1024,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.044486400485038755
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,128,512,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.06802560091018676
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,256,1024,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.0563103973865509
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,128,512,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.07839679718017578
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,8192,16384,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.14664000272750854
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,8192,16384,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,1.0749664306640625
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,512,2048,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.0466592013835907
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,256,1024,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.06928319931030273
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,512,2048,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.09455040097236633
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,256,1024,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.13006399869918822
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,16384,32768,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.28072640895843504
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,16384,32768,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,2.1109088897705077
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,1024,4096,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.04892480075359344
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,512,2048,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.07208960056304932
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,1024,4096,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.1768671989440918
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,512,2048,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.24868481159210204
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,32768,65536,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.5497471809387207
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,2048,8192,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.08223680257797242
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,1024,4096,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.0779583990573883
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,2048,8192,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.3396768093109131
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,1024,4096,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.4819808006286621
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,32768,65536,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,4.2285888671875
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,4096,16384,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.1455456018447876
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,2048,8192,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.1301792025566101
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,1,4,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.0546239972114563
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,4096,16384,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.6484127998352051
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,1,4,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.05379520058631897
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,2048,8192,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.9497407913208008
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,2,8,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.054396802186965944
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,2,8,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.05382080078125
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,8192,32768,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.2732640027999878
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,4096,16384,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.23292479515075684
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,8192,32768,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,1.2675423622131348
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,4096,16384,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,1.8742271423339845
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,4,16,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.05427520275115967
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,4,16,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.054313600063323975
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,16384,65536,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.5287199974060058
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,8192,32768,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.44023680686950684
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,16384,65536,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,2.548320007324219
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,8192,32768,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,3.723929595947266
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,8,32,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.042166399955749514
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,8,32,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.0546176016330719
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,32768,131072,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,1.0396127700805664
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,16384,65536,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.8555808067321777
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,16,64,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.04195519983768463
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,16,64,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.054262399673461914
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,32768,131072,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,5.115465545654297
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,16384,65536,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,7.507711791992188
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,1,1,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.01767359972000122
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,32,128,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.04338560104370117
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,32,128,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.05581759810447693
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,1,8,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.10560640096664428
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,1,8,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.04216000139713287
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,32768,131072,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,1.6876800537109375
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,2,2,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.008953599631786347
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,64,256,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.04359680116176605
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,64,256,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.054630398750305176
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,2,16,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.10538560152053833
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,2,16,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.04169279932975769
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,32768,131072,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,15.098687744140625
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,128,512,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.043971198797225955
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,128,512,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.0710752010345459
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,4,32,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.10363839864730835
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,4,32,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.04207679927349091
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,1,8,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.1711392045021057
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,256,1024,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.04425599873065948
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,256,1024,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.09140480160713196
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,1,8,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.09922879934310913
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,8,64,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.07777600288391114
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,8,64,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.04185599982738495
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,512,2048,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.04671359956264496
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,2,16,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.17066880464553832
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,16,128,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.07826560139656066
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,2,16,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.09921600222587586
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,512,2048,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.15592960119247437
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,16,128,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.04259839951992035
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,32,256,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.07904959917068481
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,4,32,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.16725120544433594
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,1024,4096,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.04802559912204742
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,32,256,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.04135999977588654
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,4,32,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.1008031964302063
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,1024,4096,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.2875648021697998
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,2,2,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.01910720020532608
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,64,512,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.08069760203361512
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,8,64,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.12479360103607177
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,64,512,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.047727999091148374
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,2048,8192,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.08155840039253234
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,8,64,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.09987199902534485
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,2048,8192,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.5541312217712402
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,4,4,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.009884800016880035
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,4,4,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.019708800315856933
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,128,1024,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.08129600286483765
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,16,128,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.1255519986152649
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,128,1024,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.05845119953155518
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,16,128,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.10122560262680054
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,4096,16384,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.14596159458160402
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,8,8,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.008764799684286118
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,8,8,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.01940480023622513
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,4096,16384,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,1.0750975608825684
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,32,256,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.12788159847259523
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,256,2048,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.08156160116195679
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,32,256,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.10145280361175538
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,256,2048,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.09520959854125977
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,16,16,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.008403199911117553
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,8192,32768,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.27314560413360595
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,8192,32768,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,2.1112831115722654
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,64,512,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.12919039726257325
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,512,4096,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.08552640080451965
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,64,512,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.10928000211715698
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,512,4096,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.17735999822616577
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,16384,65536,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.5275968074798584
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,16384,65536,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,4.227321624755859
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,128,1024,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.13069440126419068
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,1024,8192,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.09301440119743347
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,128,1024,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.1323456048965454
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,1024,8192,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.3400768041610718
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,32768,131072,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,1.0396127700805664
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,256,2048,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.131113600730896
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,2048,16384,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.15463999509811402
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,256,2048,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.2524224042892456
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,2048,16384,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.6497407913208008
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,32768,131072,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,8.476390075683593
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,512,4096,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.13853759765625
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,512,4096,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.4791327953338623
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,4096,32768,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.27826240062713625
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,4096,32768,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,1.271497631072998
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,1,8,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.105513596534729
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,1,8,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.08725759983062745
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,16,16,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.019094400107860565
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,1024,8192,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.1510591983795166
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,1024,8192,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.9376352310180665
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,8192,65536,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.5272895812988281
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,8192,65536,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,2.5526784896850585
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,2,16,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.10503360033035278
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,32,32,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.008790399879217148
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,32,32,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.019420799612998963
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,2,16,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.08677759766578674
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,2048,16384,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.2519551992416382
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,2048,16384,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,1.8437503814697265
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,64,64,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.008563199639320373
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,4,32,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.10330560207366943
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,64,64,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.021004800498485566
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,4,32,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.08728640079498291
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,16384,131072,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,1.0255871772766114
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,16384,131072,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,5.13070068359375
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,4096,32768,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.45354881286621096
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,128,128,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.008912000060081481
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,128,128,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.025251200795173644
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,8,64,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.07732160091400146
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,8,64,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.0872160017490387
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,4096,32768,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,3.663155364990234
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,32768,262144,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,2.0220224380493166
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,256,256,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.009065599739551544
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,16,128,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.07825279831886292
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,16,128,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.08689280152320862
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,8192,65536,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.8618880271911621
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,256,256,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.025200000405311583
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,32768,262144,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,10.24554901123047
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,8192,65536,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,7.38031005859375
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,32,256,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.07966399788856507
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,32,256,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.08725759983062745
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,512,512,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.009574399888515472
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,512,512,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.028012800216674804
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,1,16,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.20890240669250487
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,64,512,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.08077120184898376
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,64,512,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.09025920033454896
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,16384,131072,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,1.679587173461914
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,1024,1024,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.008892799913883209
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,1024,1024,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.031040000915527343
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,128,1024,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.08159040212631226
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,128,1024,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.11984319686889648
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,16384,131072,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,14.843525695800782
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,2048,2048,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.011772800236940384
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,2048,2048,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.03947199881076813
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,256,2048,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.08217599987983704
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,256,2048,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.16043200492858886
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,4096,4096,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.01926400065422058
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,4096,4096,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.06657599806785583
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,1,16,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.34365119934082033
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,512,4096,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.08565760254859925
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,512,4096,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.2893471956253052
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,8192,8192,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.03182399868965149
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,8192,8192,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.11740800142288207
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,1024,8192,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.09331520199775696
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,1024,8192,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.5555039882659912
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,16384,16384,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.057222402095794676
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,16384,16384,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.2474208116531372
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,1,16,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.06614400148391723
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,2048,16384,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.15520960092544556
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,2048,16384,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,1.0733311653137207
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,1,32768,32768,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.11384960412979125
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,1,32768,32768,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.5541567802429199
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,2,32,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.2036736011505127
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,4096,32768,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.2779839992523193
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,1,2,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.011462400108575821
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,4096,32768,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,2.106928062438965
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,8192,65536,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.5260831832885742
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,1,16,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.1869055986404419
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,8192,65536,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,4.215280151367187
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,2,32,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.33546559810638427
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,16384,131072,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,1.0227519989013671
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,16384,131072,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,8.444678497314452
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,1,2,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.020713600516319274
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,32768,262144,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,2.0169631958007814
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,2,32,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.06642240285873413
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,2,4,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.011347199976444244
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,4,64,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.20081920623779298
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,4,64,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.0674560010433197
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,32768,262144,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,16.8577880859375
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,8,128,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.148745596408844
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,8,128,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.06698560118675231
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,1,16,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.20842878818511962
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,2,32,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.1889024019241333
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,16,256,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.1499519944190979
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,4,64,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.3296639919281006
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,4,64,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.1914687991142273
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,2,4,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.01903360038995743
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,8,128,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.24353280067443847
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,8,128,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.19246400594711305
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,4,8,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.011535999923944473
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,4,8,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.01958719938993454
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,16,256,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.2457119941711426
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,8,16,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.010016000270843506
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,8,16,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.01951040029525757
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,16,32,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.01109120026230812
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,1,16,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.16055999994277953
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,16,256,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.06782079935073852
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,2,32,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.2029184103012085
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,32,512,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.15250879526138306
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,32,512,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.06899840235710145
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,16,256,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.1930176019668579
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,64,1024,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.15370559692382812
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,64,1024,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.07847679853439331
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,16,32,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.021001599729061127
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,128,2048,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.15543359518051147
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,32,512,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.2486624002456665
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,128,2048,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.0993664026260376
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,32,512,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.19516799449920655
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,32,64,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.011187200248241425
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,32,64,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.019475199282169342
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,64,1024,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.2527712106704712
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,256,4096,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.15663679838180541
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,64,1024,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.20544960498809814
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,256,4096,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.18127039670944214
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,64,128,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.01122559979557991
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,64,128,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.0213919997215271
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,128,2048,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.2542367935180664
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,512,8192,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.1662495970726013
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,128,2048,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.25339839458465574
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,512,8192,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.3414207935333252
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,128,256,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.011318399757146835
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,128,256,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.025868800282478333
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,256,4096,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.2578144073486328
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,1024,16384,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.18045120239257811
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,256,4096,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.47990078926086427
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,1024,16384,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.6508224010467529
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,2,32,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.16222399473190308
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,256,512,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.010892800241708755
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,256,512,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.02587839961051941
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,512,8192,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.2731071949005127
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,2048,32768,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.3002239942550659
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,512,8192,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.9338687896728516
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,4,64,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.20024960041046141
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,2048,32768,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,1.2691871643066406
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,4,64,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.16330879926681519
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,512,1024,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.011068800091743469
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,512,1024,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.027926400303840637
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,1024,16384,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.29392640590667723
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,8,128,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.14790400266647338
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,8,128,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.16230080127716065
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,4096,65536,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.5423808097839355
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,1024,16384,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,1.8328191757202148
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,1024,2048,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.01114879995584488
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,4096,65536,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,2.543801689147949
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,1024,2048,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.03595840036869049
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,16,256,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.14970240592956544
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,2048,32768,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.4928256034851074
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,2048,4096,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.016809600591659545
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,2048,4096,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.05768640041351318
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,8192,131072,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,1.033244800567627
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,2048,32768,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,3.637004852294922
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,8192,131072,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,5.111468887329101
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,4096,8192,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.027718400955200194
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,4096,8192,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.0994271993637085
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,4096,65536,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.8936767578125
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,8192,16384,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.04877119958400726
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,4096,65536,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,7.319929504394532
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,8192,16384,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.19866880178451538
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,16384,262144,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,2.016342353820801
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,16384,262144,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,10.217346954345704
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,16384,32768,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.09405120015144348
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,16384,32768,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.41593279838562014
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,8192,131072,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,1.7040672302246094
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,2,32768,65536,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.18067840337753296
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,2,32768,65536,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.8139904022216797
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,8192,131072,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,14.715411376953124
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,1,32,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.40839037895202634
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,1,32,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.120796799659729
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,1,4,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.017539200186729432
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,1,4,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.019571200013160706
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,16,256,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.1625983953475952
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,2,64,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.4044384002685547
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,2,64,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.1211583971977234
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,2,8,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.017667199671268462
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,2,8,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.021382400393486024
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,32,512,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.15231679677963256
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,32,512,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.165011203289032
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,1,32,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.6770431995391846
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,4,128,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.39473600387573243
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,4,128,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.11950080394744873
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,1,32,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.36103999614715576
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,4,16,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.01759680062532425
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,4,16,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.02144320011138916
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,64,1024,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.15384639501571656
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,64,1024,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.1696287989616394
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,8,256,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.29109439849853513
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,8,256,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.12043520212173461
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,2,64,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.6682464122772217
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,8,32,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.01494400054216385
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,2,64,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.365775990486145
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,8,32,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.019420799612998963
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,128,2048,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.1553056001663208
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,128,2048,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.2306528091430664
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,16,512,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.2933759927749634
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,16,512,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.12207679748535157
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,16,64,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.014918400347232819
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,4,128,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.6534719944000245
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,16,64,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.01940480023622513
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,4,128,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.36797120571136477
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,256,4096,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.15693440437316894
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,32,1024,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.29810879230499265
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,256,4096,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.2998847961425781
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,32,1024,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.12453440427780152
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,32,128,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.015376000106334687
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,32,128,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.020451200008392335
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,8,256,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.48088960647583007
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,8,256,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.3692255973815918
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,64,2048,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.30157759189605715
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,512,8192,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.16635520458221437
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,64,2048,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.14481600522994995
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,64,256,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.01499200016260147
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,512,8192,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.5590112209320068
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,64,256,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.02149440050125122
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,16,512,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.4850143909454346
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,16,512,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.37173440456390383
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,128,4096,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.3034463882446289
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,128,4096,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.1857184052467346
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,128,512,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.015097600221633912
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,128,512,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.025360000133514405
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,1024,16384,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.17944960594177245
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,32,1024,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.49147520065307615
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,32,1024,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.37641921043396
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,1024,16384,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,1.075164794921875
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,256,8192,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.3098304033279419
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,256,1024,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.01637440025806427
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,256,8192,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.34513280391693113
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,256,1024,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.027724799513816834
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,64,2048,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.49747838973999026
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,64,2048,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.3936095952987671
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,2048,32768,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.2994368076324463
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,2048,32768,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,2.1088768005371095
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,512,2048,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.01688639968633652
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,512,16384,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.3255199909210205
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,512,2048,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.033276799321174624
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,512,16384,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.6536320209503174
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,128,4096,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.5036575794219971
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,128,4096,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.48555521965026854
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,4096,65536,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.5436319828033447
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,4096,65536,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,4.218243026733399
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,1024,4096,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.016915200650691985
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,1024,4096,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.05297279953956604
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,1024,32768,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.35042240619659426
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,256,8192,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.5111999988555909
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,1024,32768,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,1.2737919807434082
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,256,8192,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.9400351524353028
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,8192,131072,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,1.032918357849121
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,2048,8192,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.02622399926185608
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,2048,8192,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.09089279770851136
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,2048,65536,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.5894847869873047
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,512,16384,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.5361343860626221
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,512,16384,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,1.838751983642578
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,2048,65536,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,2.550547218322754
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,8192,131072,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,8.447869110107423
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,4096,16384,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.043647998571395875
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,4096,16384,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.1745471954345703
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,1024,32768,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.5774464130401611
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,4096,131072,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,1.0743904113769531
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,1024,32768,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,3.6391326904296877
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,16384,262144,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,2.0115968704223635
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,4096,131072,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,5.110287857055664
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,8192,32768,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.08156800270080566
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,8192,32768,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.3454432010650635
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,2048,65536,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.974886417388916
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,16384,65536,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.156278395652771
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,16384,262144,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,16.853964233398436
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,16384,65536,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.672649621963501
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,8192,262144,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,2.048761558532715
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,2048,65536,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,7.309465789794922
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,8192,262144,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,10.213667297363282
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,4,32768,131072,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.3034111976623535
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,4,32768,131072,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,1.3248640060424806
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,4096,131072,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,1.7747583389282227
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,1,32,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.408080005645752
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,1,32,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.3079040050506592
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,1,8,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.02961280047893524
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,1,8,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.021305599808692934
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,4096,131072,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,14.673747253417968
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,2,64,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.4033215999603271
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,2,64,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.30756800174713134
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,2,16,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.02972480058670044
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,2,16,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.01942719966173172
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,1,64,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.8001664161682129
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,1,64,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.22603518962860109
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,4,128,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.3944256067276001
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,4,32,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.02940160036087036
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,4,128,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.30566720962524413
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,4,32,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.021328000724315642
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,2,128,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.7986368179321289
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,2,128,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.22815361022949218
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,8,64,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.023311999440193177
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,8,256,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.290451192855835
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,8,64,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.021132799983024596
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,8,256,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.30681281089782714
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,1,64,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,1.3308287620544434
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,4,256,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.7864543914794921
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,4,256,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.2296191930770874
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,1,64,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.7033376216888427
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,16,128,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.02332800030708313
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,16,128,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.020102399587631225
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,16,512,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.29333760738372805
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,16,512,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.30842878818511965
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,8,512,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.5756832122802734
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,8,512,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.22922880649566652
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,2,128,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,1.3263936042785645
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,2,128,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.7154208183288574
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,32,256,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.024374400079250336
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,32,256,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.020979200303554536
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,32,1024,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.2976032018661499
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,32,1024,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.3104768037796021
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,16,1024,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.5806111812591552
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,16,1024,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.23226559162139893
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,4,256,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,1.3058143615722657
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,4,256,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.7172351837158203
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,64,512,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.024963200092315674
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,64,512,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.02341119945049286
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,64,2048,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.30168321132659914
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,64,2048,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.3218976020812988
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,32,2048,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.5890751838684082
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,32,2048,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.23447680473327637
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,128,1024,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.025097599625587462
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,8,512,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.9548128128051758
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,128,1024,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.027459201216697694
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,8,512,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.7192800045013428
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,128,4096,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.30323519706726076
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,128,4096,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.4387487888336182
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,64,4096,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.5965343952178955
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,64,4096,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.2749824047088623
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,256,2048,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.0251008003950119
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,256,2048,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.031632000207901
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,16,1024,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.9623167991638184
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,256,8192,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.30944960117340087
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,256,8192,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.5688096046447754
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,16,1024,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.7254591941833496
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,128,8192,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.6054175853729248
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,128,8192,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.35243520736694334
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,512,4096,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.026262399554252625
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,512,4096,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.049753600358963014
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,512,16384,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.3250720024108887
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,32,2048,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.976899242401123
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,512,16384,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,1.08089599609375
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,32,2048,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.7344223976135253
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,256,16384,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.6108320236206055
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,256,16384,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.6630047798156739
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,1024,8192,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.02739199995994568
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,1024,8192,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.0847648024559021
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,1024,32768,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.34965760707855226
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,64,4096,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.9925215721130372
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,64,4096,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.7660863876342774
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,1024,32768,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,2.1147647857666017
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,512,32768,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.6427072048187256
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,512,32768,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,1.2835968017578125
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,2048,16384,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.043689599633216857
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,2048,16384,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.16183680295944214
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,128,8192,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,1.0043135643005372
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,2048,65536,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.5892096042633057
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,128,8192,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,0.9526975631713868
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,2048,65536,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,4.234991836547851
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,1024,65536,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,0.690831995010376
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,4096,32768,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.07773759961128235
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,4096,32768,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.3114239931106567
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,1024,65536,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,2.558323287963867
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,256,16384,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,1.0129440307617188
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,256,16384,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,1.8562623977661132
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,4096,131072,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,1.0716032028198241
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,8192,65536,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.1455008029937744
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,8192,65536,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.605456018447876
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,2048,131072,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,1.1713184356689452
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,4096,131072,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,8.452982330322266
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,512,32768,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,1.0639264106750488
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,2048,131072,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,5.116387176513672
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,512,32768,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,3.6575649261474608
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,16384,131072,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.27902400493621826
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,16384,131072,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,1.1868800163269042
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,8192,262144,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,2.0447263717651367
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,1024,65536,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,1.1430208206176757
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,4096,262144,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,2.1350143432617186
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,8,32768,262144,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.5446847915649414
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,8,32768,262144,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,2.341548728942871
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,1024,65536,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,7.328342437744141
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,4096,262144,2688,128,4,64,64,8,128,NEMOTRON_H_3_Nano,10.216175842285157
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,8192,262144,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,16.853395080566408
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,1,16,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.05641279816627502
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,2048,131072,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,1.9395391464233398
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,2048,131072,4096,128,4,128,64,8,128,NEMOTRON_H_3_Super,14.694947814941406
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,1,64,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.8004223823547363
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,1,64,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.5907999992370605
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,2,128,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.798579216003418
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,2,128,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.5899263858795166
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,1,16,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.02301120012998581
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,4,256,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.785807991027832
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,4,256,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.5909056186676025
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,2,32,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.05432000160217285
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,8,512,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.5754879951477051
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,8,512,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.5911712169647216
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,16,1024,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.5802624225616455
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,16,1024,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.5935616016387939
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,32,2048,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.5888288021087646
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,32,2048,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.596124792098999
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,64,4096,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.596070384979248
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,64,4096,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.6177855968475342
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,2,32,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.02502079904079437
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,4,64,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.05405759811401367
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,128,8192,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.6051328182220459
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,4,64,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.02500160038471222
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,128,8192,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.8459775924682618
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,8,128,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.04169920086860657
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,8,128,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.023478400707244874
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,256,16384,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.6100736141204834
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,256,16384,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,1.1000703811645507
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,16,256,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.041503998637199405
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,512,32768,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.6420383930206299
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,512,32768,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,2.128495979309082
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,1024,65536,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,0.6893856048583984
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,1024,65536,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,4.236771011352539
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,2048,131072,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,1.1689279556274415
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,16,256,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.024928000569343568
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,2048,131072,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,8.46171875
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,32,512,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.04260480105876922
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,32,512,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.025241601467132568
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,4096,262144,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,2.1312959671020506
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,64,1024,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.04297600090503693
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,64,1024,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.027599999308586122
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,4096,262144,8192,128,4,64,64,8,256,MAMBA2_GENERIC_4K,16.865072631835936
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,128,2048,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.04344640076160431
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,128,2048,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.03151040077209473
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,256,4096,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.044121599197387694
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,256,4096,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.049660798907279965
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,512,8192,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.04559679925441742
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,512,8192,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.08303359746932984
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,1024,16384,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.04822399914264679
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,1024,16384,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.15403200387954713
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,2048,32768,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.08104320168495179
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,2048,32768,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.29494400024414064
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,4096,65536,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.14502079486846925
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,4096,65536,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.5697855949401855
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,8192,131072,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.2726464033126831
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,8192,131072,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,1.1177056312561036
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,16384,262144,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.526364803314209
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,16384,262144,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,2.204256057739258
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,16,32768,524288,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,1.0369855880737304
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,16,32768,524288,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,4.368127822875977
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,1,32,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.10732799768447876
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,1,32,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.035622400045394895
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,2,64,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.1052672028541565
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,2,64,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.03520320057868957
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,4,128,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.10312639474868775
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,4,128,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.03481279909610748
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,8,256,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.07701119780540466
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,8,256,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.03553279936313629
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,16,512,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.07795519828796386
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,16,512,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.035769599676132205
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,32,1024,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.07951359748840332
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,32,1024,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.03553600013256073
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,64,2048,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.07968000173568726
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,64,2048,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.03983039855957031
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,128,4096,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.0802944004535675
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,128,4096,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.04975680112838745
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,256,8192,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.08056960105895997
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,256,8192,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.08224639892578126
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,512,16384,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.08495039939880371
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,512,16384,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.15096319913864137
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,1024,32768,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.0921567976474762
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,1024,32768,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.2880768060684204
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,2048,65536,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.15456639528274535
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,2048,65536,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.5562911987304687
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,4096,131072,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.277455997467041
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,4096,131072,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,1.0880864143371582
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,8192,262144,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.5245535850524903
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,8192,262144,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,2.1470016479492187
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,16384,524288,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,1.0207200050354004
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,16384,524288,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,4.249923324584961
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,32,32768,1048576,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,2.0152320861816406
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,32,32768,1048576,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,8.460950469970703
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,1,64,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.2042367935180664
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,1,64,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.05400320291519165
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,2,128,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.20366079807281495
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,2,128,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.05628479719161987
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,4,256,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.20101759433746338
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,4,256,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.05604479908943176
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,8,512,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.14824639558792113
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,8,512,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.05370879769325256
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,16,1024,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.14895360469818114
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,16,1024,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.05366399884223938
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,32,2048,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.15115200281143187
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,32,2048,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.05553920269012451
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,64,4096,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.15387840270996095
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,64,4096,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.06637120246887207
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,128,8192,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.1552448034286499
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,128,8192,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.08551999926567078
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,1,1,1,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.006777600198984146
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,256,16384,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.15605440139770507
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,256,16384,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.1511199951171875
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,512,32768,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.16631040573120118
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,512,32768,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.2898655891418457
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,1,1,1,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.005628800019621849
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,1024,65536,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.178166401386261
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,1024,65536,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.5563168048858642
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,2048,131072,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.29941439628601074
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,2,1,2,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.009296000003814697
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,2048,131072,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,1.0882111549377442
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,2,1,2,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.00697920024394989
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,4096,262144,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.5424287796020508
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,4096,262144,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,2.148214340209961
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,4,1,4,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.00958079993724823
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,4,1,4,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.008777599781751633
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,8192,524288,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,1.0322336196899413
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,8192,524288,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,4.253081512451172
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,8,1,8,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.008928000181913375
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,8,1,8,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.008108799904584884
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,16,1,16,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.009084799885749817
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_fn,context,64,16384,1048576,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,2.012719917297363
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,16,1,16,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.008160000294446945
TRTLLM,1.1.0,NVIDIA GB200,mamba2,mamba_chunk_scan_combined,context,64,16384,1048576,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,8.45544662475586
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,32,1,32,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.009344000369310379
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,32,1,32,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.010892800241708755
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,64,1,64,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.009302400052547455
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,64,1,64,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.015798400342464446
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,128,1,128,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.008582399785518646
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,128,1,128,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.025862398743629455
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,256,1,256,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.00971520021557808
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,256,1,256,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.04625279903411865
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,512,1,512,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.013545599579811097
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,512,1,512,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.0871936023235321
TRTLLM,1.1.0,NVIDIA GB200,mamba2,causal_conv1d_update,generation,1024,1,1024,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.01964160054922104
TRTLLM,1.1.0,NVIDIA GB200,mamba2,selective_state_update,generation,1024,1,1024,1024,64,4,16,64,4,128,MAMBA2_GENERIC_1K,0.16902400255203248
