framework,version,device,op_name,kernel_source,moe_dtype,num_tokens,hidden_size,topk,num_experts,moe_ep_size,distribution,latency
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1,7168,8,256,2,balanced,0.06271679885685444
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1,7168,8,256,2,balanced,0.0553056001663208
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1,7168,8,256,2,balanced,0.0740959994494915
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1,7168,8,256,2,balanced,0.06200000047683716
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1,7168,8,256,2,balanced,0.05024640001356602
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1,7168,8,256,2,balanced,0.07617599964141845
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1,7168,8,256,2,balanced,0.07222400084137917
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1,7168,8,256,2,balanced,0.062111999839544296
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1,7168,8,256,2,balanced,0.07550400048494339
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1,7168,8,256,2,balanced,0.07373439967632293
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,2,7168,8,256,2,balanced,0.07334719970822334
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,2,7168,8,256,2,balanced,0.05996800065040588
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,2,7168,8,256,2,balanced,0.07353279888629913
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,2,7168,8,256,2,balanced,0.06476799957454205
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,2,7168,8,256,2,balanced,0.06210239939391613
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,2,7168,8,256,2,balanced,0.07328319847583771
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,2,7168,8,256,2,balanced,0.06524159908294677
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,2,7168,8,256,2,balanced,0.062403199076652524
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,2,7168,8,256,2,balanced,0.07697280049324036
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,2,7168,8,256,2,balanced,0.07076799906790257
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,4,7168,8,256,2,balanced,0.06818880066275597
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,4,7168,8,256,2,balanced,0.055052799731493
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,4,7168,8,256,2,balanced,0.0676223985850811
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,4,7168,8,256,2,balanced,0.06075199991464615
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,4,7168,8,256,2,balanced,0.05310079976916313
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,4,7168,8,256,2,balanced,0.0689312007278204
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,4,7168,8,256,2,balanced,0.06445440016686917
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,4,7168,8,256,2,balanced,0.05757439993321896
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,4,7168,8,256,2,balanced,0.06877120062708855
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,4,7168,8,256,2,balanced,0.07174079865217209
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,8,7168,8,256,2,balanced,0.06010239981114864
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,8,7168,8,256,2,balanced,0.052054399996995925
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,8,7168,8,256,2,balanced,0.07213120087981224
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,8,7168,8,256,2,balanced,0.06456960029900075
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,8,7168,8,256,2,balanced,0.051587199419736864
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,8,7168,8,256,2,balanced,0.06813440062105655
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,8,7168,8,256,2,balanced,0.05733120031654835
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,8,7168,8,256,2,balanced,0.05801919996738434
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,8,7168,8,256,2,balanced,0.06868479922413825
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,8,7168,8,256,2,balanced,0.06993279904127121
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,16,7168,8,256,2,balanced,0.05991039909422398
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,16,7168,8,256,2,balanced,0.050499200075864795
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,16,7168,8,256,2,balanced,0.07621760070323944
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,16,7168,8,256,2,balanced,0.0582847997546196
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,16,7168,8,256,2,balanced,0.04838079959154129
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,16,7168,8,256,2,balanced,0.07004800029098987
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,16,7168,8,256,2,balanced,0.05823359936475754
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,16,7168,8,256,2,balanced,0.052054399624466896
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,16,7168,8,256,2,balanced,0.0692768007516861
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,16,7168,8,256,2,balanced,0.06577279940247535
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,32,7168,8,256,2,balanced,0.05593919977545738
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,32,7168,8,256,2,balanced,0.04832639954984188
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,32,7168,8,256,2,balanced,0.07588799893856049
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,32,7168,8,256,2,balanced,0.05512959994375706
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,32,7168,8,256,2,balanced,0.0459136001765728
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,32,7168,8,256,2,balanced,0.07645120024681092
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,32,7168,8,256,2,balanced,0.05273279882967472
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,32,7168,8,256,2,balanced,0.045081600174307825
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,32,7168,8,256,2,balanced,0.07292799949645996
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,32,7168,8,256,2,balanced,0.0759936012327671
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,48,7168,8,256,2,balanced,0.05714880041778088
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,48,7168,8,256,2,balanced,0.04428480006754398
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,48,7168,8,256,2,balanced,0.07765759974718094
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,48,7168,8,256,2,balanced,0.04962879978120327
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,48,7168,8,256,2,balanced,0.04587519988417625
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,48,7168,8,256,2,balanced,0.06945920065045356
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,48,7168,8,256,2,balanced,0.05150079987943172
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,48,7168,8,256,2,balanced,0.04634560011327267
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,48,7168,8,256,2,balanced,0.07634880021214485
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,48,7168,8,256,2,balanced,0.06862720027565956
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,64,7168,8,256,2,balanced,0.051583999395370485
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,64,7168,8,256,2,balanced,0.044076799601316455
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,64,7168,8,256,2,balanced,0.07356160134077072
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,64,7168,8,256,2,balanced,0.051654399186372754
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,64,7168,8,256,2,balanced,0.04197440035641194
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,64,7168,8,256,2,balanced,0.07317759916186332
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,64,7168,8,256,2,balanced,0.056249599903821945
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,64,7168,8,256,2,balanced,0.04718399979174137
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,64,7168,8,256,2,balanced,0.07696959972381592
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,64,7168,8,256,2,balanced,0.07195839881896973
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,80,7168,8,256,2,balanced,0.05490560084581375
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,80,7168,8,256,2,balanced,0.043516799435019494
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,80,7168,8,256,2,balanced,0.06640319973230362
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,80,7168,8,256,2,balanced,0.05090240016579628
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,80,7168,8,256,2,balanced,0.045081600174307825
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,80,7168,8,256,2,balanced,0.06848319955170154
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,80,7168,8,256,2,balanced,0.051615999266505244
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,80,7168,8,256,2,balanced,0.046041599661111834
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,80,7168,8,256,2,balanced,0.06976000033318996
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,80,7168,8,256,2,balanced,0.0742976002395153
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,96,7168,8,256,2,balanced,0.0500160001218319
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,96,7168,8,256,2,balanced,0.04352639988064766
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,96,7168,8,256,2,balanced,0.07051840014755725
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,96,7168,8,256,2,balanced,0.054371199756860736
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,96,7168,8,256,2,balanced,0.04521919973194599
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,96,7168,8,256,2,balanced,0.07757120057940484
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,96,7168,8,256,2,balanced,0.05167359970510006
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,96,7168,8,256,2,balanced,0.04768000058829784
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,96,7168,8,256,2,balanced,0.07442880123853683
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,96,7168,8,256,2,balanced,0.07338559925556183
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,128,7168,8,256,2,balanced,0.05216319970786572
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,128,7168,8,256,2,balanced,0.044428800046443936
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,128,7168,8,256,2,balanced,0.0689856007695198
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,128,7168,8,256,2,balanced,0.05160319954156876
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,128,7168,8,256,2,balanced,0.04278399981558323
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,128,7168,8,256,2,balanced,0.07678080052137375
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,128,7168,8,256,2,balanced,0.05169919990003109
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,128,7168,8,256,2,balanced,0.044556799903512
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,128,7168,8,256,2,balanced,0.0763839989900589
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,128,7168,8,256,2,balanced,0.07607039958238601
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,160,7168,8,256,2,balanced,0.05051839984953403
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,160,7168,8,256,2,balanced,0.04314879961311817
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,160,7168,8,256,2,balanced,0.07455359920859336
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,160,7168,8,256,2,balanced,0.05388480015099049
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,160,7168,8,256,2,balanced,0.04381760023534298
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,160,7168,8,256,2,balanced,0.07172480039298534
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,160,7168,8,256,2,balanced,0.05273599959909916
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,160,7168,8,256,2,balanced,0.04366080015897751
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,160,7168,8,256,2,balanced,0.06952640041708946
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,160,7168,8,256,2,balanced,0.07504000067710877
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,192,7168,8,256,2,balanced,0.04985600002110004
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,192,7168,8,256,2,balanced,0.043193599209189415
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,192,7168,8,256,2,balanced,0.06813440062105655
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,192,7168,8,256,2,balanced,0.051343999803066254
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,192,7168,8,256,2,balanced,0.04144319966435432
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,192,7168,8,256,2,balanced,0.07310400009155274
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,192,7168,8,256,2,balanced,0.04798079989850521
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,192,7168,8,256,2,balanced,0.044572800025343896
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,192,7168,8,256,2,balanced,0.06487360037863255
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,192,7168,8,256,2,balanced,0.06280000023543834
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,256,7168,8,256,2,balanced,0.06218560002744198
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,256,7168,8,256,2,balanced,0.053862399980425836
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,256,7168,8,256,2,balanced,0.08071360141038894
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,256,7168,8,256,2,balanced,0.05586880035698414
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,256,7168,8,256,2,balanced,0.049904000014066696
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,256,7168,8,256,2,balanced,0.07367039918899536
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,256,7168,8,256,2,balanced,0.06022399961948395
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,256,7168,8,256,2,balanced,0.05045120045542717
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,256,7168,8,256,2,balanced,0.07400959953665734
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,256,7168,8,256,2,balanced,0.07608319967985153
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,320,7168,8,256,2,balanced,0.05751040056347847
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,320,7168,8,256,2,balanced,0.050016000494360925
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,320,7168,8,256,2,balanced,0.07291199937462807
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,320,7168,8,256,2,balanced,0.052287999913096427
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,320,7168,8,256,2,balanced,0.04664639979600906
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,320,7168,8,256,2,balanced,0.07143040001392365
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,320,7168,8,256,2,balanced,0.052566399425268175
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,320,7168,8,256,2,balanced,0.048268799856305125
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,320,7168,8,256,2,balanced,0.07637119889259339
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,320,7168,8,256,2,balanced,0.07637120112776756
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,384,7168,8,256,2,balanced,0.0559712003916502
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,384,7168,8,256,2,balanced,0.0483424000442028
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,384,7168,8,256,2,balanced,0.07270080000162124
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,384,7168,8,256,2,balanced,0.056931200250983235
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,384,7168,8,256,2,balanced,0.04731519967317581
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,384,7168,8,256,2,balanced,0.07401279956102372
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,384,7168,8,256,2,balanced,0.05713920034468174
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,384,7168,8,256,2,balanced,0.050057600066065785
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,384,7168,8,256,2,balanced,0.07472000047564506
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,384,7168,8,256,2,balanced,0.07685439884662629
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,512,7168,8,256,2,balanced,0.058300799876451495
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,512,7168,8,256,2,balanced,0.04918080009520054
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,512,7168,8,256,2,balanced,0.07344960048794746
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,512,7168,8,256,2,balanced,0.05678720027208328
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,512,7168,8,256,2,balanced,0.04935359992086887
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,512,7168,8,256,2,balanced,0.07483520060777664
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,512,7168,8,256,2,balanced,0.05305279977619648
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,512,7168,8,256,2,balanced,0.04900799989700318
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,512,7168,8,256,2,balanced,0.0716320000588894
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,512,7168,8,256,2,balanced,0.08660160005092621
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,768,7168,8,256,2,balanced,0.054252799972891806
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,768,7168,8,256,2,balanced,0.05475519932806492
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,768,7168,8,256,2,balanced,0.09063360020518303
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,768,7168,8,256,2,balanced,0.05498240031301975
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,768,7168,8,256,2,balanced,0.049315200001001355
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,768,7168,8,256,2,balanced,0.08973760008811951
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,768,7168,8,256,2,balanced,0.05602880045771599
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,768,7168,8,256,2,balanced,0.05688959956169128
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,768,7168,8,256,2,balanced,0.08971840143203735
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,768,7168,8,256,2,balanced,0.11415040120482445
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1024,7168,8,256,2,balanced,0.05144319981336594
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1024,7168,8,256,2,balanced,0.06066240072250366
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1024,7168,8,256,2,balanced,0.10530559942126275
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1024,7168,8,256,2,balanced,0.05524800010025501
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1024,7168,8,256,2,balanced,0.0529823999851942
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1024,7168,8,256,2,balanced,0.1023424006998539
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1024,7168,8,256,2,balanced,0.050911999121308324
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1024,7168,8,256,2,balanced,0.05774080008268356
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1024,7168,8,256,2,balanced,0.10363199934363365
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1024,7168,8,256,2,balanced,0.13782399892807007
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1536,7168,8,256,2,balanced,0.053359999880194665
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1536,7168,8,256,2,balanced,0.0762304000556469
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1536,7168,8,256,2,balanced,0.13417279869318008
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1536,7168,8,256,2,balanced,0.05664959996938705
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1536,7168,8,256,2,balanced,0.06289280094206333
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1536,7168,8,256,2,balanced,0.13363840132951738
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1536,7168,8,256,2,balanced,0.051628800481557845
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1536,7168,8,256,2,balanced,0.06632639914751053
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1536,7168,8,256,2,balanced,0.13316799998283385
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1536,7168,8,256,2,balanced,0.17704000025987626
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,2048,7168,8,256,2,balanced,0.05382080003619194
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,2048,7168,8,256,2,balanced,0.08853119984269142
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,2048,7168,8,256,2,balanced,0.16324480026960372
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,2048,7168,8,256,2,balanced,0.05698239989578724
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,2048,7168,8,256,2,balanced,0.07129919901490211
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,2048,7168,8,256,2,balanced,0.1611775979399681
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,2048,7168,8,256,2,balanced,0.055296000093221664
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,2048,7168,8,256,2,balanced,0.07615040019154548
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,2048,7168,8,256,2,balanced,0.1627008005976677
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,2048,7168,8,256,2,balanced,0.2197855979204178
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,3072,7168,8,256,2,balanced,0.06087360009551048
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,3072,7168,8,256,2,balanced,0.12037120014429092
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,3072,7168,8,256,2,balanced,0.22035199999809266
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,3072,7168,8,256,2,balanced,0.061008000001311305
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,3072,7168,8,256,2,balanced,0.09140480160713196
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,3072,7168,8,256,2,balanced,0.21928320080041885
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,3072,7168,8,256,2,balanced,0.061532800644636156
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,3072,7168,8,256,2,balanced,0.09713279977440833
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,3072,7168,8,256,2,balanced,0.21852799952030183
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,3072,7168,8,256,2,balanced,0.30819199681282045
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,4096,7168,8,256,2,balanced,0.06823039948940277
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,4096,7168,8,256,2,balanced,0.14822400063276292
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,4096,7168,8,256,2,balanced,0.2888896018266678
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,4096,7168,8,256,2,balanced,0.0687648005783558
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,4096,7168,8,256,2,balanced,0.11672319993376731
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,4096,7168,8,256,2,balanced,0.2789248049259186
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,4096,7168,8,256,2,balanced,0.06862719953060151
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,4096,7168,8,256,2,balanced,0.11989760026335716
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,4096,7168,8,256,2,balanced,0.27874560058116915
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,4096,7168,8,256,2,balanced,0.393027201294899
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,6144,7168,8,256,2,balanced,0.0810591995716095
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,6144,7168,8,256,2,balanced,0.20437439978122712
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,6144,7168,8,256,2,balanced,0.39664640128612516
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,6144,7168,8,256,2,balanced,0.08231360018253327
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,6144,7168,8,256,2,balanced,0.16828480064868928
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,6144,7168,8,256,2,balanced,0.3934688001871109
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,6144,7168,8,256,2,balanced,0.08046400025486947
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,6144,7168,8,256,2,balanced,0.15996479839086533
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,6144,7168,8,256,2,balanced,0.3924223929643631
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,6144,7168,8,256,2,balanced,0.5660032033920288
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,8192,7168,8,256,2,balanced,0.09397440105676651
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,8192,7168,8,256,2,balanced,0.2585376024246216
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,8192,7168,8,256,2,balanced,0.5072704076766967
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,8192,7168,8,256,2,balanced,0.09154879897832871
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,8192,7168,8,256,2,balanced,0.2108287990093231
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,8192,7168,8,256,2,balanced,0.5047232002019882
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,8192,7168,8,256,2,balanced,0.09426879957318306
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,8192,7168,8,256,2,balanced,0.202463997900486
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,8192,7168,8,256,2,balanced,0.5048000067472458
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,8192,7168,8,256,2,balanced,0.7363488018512726
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1,7168,8,256,4,balanced,0.056614399701356885
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1,7168,8,256,4,balanced,0.04973760023713112
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1,7168,8,256,4,balanced,0.07206400036811829
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1,7168,8,256,4,balanced,0.06242239996790886
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1,7168,8,256,4,balanced,0.046332800015807155
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1,7168,8,256,4,balanced,0.07660480067133904
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1,7168,8,256,4,balanced,0.06167999990284443
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1,7168,8,256,4,balanced,0.05214719995856285
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1,7168,8,256,4,balanced,0.0758943997323513
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1,7168,8,256,4,balanced,0.07315840050578118
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,2,7168,8,256,4,balanced,0.05631359927356243
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,2,7168,8,256,4,balanced,0.05004799962043762
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,2,7168,8,256,4,balanced,0.07380799949169159
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,2,7168,8,256,4,balanced,0.05574080124497414
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,2,7168,8,256,4,balanced,0.04642880037426948
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,2,7168,8,256,4,balanced,0.0723583996295929
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,2,7168,8,256,4,balanced,0.058611200749874116
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,2,7168,8,256,4,balanced,0.04787520058453083
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,2,7168,8,256,4,balanced,0.07218879908323288
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,2,7168,8,256,4,balanced,0.06956800073385239
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,4,7168,8,256,4,balanced,0.05353280007839203
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,4,7168,8,256,4,balanced,0.046582399681210515
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,4,7168,8,256,4,balanced,0.06839359998703003
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,4,7168,8,256,4,balanced,0.05445760004222393
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,4,7168,8,256,4,balanced,0.044697600230574605
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,4,7168,8,256,4,balanced,0.07248640060424805
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,4,7168,8,256,4,balanced,0.05138559974730015
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,4,7168,8,256,4,balanced,0.046483200415968894
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,4,7168,8,256,4,balanced,0.06917440071702004
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,4,7168,8,256,4,balanced,0.06640640087425709
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,8,7168,8,256,4,balanced,0.05230400003492832
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,8,7168,8,256,4,balanced,0.04440640024840832
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,8,7168,8,256,4,balanced,0.06932160034775733
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,12288,7168,8,256,2,balanced,0.12104320004582406
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,12288,7168,8,256,2,balanced,0.36890239715576173
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,8,7168,8,256,4,balanced,0.052108800411224364
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,12288,7168,8,256,2,balanced,0.7338111937046051
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,8,7168,8,256,4,balanced,0.04643200002610683
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,8,7168,8,256,4,balanced,0.05973440036177635
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,8,7168,8,256,4,balanced,0.05450559966266155
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,8,7168,8,256,4,balanced,0.053571199625730516
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,8,7168,8,256,4,balanced,0.06413759961724282
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,8,7168,8,256,4,balanced,0.06438080035150051
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,16,7168,8,256,4,balanced,0.052553600072860716
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,16,7168,8,256,4,balanced,0.046719999983906746
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,16,7168,8,256,4,balanced,0.07795199975371361
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,16,7168,8,256,4,balanced,0.05204159989953041
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,16,7168,8,256,4,balanced,0.041715200245380404
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,16,7168,8,256,4,balanced,0.06440959982573986
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,16,7168,8,256,4,balanced,0.052137600257992744
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,16,7168,8,256,4,balanced,0.04784000031650067
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,16,7168,8,256,4,balanced,0.06969280019402505
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,16,7168,8,256,4,balanced,0.06615680009126663
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,32,7168,8,256,4,balanced,0.05393600016832352
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,32,7168,8,256,4,balanced,0.04279039986431599
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,32,7168,8,256,4,balanced,0.06824960038065911
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,32,7168,8,256,4,balanced,0.052383999526500705
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,32,7168,8,256,4,balanced,0.04839680008590221
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,32,7168,8,256,4,balanced,0.06855679973959923
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,32,7168,8,256,4,balanced,0.05236479975283146
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,32,7168,8,256,4,balanced,0.045001599565148355
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,32,7168,8,256,4,balanced,0.06993919946253299
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,32,7168,8,256,4,balanced,0.06784000024199485
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,48,7168,8,256,4,balanced,0.057747199386358264
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,48,7168,8,256,4,balanced,0.04701439961791039
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,48,7168,8,256,4,balanced,0.0661248005926609
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,48,7168,8,256,4,balanced,0.05668800063431263
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,48,7168,8,256,4,balanced,0.043648000061511996
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,48,7168,8,256,4,balanced,0.06754560135304928
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,48,7168,8,256,4,balanced,0.051734399795532224
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,48,7168,8,256,4,balanced,0.04378559961915016
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,48,7168,8,256,4,balanced,0.06678399965167045
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,48,7168,8,256,4,balanced,0.06561599932610988
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,64,7168,8,256,4,balanced,0.05275519974529743
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,64,7168,8,256,4,balanced,0.041420799866318705
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,64,7168,8,256,4,balanced,0.0653087992221117
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,64,7168,8,256,4,balanced,0.05493759959936142
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,64,7168,8,256,4,balanced,0.041360000148415565
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,64,7168,8,256,4,balanced,0.06622399985790253
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1,7168,8,256,8,balanced,0.07006720080971718
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1,7168,8,256,8,balanced,0.05698879994452
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,64,7168,8,256,4,balanced,0.05239679962396622
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,64,7168,8,256,4,balanced,0.04301760047674179
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,64,7168,8,256,4,balanced,0.06464959979057312
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,64,7168,8,256,4,balanced,0.06904639974236489
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,80,7168,8,256,4,balanced,0.0588832002133131
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,80,7168,8,256,4,balanced,0.045654400065541265
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,80,7168,8,256,4,balanced,0.06897279992699623
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,80,7168,8,256,4,balanced,0.05392640009522438
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,80,7168,8,256,4,balanced,0.041075199842453
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,80,7168,8,256,4,balanced,0.06760639995336533
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,80,7168,8,256,4,balanced,0.05218240059912205
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,80,7168,8,256,4,balanced,0.04208959974348545
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,80,7168,8,256,4,balanced,0.0701375987380743
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,80,7168,8,256,4,balanced,0.0700095996260643
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,96,7168,8,256,4,balanced,0.052937600389122964
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,96,7168,8,256,4,balanced,0.043190399557352065
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,96,7168,8,256,4,balanced,0.07285760045051574
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,96,7168,8,256,4,balanced,0.0514335997402668
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,96,7168,8,256,4,balanced,0.04089280031621456
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,96,7168,8,256,4,balanced,0.06760000027716159
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,96,7168,8,256,4,balanced,0.05046719908714294
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,96,7168,8,256,4,balanced,0.04173760004341602
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,96,7168,8,256,4,balanced,0.06937279999256134
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,96,7168,8,256,4,balanced,0.06833279915153981
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,128,7168,8,256,4,balanced,0.05289919972419739
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,128,7168,8,256,4,balanced,0.04283199980854988
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,128,7168,8,256,4,balanced,0.06693440042436123
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,128,7168,8,256,4,balanced,0.05488640069961548
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,128,7168,8,256,4,balanced,0.04132799990475178
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,128,7168,8,256,4,balanced,0.07028160020709037
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,128,7168,8,256,4,balanced,0.052003201097249985
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,128,7168,8,256,4,balanced,0.04493759945034981
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,128,7168,8,256,4,balanced,0.06338240019977093
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,128,7168,8,256,4,balanced,0.06514879949390888
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,160,7168,8,256,4,balanced,0.05485759936273098
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,160,7168,8,256,4,balanced,0.043644799292087554
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,160,7168,8,256,4,balanced,0.06554880067706108
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,160,7168,8,256,4,balanced,0.053619200363755226
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,160,7168,8,256,4,balanced,0.04122560061514378
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,160,7168,8,256,4,balanced,0.06538560055196285
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,160,7168,8,256,4,balanced,0.05192000009119511
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,160,7168,8,256,4,balanced,0.043753599375486375
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,160,7168,8,256,4,balanced,0.06597119942307472
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,160,7168,8,256,4,balanced,0.0700543999671936
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,192,7168,8,256,4,balanced,0.05305600017309189
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,192,7168,8,256,4,balanced,0.043952000141143796
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,192,7168,8,256,4,balanced,0.06779840104281902
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,192,7168,8,256,4,balanced,0.05164160020649433
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,192,7168,8,256,4,balanced,0.041068800166249274
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,192,7168,8,256,4,balanced,0.06447680108249187
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,192,7168,8,256,4,balanced,0.053593600168824196
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,192,7168,8,256,4,balanced,0.04432320035994053
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,192,7168,8,256,4,balanced,0.06591360047459602
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,192,7168,8,256,4,balanced,0.06482239887118339
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,256,7168,8,256,4,balanced,0.053814399614930154
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,256,7168,8,256,4,balanced,0.04407680034637451
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,256,7168,8,256,4,balanced,0.07839679941534997
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,256,7168,8,256,4,balanced,0.05257919952273369
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,256,7168,8,256,4,balanced,0.04021120071411133
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,256,7168,8,256,4,balanced,0.06517439968883991
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,256,7168,8,256,4,balanced,0.04863360039889812
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,256,7168,8,256,4,balanced,0.04396160021424293
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,256,7168,8,256,4,balanced,0.06584319993853568
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,256,7168,8,256,4,balanced,0.06699520014226437
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,320,7168,8,256,4,balanced,0.05001279972493648
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,320,7168,8,256,4,balanced,0.04329920038580894
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,320,7168,8,256,4,balanced,0.06516160070896149
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,320,7168,8,256,4,balanced,0.05103360041975975
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,320,7168,8,256,4,balanced,0.04549439996480942
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,320,7168,8,256,4,balanced,0.06935039944946766
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,320,7168,8,256,4,balanced,0.052768000215291974
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,320,7168,8,256,4,balanced,0.045443199947476386
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,320,7168,8,256,4,balanced,0.06658880077302456
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,320,7168,8,256,4,balanced,0.06618240028619767
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,384,7168,8,256,4,balanced,0.05440640039741993
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,384,7168,8,256,4,balanced,0.04495680034160614
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,384,7168,8,256,4,balanced,0.07152320072054863
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,384,7168,8,256,4,balanced,0.04766719974577427
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,384,7168,8,256,4,balanced,0.04168320000171662
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,384,7168,8,256,4,balanced,0.06219520084559917
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,384,7168,8,256,4,balanced,0.053692799806594846
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,384,7168,8,256,4,balanced,0.04443200044333935
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,384,7168,8,256,4,balanced,0.06756160072982312
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,384,7168,8,256,4,balanced,0.06970880031585694
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,512,7168,8,256,4,balanced,0.05349759981036186
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,512,7168,8,256,4,balanced,0.04576000012457371
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,512,7168,8,256,4,balanced,0.06919039860367775
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1,7168,8,256,8,balanced,0.07264960035681725
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1,7168,8,256,8,balanced,0.06739520020782948
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1,7168,8,256,8,balanced,0.05926399901509285
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1,7168,8,256,8,balanced,0.07278079986572265
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1,7168,8,256,8,balanced,0.06864320039749146
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1,7168,8,256,8,balanced,0.11771840006113052
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1,7168,8,256,8,balanced,0.09365119934082031
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1,7168,8,256,8,balanced,0.13347200006246568
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,2,7168,8,256,8,balanced,0.10450559929013252
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,2,7168,8,256,8,balanced,0.09023999944329261
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,2,7168,8,256,8,balanced,0.0796671986579895
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,512,7168,8,256,4,balanced,0.05313599966466427
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,2,7168,8,256,8,balanced,0.097104000300169
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,512,7168,8,256,4,balanced,0.04344319961965084
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,2,7168,8,256,8,balanced,0.08143360018730164
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,512,7168,8,256,4,balanced,0.07111679911613464
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,2,7168,8,256,8,balanced,0.08256960026919842
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,2,7168,8,256,8,balanced,0.09629439860582352
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,2,7168,8,256,8,balanced,0.07764159962534904
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,2,7168,8,256,8,balanced,0.07848320007324219
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,2,7168,8,256,8,balanced,0.10241279974579812
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,4,7168,8,256,8,balanced,0.09182079955935478
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,4,7168,8,256,8,balanced,0.07958079949021339
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,4,7168,8,256,8,balanced,0.07624959982931614
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,4,7168,8,256,8,balanced,0.09050879925489426
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,4,7168,8,256,8,balanced,0.0756831981241703
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,4,7168,8,256,8,balanced,0.07532479912042618
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,4,7168,8,256,8,balanced,0.08328959867358207
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,4,7168,8,256,8,balanced,0.07206079922616482
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,4,7168,8,256,8,balanced,0.07388480007648468
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,4,7168,8,256,8,balanced,0.09943039938807488
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,8,7168,8,256,8,balanced,0.0883360005915165
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,8,7168,8,256,8,balanced,0.07350080013275147
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,512,7168,8,256,4,balanced,0.05421439930796623
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,512,7168,8,256,4,balanced,0.0450655996799469
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,512,7168,8,256,4,balanced,0.07041279971599579
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,512,7168,8,256,4,balanced,0.07289280071854591
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,12288,7168,8,256,2,balanced,0.12112959921360016
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,12288,7168,8,256,2,balanced,0.3009503960609436
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,12288,7168,8,256,2,balanced,0.7328447997570038
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,768,7168,8,256,4,balanced,0.051238400116562846
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,768,7168,8,256,4,balanced,0.05137280039489269
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,768,7168,8,256,4,balanced,0.08040319979190827
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,768,7168,8,256,4,balanced,0.05226880051195622
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,768,7168,8,256,4,balanced,0.042422400414943696
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,768,7168,8,256,4,balanced,0.08033600151538849
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,768,7168,8,256,4,balanced,0.04887680001556873
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,768,7168,8,256,4,balanced,0.04416319988667965
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,768,7168,8,256,4,balanced,0.0807839997112751
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,768,7168,8,256,4,balanced,0.09015680104494095
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1024,7168,8,256,4,balanced,0.05596799999475479
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1024,7168,8,256,4,balanced,0.05647360011935234
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1024,7168,8,256,4,balanced,0.09389120042324066
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1024,7168,8,256,4,balanced,0.05644800066947937
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1024,7168,8,256,4,balanced,0.04443520046770573
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1024,7168,8,256,4,balanced,0.09379520043730735
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1024,7168,8,256,4,balanced,0.054345599934458734
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1024,7168,8,256,4,balanced,0.04613120034337044
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1024,7168,8,256,4,balanced,0.09521600008010864
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1024,7168,8,256,4,balanced,0.10609280094504356
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1536,7168,8,256,4,balanced,0.05379520058631897
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1536,7168,8,256,4,balanced,0.06726720035076142
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1536,7168,8,256,4,balanced,0.1216511994600296
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,8,7168,8,256,8,balanced,0.07770240008831024
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,8,7168,8,256,8,balanced,0.09441279992461205
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,8,7168,8,256,8,balanced,0.07411839962005615
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,8,7168,8,256,8,balanced,0.07552959993481637
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,8,7168,8,256,8,balanced,0.08939200043678283
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,8,7168,8,256,8,balanced,0.07320320010185241
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,8,7168,8,256,8,balanced,0.07575680091977119
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,8,7168,8,256,8,balanced,0.0955520000308752
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,16,7168,8,256,8,balanced,0.08606079891324044
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,16,7168,8,256,8,balanced,0.07954559922218322
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,16,7168,8,256,8,balanced,0.11665279865264892
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,16,7168,8,256,8,balanced,0.08244160041213036
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,16,7168,8,256,8,balanced,0.07255359962582589
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,16,7168,8,256,8,balanced,0.11205440014600754
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,16,7168,8,256,8,balanced,0.0803007997572422
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,16,7168,8,256,8,balanced,0.07192960008978844
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,16,7168,8,256,8,balanced,0.11240000054240226
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,16,7168,8,256,8,balanced,0.10782080069184304
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,32,7168,8,256,8,balanced,0.09304320141673088
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,32,7168,8,256,8,balanced,0.0732512004673481
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,32,7168,8,256,8,balanced,0.1170975998044014
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,32,7168,8,256,8,balanced,0.07835199981927872
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,32,7168,8,256,8,balanced,0.07133120000362396
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,32,7168,8,256,8,balanced,0.11700799986720085
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,32,7168,8,256,8,balanced,0.08598080053925514
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,32,7168,8,256,8,balanced,0.06998720094561577
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,32,7168,8,256,8,balanced,0.10780799984931946
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,32,7168,8,256,8,balanced,0.10362560003995895
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,48,7168,8,256,8,balanced,0.08106560036540031
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,48,7168,8,256,8,balanced,0.06873920112848282
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,48,7168,8,256,8,balanced,0.11126400008797646
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1536,7168,8,256,4,balanced,0.05175360031425953
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1536,7168,8,256,4,balanced,0.05026560053229332
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1536,7168,8,256,4,balanced,0.12001279890537261
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,48,7168,8,256,8,balanced,0.07787200100719929
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,48,7168,8,256,8,balanced,0.07216959930956364
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,48,7168,8,256,8,balanced,0.11240640133619309
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,48,7168,8,256,8,balanced,0.07470400035381317
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,48,7168,8,256,8,balanced,0.06954879947006702
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,48,7168,8,256,8,balanced,0.11360320001840592
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,48,7168,8,256,8,balanced,0.10755839869379998
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,64,7168,8,256,8,balanced,0.08225279971957207
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,64,7168,8,256,8,balanced,0.06681920066475869
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,64,7168,8,256,8,balanced,0.10742719992995262
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,64,7168,8,256,8,balanced,0.0764320008456707
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,64,7168,8,256,8,balanced,0.06878399923443794
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,64,7168,8,256,8,balanced,0.1156800001859665
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,64,7168,8,256,8,balanced,0.07927359938621521
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,64,7168,8,256,8,balanced,0.0701311994343996
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,64,7168,8,256,8,balanced,0.10778240114450455
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,64,7168,8,256,8,balanced,0.10824000015854836
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,80,7168,8,256,8,balanced,0.07571200020611286
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,80,7168,8,256,8,balanced,0.06741119958460332
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,80,7168,8,256,8,balanced,0.0989376001060009
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,80,7168,8,256,8,balanced,0.07271040044724941
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,80,7168,8,256,8,balanced,0.07070400044322014
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,80,7168,8,256,8,balanced,0.1012991987168789
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,80,7168,8,256,8,balanced,0.08053760081529618
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,80,7168,8,256,8,balanced,0.06803520023822784
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,80,7168,8,256,8,balanced,0.1093120001256466
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,80,7168,8,256,8,balanced,0.10621760040521622
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,96,7168,8,256,8,balanced,0.0750815998762846
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,96,7168,8,256,8,balanced,0.07019519954919815
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,96,7168,8,256,8,balanced,0.11039680093526841
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1536,7168,8,256,4,balanced,0.05093120038509369
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1536,7168,8,256,4,balanced,0.05403199978172779
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1536,7168,8,256,4,balanced,0.11761600002646447
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1536,7168,8,256,4,balanced,0.13665599822998048
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,96,7168,8,256,8,balanced,0.08098559975624084
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,96,7168,8,256,8,balanced,0.06614400036633014
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,96,7168,8,256,8,balanced,0.11087680086493493
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,96,7168,8,256,8,balanced,0.07320319935679435
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,96,7168,8,256,8,balanced,0.0723040010780096
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,96,7168,8,256,8,balanced,0.1146879993379116
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,96,7168,8,256,8,balanced,0.10316800028085708
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,128,7168,8,256,8,balanced,0.07178240045905113
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,128,7168,8,256,8,balanced,0.07003519907593728
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,128,7168,8,256,8,balanced,0.11483200043439865
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,128,7168,8,256,8,balanced,0.0809663999825716
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,128,7168,8,256,8,balanced,0.06964799985289574
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,128,7168,8,256,8,balanced,0.11391680091619491
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,128,7168,8,256,8,balanced,0.0744703996926546
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,128,7168,8,256,8,balanced,0.06503680013120175
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,128,7168,8,256,8,balanced,0.11706240102648735
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,128,7168,8,256,8,balanced,0.09984640032052994
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,12288,7168,8,256,2,balanced,0.12015039920806884
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,12288,7168,8,256,2,balanced,0.2860159993171692
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,160,7168,8,256,8,balanced,0.07643840163946151
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,160,7168,8,256,8,balanced,0.06923840008676052
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,160,7168,8,256,8,balanced,0.10807999894022942
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,160,7168,8,256,8,balanced,0.0706144005060196
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,160,7168,8,256,8,balanced,0.06657919883728028
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,160,7168,8,256,8,balanced,0.10103999972343444
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,160,7168,8,256,8,balanced,0.07362880110740662
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,160,7168,8,256,8,balanced,0.06624959893524647
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,160,7168,8,256,8,balanced,0.11375679820775986
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,160,7168,8,256,8,balanced,0.10343680083751679
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,2048,7168,8,256,4,balanced,0.055769600346684454
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,2048,7168,8,256,4,balanced,0.0761920005083084
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,2048,7168,8,256,4,balanced,0.14494719952344895
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,192,7168,8,256,8,balanced,0.0779424000531435
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,192,7168,8,256,8,balanced,0.08991359993815422
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,192,7168,8,256,8,balanced,0.10679040029644966
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,192,7168,8,256,8,balanced,0.07147839926183223
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,192,7168,8,256,8,balanced,0.07096319906413555
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,192,7168,8,256,8,balanced,0.10503040105104447
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,192,7168,8,256,8,balanced,0.07226879969239235
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,192,7168,8,256,8,balanced,0.07399680018424988
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,192,7168,8,256,8,balanced,0.10297920033335686
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,192,7168,8,256,8,balanced,0.10559359937906265
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,256,7168,8,256,8,balanced,0.07252799943089486
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,256,7168,8,256,8,balanced,0.07912959977984428
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,256,7168,8,256,8,balanced,0.1100608006119728
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,256,7168,8,256,8,balanced,0.07390079870820046
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,256,7168,8,256,8,balanced,0.06354239992797375
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,256,7168,8,256,8,balanced,0.10828480049967766
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,256,7168,8,256,8,balanced,0.0725823998451233
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,256,7168,8,256,8,balanced,0.06833919994533062
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,256,7168,8,256,8,balanced,0.11319680064916611
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,256,7168,8,256,8,balanced,0.10568000003695488
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,2048,7168,8,256,4,balanced,0.0534911997616291
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,2048,7168,8,256,4,balanced,0.05467199981212616
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,2048,7168,8,256,4,balanced,0.14435200095176698
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,320,7168,8,256,8,balanced,0.07380159944295883
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,320,7168,8,256,8,balanced,0.07542399987578392
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,320,7168,8,256,8,balanced,0.11555840000510216
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,320,7168,8,256,8,balanced,0.07096000015735626
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,320,7168,8,256,8,balanced,0.06447039991617202
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,320,7168,8,256,8,balanced,0.11023040115833282
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,320,7168,8,256,8,balanced,0.07110079973936081
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,320,7168,8,256,8,balanced,0.06479679979383945
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,320,7168,8,256,8,balanced,0.09528320059180259
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,320,7168,8,256,8,balanced,0.09723839983344078
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,384,7168,8,256,8,balanced,0.07397120073437691
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,384,7168,8,256,8,balanced,0.07656640000641346
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,384,7168,8,256,8,balanced,0.11345599815249444
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,384,7168,8,256,8,balanced,0.07038720063865185
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,384,7168,8,256,8,balanced,0.06400959938764572
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,384,7168,8,256,8,balanced,0.10706560090184211
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,2048,7168,8,256,4,balanced,0.055824000388383865
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,2048,7168,8,256,4,balanced,0.055657599866390225
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,2048,7168,8,256,4,balanced,0.1444992020726204
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,2048,7168,8,256,4,balanced,0.16453759968280793
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,384,7168,8,256,8,balanced,0.07190720029175282
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,384,7168,8,256,8,balanced,0.0671647995710373
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,384,7168,8,256,8,balanced,0.10045759826898575
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,384,7168,8,256,8,balanced,0.0963680014014244
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,512,7168,8,256,8,balanced,0.07484160140156745
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,512,7168,8,256,8,balanced,0.0723904013633728
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,512,7168,8,256,8,balanced,0.11301440075039863
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,512,7168,8,256,8,balanced,0.07058560028672219
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,512,7168,8,256,8,balanced,0.06625279933214187
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,512,7168,8,256,8,balanced,0.10300800055265427
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,512,7168,8,256,8,balanced,0.07646400034427643
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,512,7168,8,256,8,balanced,0.06730560027062893
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,512,7168,8,256,8,balanced,0.1042047992348671
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,512,7168,8,256,8,balanced,0.10581759959459305
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,768,7168,8,256,8,balanced,0.07552000023424625
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,768,7168,8,256,8,balanced,0.08015680052340031
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,12288,7168,8,256,2,balanced,0.7348576009273529
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,12288,7168,8,256,2,balanced,1.0822175979614257
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,3072,7168,8,256,4,balanced,0.05963199995458126
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,3072,7168,8,256,4,balanced,0.0981183998286724
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,3072,7168,8,256,4,balanced,0.20182080268859864
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,3072,7168,8,256,4,balanced,0.06009920053184033
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,3072,7168,8,256,4,balanced,0.06594239994883537
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,3072,7168,8,256,4,balanced,0.194377601146698
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,3072,7168,8,256,4,balanced,0.05975040011107922
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,3072,7168,8,256,4,balanced,0.0673567995429039
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,3072,7168,8,256,4,balanced,0.1962815999984741
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,3072,7168,8,256,4,balanced,0.22394880056381225
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,768,7168,8,256,8,balanced,0.11808639988303185
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,768,7168,8,256,8,balanced,0.07700159884989262
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,768,7168,8,256,8,balanced,0.0644799992442131
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,768,7168,8,256,8,balanced,0.10582399964332581
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,768,7168,8,256,8,balanced,0.0745023988187313
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,768,7168,8,256,8,balanced,0.06695040091872215
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,768,7168,8,256,8,balanced,0.10808000117540359
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,768,7168,8,256,8,balanced,0.10405759960412979
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,4096,7168,8,256,4,balanced,0.06585279926657676
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,4096,7168,8,256,4,balanced,0.12043200060725212
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,4096,7168,8,256,4,balanced,0.2488704040646553
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1024,7168,8,256,8,balanced,0.07511680088937282
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1024,7168,8,256,8,balanced,0.07771200090646743
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1024,7168,8,256,8,balanced,0.1221568003296852
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1024,7168,8,256,8,balanced,0.07953279986977577
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1024,7168,8,256,8,balanced,0.06848960071802139
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1024,7168,8,256,8,balanced,0.11966080144047737
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1024,7168,8,256,8,balanced,0.07793599963188172
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1024,7168,8,256,8,balanced,0.06757119968533516
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1024,7168,8,256,8,balanced,0.11557760015130043
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1024,7168,8,256,8,balanced,0.10952320024371147
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,4096,7168,8,256,4,balanced,0.06761599965393543
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,4096,7168,8,256,4,balanced,0.0767103984951973
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,4096,7168,8,256,4,balanced,0.24883520007133483
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1536,7168,8,256,8,balanced,0.07352960109710693
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1536,7168,8,256,8,balanced,0.08970239907503127
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1536,7168,8,256,8,balanced,0.1405696004629135
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1536,7168,8,256,8,balanced,0.07667199894785881
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1536,7168,8,256,8,balanced,0.06689280010759831
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1536,7168,8,256,8,balanced,0.14072320312261583
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1536,7168,8,256,8,balanced,0.07177599892020226
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1536,7168,8,256,8,balanced,0.06690239943563939
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1536,7168,8,256,8,balanced,0.1390816017985344
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1536,7168,8,256,8,balanced,0.14049600064754486
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,4096,7168,8,256,4,balanced,0.06534400023519993
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,4096,7168,8,256,4,balanced,0.07663039937615394
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,4096,7168,8,256,4,balanced,0.24684160053730012
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,4096,7168,8,256,4,balanced,0.2837952017784119
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,2048,7168,8,256,8,balanced,0.08383680060505867
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,2048,7168,8,256,8,balanced,0.09957439973950385
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,2048,7168,8,256,8,balanced,0.16198400259017945
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,2048,7168,8,256,8,balanced,0.07640640065073967
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,2048,7168,8,256,8,balanced,0.07426559999585151
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,2048,7168,8,256,8,balanced,0.16462080031633378
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,16384,7168,8,256,2,balanced,0.14722879827022553
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,16384,7168,8,256,2,balanced,0.47986880242824553
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,16384,7168,8,256,2,balanced,0.9651392042636872
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,6144,7168,8,256,4,balanced,0.08264319971203804
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,6144,7168,8,256,4,balanced,0.16023999750614165
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,6144,7168,8,256,4,balanced,0.35183999836444857
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,6144,7168,8,256,4,balanced,0.07995839938521385
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,6144,7168,8,256,4,balanced,0.10410239920020103
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,6144,7168,8,256,4,balanced,0.3472640007734299
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,2048,7168,8,256,8,balanced,0.07322239950299263
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,2048,7168,8,256,8,balanced,0.06731519959867001
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,2048,7168,8,256,8,balanced,0.1692960023880005
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,2048,7168,8,256,8,balanced,0.15809920132160188
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,3072,7168,8,256,8,balanced,0.07934719994664192
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,3072,7168,8,256,8,balanced,0.11793279945850373
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,3072,7168,8,256,8,balanced,0.21656640172004699
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,6144,7168,8,256,4,balanced,0.08183680027723313
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,6144,7168,8,256,4,balanced,0.09825599938631058
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,6144,7168,8,256,4,balanced,0.3492095977067947
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,6144,7168,8,256,4,balanced,0.4000512033700943
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,16384,7168,8,256,2,balanced,0.1452320024371147
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,16384,7168,8,256,2,balanced,0.3838399976491928
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,16384,7168,8,256,2,balanced,0.9621887981891633
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,3072,7168,8,256,8,balanced,0.07546240016818047
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,3072,7168,8,256,8,balanced,0.08088959977030755
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,3072,7168,8,256,8,balanced,0.21889920234680177
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,3072,7168,8,256,8,balanced,0.07856639996170997
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,3072,7168,8,256,8,balanced,0.07237120047211647
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,3072,7168,8,256,8,balanced,0.21571840047836305
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,3072,7168,8,256,8,balanced,0.20735679864883422
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,8192,7168,8,256,4,balanced,0.09363519921898841
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,8192,7168,8,256,4,balanced,0.20634240061044692
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,8192,7168,8,256,4,balanced,0.4516608029603958
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,4096,7168,8,256,8,balanced,0.07962239906191826
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,4096,7168,8,256,8,balanced,0.14124800264835358
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,4096,7168,8,256,8,balanced,0.2687296003103256
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,4096,7168,8,256,8,balanced,0.0812608003616333
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,4096,7168,8,256,8,balanced,0.09727039933204651
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,4096,7168,8,256,8,balanced,0.26761600077152253
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,16384,7168,8,256,2,balanced,0.14450879991054535
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,16384,7168,8,256,2,balanced,0.3694911926984787
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,8192,7168,8,256,4,balanced,0.09100160002708435
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,8192,7168,8,256,4,balanced,0.12671679854393006
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,8192,7168,8,256,4,balanced,0.45012799799442293
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,4096,7168,8,256,8,balanced,0.07902080044150353
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,4096,7168,8,256,8,balanced,0.08070079982280731
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,4096,7168,8,256,8,balanced,0.2690623998641968
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,4096,7168,8,256,8,balanced,0.25830080062150956
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,6144,7168,8,256,8,balanced,0.09174719974398612
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,6144,7168,8,256,8,balanced,0.17864640206098556
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,6144,7168,8,256,8,balanced,0.3730879992246628
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,16384,7168,8,256,2,balanced,0.9606944024562836
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,16384,7168,8,256,2,balanced,1.4245759963989257
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,8192,7168,8,256,4,balanced,0.09265599995851517
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,8192,7168,8,256,4,balanced,0.11917439997196197
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,8192,7168,8,256,4,balanced,0.45123199820518495
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,8192,7168,8,256,4,balanced,0.5188768029212951
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,6144,7168,8,256,8,balanced,0.09199360013008118
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,6144,7168,8,256,8,balanced,0.11980799958109856
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,6144,7168,8,256,8,balanced,0.3732192009687424
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,6144,7168,8,256,8,balanced,0.08672320023179055
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,6144,7168,8,256,8,balanced,0.09290880039334297
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,6144,7168,8,256,8,balanced,0.3669920027256012
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,6144,7168,8,256,8,balanced,0.3494623988866806
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,12288,7168,8,256,4,balanced,0.11762239933013915
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,12288,7168,8,256,4,balanced,0.28387199640274047
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,12288,7168,8,256,4,balanced,0.6572735965251922
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,20480,7168,8,256,2,balanced,0.1710335999727249
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,20480,7168,8,256,2,balanced,0.5904576003551483
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,20480,7168,8,256,2,balanced,1.1972352027893067
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,8192,7168,8,256,8,balanced,0.10623680055141449
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,8192,7168,8,256,8,balanced,0.223990398645401
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,8192,7168,8,256,8,balanced,0.47265600264072416
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,12288,7168,8,256,4,balanced,0.11833280101418495
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,12288,7168,8,256,4,balanced,0.17301120162010192
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,12288,7168,8,256,4,balanced,0.6541568040847778
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,8192,7168,8,256,8,balanced,0.10659520030021667
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,8192,7168,8,256,8,balanced,0.14103679805994035
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,8192,7168,8,256,8,balanced,0.47207040190696714
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,12288,7168,8,256,4,balanced,0.11830720007419586
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,12288,7168,8,256,4,balanced,0.16346240043640137
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,12288,7168,8,256,4,balanced,0.652102392911911
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,12288,7168,8,256,4,balanced,0.7516735911369323
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,8192,7168,8,256,8,balanced,0.09981759935617447
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,8192,7168,8,256,8,balanced,0.10152000039815903
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,8192,7168,8,256,8,balanced,0.47073599696159363
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,8192,7168,8,256,8,balanced,0.4422592043876648
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,20480,7168,8,256,2,balanced,0.17236160039901732
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,20480,7168,8,256,2,balanced,0.4731808006763458
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,20480,7168,8,256,2,balanced,1.1953312039375306
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,12288,7168,8,256,8,balanced,0.12773120105266572
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,12288,7168,8,256,8,balanced,0.30648320019245145
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,12288,7168,8,256,8,balanced,0.6834592044353485
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,16384,7168,8,256,4,balanced,0.1433568000793457
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,16384,7168,8,256,4,balanced,0.37005760371685026
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,16384,7168,8,256,4,balanced,0.862553596496582
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,20480,7168,8,256,2,balanced,0.17227840125560762
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,20480,7168,8,256,2,balanced,0.4501824021339417
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,20480,7168,8,256,2,balanced,1.195193588733673
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,20480,7168,8,256,2,balanced,1.7730464100837708
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,12288,7168,8,256,8,balanced,0.127199999243021
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,12288,7168,8,256,8,balanced,0.19139840006828307
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,12288,7168,8,256,8,balanced,0.6771488010883331
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,16384,7168,8,256,4,balanced,0.1425152003765106
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,16384,7168,8,256,4,balanced,0.22003840208053588
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,16384,7168,8,256,4,balanced,0.8577567994594574
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,12288,7168,8,256,8,balanced,0.1264447994530201
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,12288,7168,8,256,8,balanced,0.12318079993128776
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,12288,7168,8,256,8,balanced,0.6778560042381286
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,12288,7168,8,256,8,balanced,0.6315551936626435
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,16384,7168,8,256,4,balanced,0.14307839870452882
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,16384,7168,8,256,4,balanced,0.20679040104150773
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,16384,7168,8,256,4,balanced,0.8547327935695648
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,16384,7168,8,256,4,balanced,0.9857216000556945
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,32768,7168,8,256,2,balanced,0.25086399763822553
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,32768,7168,8,256,2,balanced,0.9175328075885772
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,32768,7168,8,256,2,balanced,1.9369728088378906
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,16384,7168,8,256,8,balanced,0.15095039904117585
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,16384,7168,8,256,8,balanced,0.3886784017086029
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,16384,7168,8,256,8,balanced,0.8744960010051728
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,20480,7168,8,256,4,balanced,0.16734079718589784
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,20480,7168,8,256,4,balanced,0.4517024040222168
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,20480,7168,8,256,4,balanced,1.0723935961723328
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,16384,7168,8,256,8,balanced,0.15190400034189225
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,16384,7168,8,256,8,balanced,0.23110720068216323
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,16384,7168,8,256,8,balanced,0.8787487983703614
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,32768,7168,8,256,2,balanced,0.24836159944534303
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,32768,7168,8,256,2,balanced,0.730377596616745
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,32768,7168,8,256,2,balanced,1.8781056165695191
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,16384,7168,8,256,8,balanced,0.15051199942827226
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,16384,7168,8,256,8,balanced,0.1450272023677826
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,16384,7168,8,256,8,balanced,0.8746464073657989
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,16384,7168,8,256,8,balanced,0.8136383950710296
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,20480,7168,8,256,4,balanced,0.16885760128498079
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,20480,7168,8,256,4,balanced,0.26483839750289917
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,20480,7168,8,256,4,balanced,1.0706016063690185
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,20480,7168,8,256,8,balanced,0.17783999890089036
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,20480,7168,8,256,8,balanced,0.46936639547348025
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,20480,7168,8,256,8,balanced,1.084716796875
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,20480,7168,8,256,4,balanced,0.16857280135154723
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,20480,7168,8,256,4,balanced,0.2475680023431778
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,20480,7168,8,256,4,balanced,1.0701663970947266
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,20480,7168,8,256,4,balanced,1.229359996318817
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,32768,7168,8,256,2,balanced,0.24996800124645233
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,32768,7168,8,256,2,balanced,0.6975360035896301
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,32768,7168,8,256,2,balanced,1.8738720059394836
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,32768,7168,8,256,2,balanced,2.79870400428772
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,20480,7168,8,256,8,balanced,0.17407359778881074
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,20480,7168,8,256,8,balanced,0.280323201417923
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,20480,7168,8,256,8,balanced,1.0870527982711793
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,32768,7168,8,256,4,balanced,0.24301759898662567
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,32768,7168,8,256,4,balanced,0.6994207918643951
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,32768,7168,8,256,4,balanced,1.67810879945755
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,20480,7168,8,256,8,balanced,0.1721951976418495
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,20480,7168,8,256,8,balanced,0.16991359889507293
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,20480,7168,8,256,8,balanced,1.0830623865127564
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,20480,7168,8,256,8,balanced,1.0098623752593994
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,65536,7168,8,256,2,balanced,0.458841598033905
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,65536,7168,8,256,2,balanced,1.7910175919532776
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,65536,7168,8,256,2,balanced,3.838969588279724
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,32768,7168,8,256,4,balanced,0.243046398460865
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,32768,7168,8,256,4,balanced,0.4030112028121948
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,32768,7168,8,256,4,balanced,1.674726402759552
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,32768,7168,8,256,8,balanced,0.24766079783439637
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,32768,7168,8,256,8,balanced,0.7253152012825013
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,32768,7168,8,256,8,balanced,1.6936256051063538
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,32768,7168,8,256,4,balanced,0.2437824010848999
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,32768,7168,8,256,4,balanced,0.3754720002412796
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,32768,7168,8,256,4,balanced,1.6788576006889344
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,32768,7168,8,256,4,balanced,1.928169596195221
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,32768,7168,8,256,8,balanced,0.24677119851112367
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,32768,7168,8,256,8,balanced,0.4105791985988617
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,32768,7168,8,256,8,balanced,1.7008511781692506
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,65536,7168,8,256,2,balanced,0.4575520038604736
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,65536,7168,8,256,2,balanced,1.422435212135315
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,65536,7168,8,256,2,balanced,3.6990336179733276
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,32768,7168,8,256,8,balanced,0.2435552015900612
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,32768,7168,8,256,8,balanced,0.24556799978017807
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,32768,7168,8,256,8,balanced,1.693334412574768
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,32768,7168,8,256,8,balanced,1.5653791904449463
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,65536,7168,8,256,4,balanced,0.44351360499858855
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,65536,7168,8,256,4,balanced,1.3630527973175048
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,65536,7168,8,256,4,balanced,3.3121920108795164
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,65536,7168,8,256,2,balanced,0.4634848028421402
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,65536,7168,8,256,2,balanced,1.3548352003097535
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,65536,7168,8,256,2,balanced,3.8068480253219605
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,65536,7168,8,256,2,balanced,5.543644809722901
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,65536,7168,8,256,8,balanced,0.44378560185432436
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,65536,7168,8,256,8,balanced,1.401644802093506
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,65536,7168,8,256,8,balanced,3.3360607862472533
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,65536,7168,8,256,4,balanced,0.4420415997505188
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,65536,7168,8,256,4,balanced,0.7677184045314789
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,65536,7168,8,256,4,balanced,3.312768006324768
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,65536,7168,8,256,8,balanced,0.4395103991031647
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,65536,7168,8,256,8,balanced,0.7803936004638672
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,65536,7168,8,256,8,balanced,3.3358719825744627
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,65536,7168,8,256,4,balanced,0.4410944014787674
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,65536,7168,8,256,4,balanced,0.707587206363678
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,65536,7168,8,256,4,balanced,3.313817596435547
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,65536,7168,8,256,4,balanced,3.810803198814392
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,65536,7168,8,256,8,balanced,0.44156799614429476
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,65536,7168,8,256,8,balanced,0.4843328058719635
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,65536,7168,8,256,8,balanced,3.3342527866363527
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,65536,7168,8,256,8,balanced,3.036406421661377
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1,7168,8,256,16,balanced,0.08259200006723404
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1,7168,8,256,16,balanced,0.06624320037662983
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1,7168,8,256,16,balanced,0.07015680000185967
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1,7168,8,256,16,balanced,0.07295040003955364
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1,7168,8,256,16,balanced,0.0686400007456541
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1,7168,8,256,16,balanced,0.07089599967002869
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1,7168,8,256,16,balanced,0.07503680102527141
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1,7168,8,256,16,balanced,0.11477120146155358
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1,7168,8,256,16,balanced,0.09828160032629966
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1,7168,8,256,16,balanced,0.13471679985523224
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,2,7168,8,256,16,balanced,0.10936960056424141
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,2,7168,8,256,16,balanced,0.0966399997472763
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,2,7168,8,256,16,balanced,0.08713600113987922
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,2,7168,8,256,16,balanced,0.09538880065083503
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,2,7168,8,256,16,balanced,0.09447679892182351
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,2,7168,8,256,16,balanced,0.08050239980220794
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,2,7168,8,256,16,balanced,0.09811519980430602
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,2,7168,8,256,16,balanced,0.08664639964699745
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,2,7168,8,256,16,balanced,0.08309120088815689
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,2,7168,8,256,16,balanced,0.11657599955797196
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,4,7168,8,256,16,balanced,0.09167999923229217
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,4,7168,8,256,16,balanced,0.08516799956560135
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,4,7168,8,256,16,balanced,0.10545600056648255
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,4,7168,8,256,16,balanced,0.08951679989695549
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,4,7168,8,256,16,balanced,0.08206719905138016
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,4,7168,8,256,16,balanced,0.07407039999961854
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,4,7168,8,256,16,balanced,0.09401599988341332
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,4,7168,8,256,16,balanced,0.07884799912571908
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,4,7168,8,256,16,balanced,0.07194560058414937
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,4,7168,8,256,16,balanced,0.10444479882717132
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,8,7168,8,256,16,balanced,0.08472959995269776
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,8,7168,8,256,16,balanced,0.07415360063314438
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,8,7168,8,256,16,balanced,0.0790752001106739
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,8,7168,8,256,16,balanced,0.08997759968042374
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,8,7168,8,256,16,balanced,0.0756351999938488
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,8,7168,8,256,16,balanced,0.08584000021219254
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,8,7168,8,256,16,balanced,0.08882240056991578
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,8,7168,8,256,16,balanced,0.07421439960598945
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,8,7168,8,256,16,balanced,0.07984640002250672
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,8,7168,8,256,16,balanced,0.09763839989900588
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,16,7168,8,256,16,balanced,0.0759359996765852
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,16,7168,8,256,16,balanced,0.07050879932940006
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,16,7168,8,256,16,balanced,0.08852160051465034
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,16,7168,8,256,16,balanced,0.0798879999667406
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,16,7168,8,256,16,balanced,0.07367039918899536
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,16,7168,8,256,16,balanced,0.09471679925918579
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,16,7168,8,256,16,balanced,0.08075520023703575
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,16,7168,8,256,16,balanced,0.06903040073812008
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,16,7168,8,256,16,balanced,0.0901696003973484
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,16,7168,8,256,16,balanced,0.09365760087966919
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,32,7168,8,256,16,balanced,0.07672000005841255
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,32,7168,8,256,16,balanced,0.06675200052559375
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,32,7168,8,256,16,balanced,0.10668799877166749
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,32,7168,8,256,16,balanced,0.07482880055904388
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,32,7168,8,256,16,balanced,0.06707840040326118
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,32,7168,8,256,16,balanced,0.10592000111937523
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,32,7168,8,256,16,balanced,0.07400639913976192
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,32,7168,8,256,16,balanced,0.06884479969739914
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,32,7168,8,256,16,balanced,0.1069056011736393
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,32,7168,8,256,16,balanced,0.10387199968099595
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,48,7168,8,256,16,balanced,0.07771520130336285
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,48,7168,8,256,16,balanced,0.07312319912016392
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,48,7168,8,256,16,balanced,0.11545279920101166
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,48,7168,8,256,16,balanced,0.07602560073137284
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,48,7168,8,256,16,balanced,0.06774399951100349
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,48,7168,8,256,16,balanced,0.12378560155630111
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,48,7168,8,256,16,balanced,0.07176000103354455
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,48,7168,8,256,16,balanced,0.06593279987573623
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,48,7168,8,256,16,balanced,0.10919039994478226
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,48,7168,8,256,16,balanced,0.09666879922151565
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,64,7168,8,256,16,balanced,0.06797120049595833
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,64,7168,8,256,16,balanced,0.0749728012830019
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,64,7168,8,256,16,balanced,0.11411520168185234
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,64,7168,8,256,16,balanced,0.06693120002746582
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,64,7168,8,256,16,balanced,0.06707520000636577
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,64,7168,8,256,16,balanced,0.11404799893498421
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,64,7168,8,256,16,balanced,0.0743648000061512
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,64,7168,8,256,16,balanced,0.06177599988877773
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,64,7168,8,256,16,balanced,0.1125567987561226
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,64,7168,8,256,16,balanced,0.11196159943938255
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,80,7168,8,256,16,balanced,0.07160640060901642
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,80,7168,8,256,16,balanced,0.07098879888653756
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,80,7168,8,256,16,balanced,0.1070080004632473
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,80,7168,8,256,16,balanced,0.07162239961326122
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,80,7168,8,256,16,balanced,0.06381119973957539
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,80,7168,8,256,16,balanced,0.10882239937782287
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,80,7168,8,256,16,balanced,0.07357759959995747
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,80,7168,8,256,16,balanced,0.063408000767231
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,80,7168,8,256,16,balanced,0.10765439867973328
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,80,7168,8,256,16,balanced,0.10089280009269715
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,96,7168,8,256,16,balanced,0.07301440052688121
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,96,7168,8,256,16,balanced,0.06959360018372536
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,96,7168,8,256,16,balanced,0.10831360071897507
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,96,7168,8,256,16,balanced,0.06931839883327484
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,96,7168,8,256,16,balanced,0.06104640103876591
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,96,7168,8,256,16,balanced,0.10847359895706177
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,96,7168,8,256,16,balanced,0.06591040045022964
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,96,7168,8,256,16,balanced,0.06869760043919086
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,96,7168,8,256,16,balanced,0.11238720193505287
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,96,7168,8,256,16,balanced,0.1075775995850563
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,128,7168,8,256,16,balanced,0.07124159969389439
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,128,7168,8,256,16,balanced,0.06761280037462711
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,128,7168,8,256,16,balanced,0.10633600056171418
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,128,7168,8,256,16,balanced,0.07057919912040234
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,128,7168,8,256,16,balanced,0.06312960013747215
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,128,7168,8,256,16,balanced,0.10738560035824776
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,128,7168,8,256,16,balanced,0.06939839906990528
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,128,7168,8,256,16,balanced,0.06257599964737892
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,128,7168,8,256,16,balanced,0.10599039867520332
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,128,7168,8,256,16,balanced,0.11039679944515228
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,160,7168,8,256,16,balanced,0.07622400037944317
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,160,7168,8,256,16,balanced,0.07039999924600124
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,160,7168,8,256,16,balanced,0.10477759912610055
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,160,7168,8,256,16,balanced,0.07360000014305115
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,160,7168,8,256,16,balanced,0.06330880001187325
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,160,7168,8,256,16,balanced,0.10278719961643219
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,160,7168,8,256,16,balanced,0.07307519987225533
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,160,7168,8,256,16,balanced,0.06144320033490658
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,160,7168,8,256,16,balanced,0.10764480009675026
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,160,7168,8,256,16,balanced,0.10037439912557602
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,192,7168,8,256,16,balanced,0.07688640058040619
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,192,7168,8,256,16,balanced,0.06847679913043976
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,192,7168,8,256,16,balanced,0.10147840082645417
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,192,7168,8,256,16,balanced,0.06969280056655407
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,192,7168,8,256,16,balanced,0.06348480060696601
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,192,7168,8,256,16,balanced,0.10823679864406585
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,192,7168,8,256,16,balanced,0.07197760045528412
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,192,7168,8,256,16,balanced,0.06154559999704361
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,192,7168,8,256,16,balanced,0.10140480026602745
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,192,7168,8,256,16,balanced,0.10315839946269989
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,256,7168,8,256,16,balanced,0.06857919991016388
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,256,7168,8,256,16,balanced,0.06946559920907021
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,256,7168,8,256,16,balanced,0.10864320024847984
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,256,7168,8,256,16,balanced,0.07167039923369885
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,256,7168,8,256,16,balanced,0.06066560037434101
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,256,7168,8,256,16,balanced,0.10693440064787865
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,256,7168,8,256,16,balanced,0.06689600050449371
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,256,7168,8,256,16,balanced,0.06604800000786781
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,256,7168,8,256,16,balanced,0.10098879933357238
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,256,7168,8,256,16,balanced,0.0971264012157917
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,320,7168,8,256,16,balanced,0.06452479921281337
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,320,7168,8,256,16,balanced,0.07259839959442616
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,320,7168,8,256,16,balanced,0.10533439889550208
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,320,7168,8,256,16,balanced,0.06972799971699714
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,320,7168,8,256,16,balanced,0.061379199847579
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,320,7168,8,256,16,balanced,0.11162239983677864
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,320,7168,8,256,16,balanced,0.0700864002108574
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,320,7168,8,256,16,balanced,0.06289280056953431
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,320,7168,8,256,16,balanced,0.10304960012435913
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,320,7168,8,256,16,balanced,0.10274560004472733
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,384,7168,8,256,16,balanced,0.06971199922263623
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,384,7168,8,256,16,balanced,0.07052479982376099
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,384,7168,8,256,16,balanced,0.10006720051169396
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,384,7168,8,256,16,balanced,0.07044800035655499
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,384,7168,8,256,16,balanced,0.06107200048863888
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,384,7168,8,256,16,balanced,0.10139519944787026
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,384,7168,8,256,16,balanced,0.0686912003904581
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,384,7168,8,256,16,balanced,0.05926720015704632
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,384,7168,8,256,16,balanced,0.10247040018439293
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,384,7168,8,256,16,balanced,0.09701759964227677
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,512,7168,8,256,16,balanced,0.06575679890811444
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,512,7168,8,256,16,balanced,0.07567040026187896
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,512,7168,8,256,16,balanced,0.10503999963402748
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,512,7168,8,256,16,balanced,0.0702143993228674
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,512,7168,8,256,16,balanced,0.06164159998297691
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,512,7168,8,256,16,balanced,0.09551040083169937
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,512,7168,8,256,16,balanced,0.06807360053062439
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,512,7168,8,256,16,balanced,0.06373120024800301
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,512,7168,8,256,16,balanced,0.09943360090255737
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,512,7168,8,256,16,balanced,0.09913600012660026
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,768,7168,8,256,16,balanced,0.06714559979736805
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,768,7168,8,256,16,balanced,0.07399679906666279
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,768,7168,8,256,16,balanced,0.1097216010093689
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,768,7168,8,256,16,balanced,0.0732448011636734
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,768,7168,8,256,16,balanced,0.061660800129175186
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,768,7168,8,256,16,balanced,0.10810879990458488
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,768,7168,8,256,16,balanced,0.06757120117545128
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,768,7168,8,256,16,balanced,0.06607680059969426
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,768,7168,8,256,16,balanced,0.10914560109376907
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,768,7168,8,256,16,balanced,0.11073920205235481
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1024,7168,8,256,16,balanced,0.06969919987022877
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1024,7168,8,256,16,balanced,0.07821759879589081
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1024,7168,8,256,16,balanced,0.11915199905633926
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1024,7168,8,256,16,balanced,0.0696704000234604
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1024,7168,8,256,16,balanced,0.06340479999780654
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1024,7168,8,256,16,balanced,0.11767359972000122
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1024,7168,8,256,16,balanced,0.0775392010807991
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1024,7168,8,256,16,balanced,0.060988799482584
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1024,7168,8,256,16,balanced,0.11750399917364121
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1024,7168,8,256,16,balanced,0.11605760157108307
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1536,7168,8,256,16,balanced,0.06945280134677886
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1536,7168,8,256,16,balanced,0.09171520024538041
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1536,7168,8,256,16,balanced,0.14311359971761703
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1536,7168,8,256,16,balanced,0.07453440017998218
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1536,7168,8,256,16,balanced,0.07151359990239144
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1536,7168,8,256,16,balanced,0.14666559994220735
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1536,7168,8,256,16,balanced,0.07199679985642433
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1536,7168,8,256,16,balanced,0.06573120020329952
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1536,7168,8,256,16,balanced,0.14438720047473907
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1536,7168,8,256,16,balanced,0.13838400095701217
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,2048,7168,8,256,16,balanced,0.06881920136511326
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,2048,7168,8,256,16,balanced,0.10215039998292923
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,2048,7168,8,256,16,balanced,0.16801920235157014
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,2048,7168,8,256,16,balanced,0.0756800003349781
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,2048,7168,8,256,16,balanced,0.0746751993894577
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,2048,7168,8,256,16,balanced,0.17131199985742568
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,2048,7168,8,256,16,balanced,0.07554560154676437
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,2048,7168,8,256,16,balanced,0.06591999903321266
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,2048,7168,8,256,16,balanced,0.17238400131464005
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,2048,7168,8,256,16,balanced,0.16134719997644426
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,3072,7168,8,256,16,balanced,0.0798111990094185
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,3072,7168,8,256,16,balanced,0.1264255978167057
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,3072,7168,8,256,16,balanced,0.2200608029961586
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,3072,7168,8,256,16,balanced,0.07583360075950622
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,3072,7168,8,256,16,balanced,0.0848064012825489
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,3072,7168,8,256,16,balanced,0.22387840002775192
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,3072,7168,8,256,16,balanced,0.08016640022397041
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,3072,7168,8,256,16,balanced,0.07412799894809723
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,3072,7168,8,256,16,balanced,0.22133439928293228
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,3072,7168,8,256,16,balanced,0.21059839874505998
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,4096,7168,8,256,16,balanced,0.08238720074295998
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,4096,7168,8,256,16,balanced,0.14926399886608124
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,4096,7168,8,256,16,balanced,0.27345280051231385
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,4096,7168,8,256,16,balanced,0.08002560064196587
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,4096,7168,8,256,16,balanced,0.09787520095705986
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,4096,7168,8,256,16,balanced,0.2753760010004044
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,4096,7168,8,256,16,balanced,0.08336639925837516
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,4096,7168,8,256,16,balanced,0.08470079973340035
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,4096,7168,8,256,16,balanced,0.27982400059700013
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,4096,7168,8,256,16,balanced,0.25925120115280154
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,6144,7168,8,256,16,balanced,0.09352640062570572
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,6144,7168,8,256,16,balanced,0.19187839776277543
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,6144,7168,8,256,16,balanced,0.37965439856052396
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,6144,7168,8,256,16,balanced,0.09935040175914764
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,6144,7168,8,256,16,balanced,0.12537919953465462
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,6144,7168,8,256,16,balanced,0.3791391998529434
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,6144,7168,8,256,16,balanced,0.09594559967517853
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,6144,7168,8,256,16,balanced,0.09875520020723343
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,6144,7168,8,256,16,balanced,0.3813536047935486
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,6144,7168,8,256,16,balanced,0.3532383948564529
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,8192,7168,8,256,16,balanced,0.10927040129899979
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,8192,7168,8,256,16,balanced,0.23836799710988998
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,8192,7168,8,256,16,balanced,0.48711040019989016
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,8192,7168,8,256,16,balanced,0.10384959802031517
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,8192,7168,8,256,16,balanced,0.15206400007009507
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,8192,7168,8,256,16,balanced,0.484208008646965
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,8192,7168,8,256,16,balanced,0.10606400072574615
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,8192,7168,8,256,16,balanced,0.10896000117063523
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,8192,7168,8,256,16,balanced,0.48899840414524076
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,8192,7168,8,256,16,balanced,0.44025920033454896
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,12288,7168,8,256,16,balanced,0.13089279979467391
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,12288,7168,8,256,16,balanced,0.3290176033973694
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,12288,7168,8,256,16,balanced,0.6952096045017242
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,12288,7168,8,256,16,balanced,0.13370879963040352
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,12288,7168,8,256,16,balanced,0.19719360172748565
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,12288,7168,8,256,16,balanced,0.6973183989524842
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,12288,7168,8,256,16,balanced,0.13344959840178489
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,12288,7168,8,256,16,balanced,0.1401728019118309
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,12288,7168,8,256,16,balanced,0.6968735992908478
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,12288,7168,8,256,16,balanced,0.6312640011310577
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,16384,7168,8,256,16,balanced,0.15987520068883895
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,16384,7168,8,256,16,balanced,0.4208032011985779
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,16384,7168,8,256,16,balanced,0.9077983975410462
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,16384,7168,8,256,16,balanced,0.1532383993268013
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,16384,7168,8,256,16,balanced,0.24648959785699845
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,16384,7168,8,256,16,balanced,0.9052671968936921
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,16384,7168,8,256,16,balanced,0.15518079847097396
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,16384,7168,8,256,16,balanced,0.17039359956979752
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,16384,7168,8,256,16,balanced,0.9050719976425171
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,16384,7168,8,256,16,balanced,0.8153696060180664
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,20480,7168,8,256,16,balanced,0.1807679995894432
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,20480,7168,8,256,16,balanced,0.5140255987644196
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,20480,7168,8,256,16,balanced,1.1240896105766296
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,20480,7168,8,256,16,balanced,0.18176319897174836
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,20480,7168,8,256,16,balanced,0.28879680037498473
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,20480,7168,8,256,16,balanced,1.1275231838226318
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,20480,7168,8,256,16,balanced,0.18139199912548065
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,20480,7168,8,256,16,balanced,0.2000031977891922
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,20480,7168,8,256,16,balanced,1.1270464181900024
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,20480,7168,8,256,16,balanced,1.0144351780414582
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,32768,7168,8,256,16,balanced,0.2580479994416237
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,32768,7168,8,256,16,balanced,0.7998687982559204
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,32768,7168,8,256,16,balanced,1.7649215936660767
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,32768,7168,8,256,16,balanced,0.25395520180463793
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,32768,7168,8,256,16,balanced,0.43675519824028014
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,32768,7168,8,256,16,balanced,1.7641216158866881
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,32768,7168,8,256,16,balanced,0.24942079931497574
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,32768,7168,8,256,16,balanced,0.28840000331401827
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,32768,7168,8,256,16,balanced,1.7653824090957642
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,32768,7168,8,256,16,balanced,1.562995195388794
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,65536,7168,8,256,16,balanced,0.4438272029161453
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,65536,7168,8,256,16,balanced,1.563308799266815
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,65536,7168,8,256,16,balanced,3.485097599029541
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,65536,7168,8,256,16,balanced,0.4443327933549881
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,65536,7168,8,256,16,balanced,0.8284127950668335
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,65536,7168,8,256,16,balanced,3.4829311847686766
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1,7168,8,256,32,balanced,0.06586240008473396
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1,7168,8,256,32,balanced,0.06359679959714412
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1,7168,8,256,32,balanced,0.0659487996250391
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1,7168,8,256,32,balanced,0.06880000047385693
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1,7168,8,256,32,balanced,0.05702719911932945
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1,7168,8,256,32,balanced,0.0682399995625019
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1,7168,8,256,32,balanced,0.06825920008122921
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1,7168,8,256,32,balanced,0.06357440054416656
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1,7168,8,256,32,balanced,0.07169600091874599
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1,7168,8,256,32,balanced,0.07474239990115165
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,2,7168,8,256,32,balanced,0.07719680033624172
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,2,7168,8,256,32,balanced,0.057225600630044934
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,2,7168,8,256,32,balanced,0.07142719998955727
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,2,7168,8,256,32,balanced,0.06417600028216838
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,2,7168,8,256,32,balanced,0.0493792001157999
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,2,7168,8,256,32,balanced,0.0694464012980461
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,2,7168,8,256,32,balanced,0.0650239996612072
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,2,7168,8,256,32,balanced,0.051401599496603015
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,2,7168,8,256,32,balanced,0.06918719932436942
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,2,7168,8,256,32,balanced,0.06830080039799213
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,4,7168,8,256,32,balanced,0.06283199936151504
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,4,7168,8,256,32,balanced,0.05349440015852451
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,4,7168,8,256,32,balanced,0.07105600014328957
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,4,7168,8,256,32,balanced,0.06091519929468632
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,4,7168,8,256,32,balanced,0.05150719992816448
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,4,7168,8,256,32,balanced,0.06670079976320267
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,4,7168,8,256,32,balanced,0.0570847999304533
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,4,7168,8,256,32,balanced,0.05151039958000183
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,4,7168,8,256,32,balanced,0.06680959984660148
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,4,7168,8,256,32,balanced,0.06492159925401211
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,8,7168,8,256,32,balanced,0.06082879900932312
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,8,7168,8,256,32,balanced,0.04816320016980171
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,8,7168,8,256,32,balanced,0.07846079915761947
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,8,7168,8,256,32,balanced,0.05913279987871647
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,8,7168,8,256,32,balanced,0.05038079991936684
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,8,7168,8,256,32,balanced,0.07683840021491051
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,8,7168,8,256,32,balanced,0.059065600484609605
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,8,7168,8,256,32,balanced,0.05023360028862953
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,8,7168,8,256,32,balanced,0.07960640043020248
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,8,7168,8,256,32,balanced,0.07837760075926781
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,16,7168,8,256,32,balanced,0.06579839996993542
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,16,7168,8,256,32,balanced,0.05106240026652813
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,16,7168,8,256,32,balanced,0.07200959958136081
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,16,7168,8,256,32,balanced,0.06266559921205044
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,16,7168,8,256,32,balanced,0.04791360050439834
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,16,7168,8,256,32,balanced,0.07398719936609269
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,16,7168,8,256,32,balanced,0.0628127995878458
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,16,7168,8,256,32,balanced,0.0517280001193285
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,16,7168,8,256,32,balanced,0.0787712000310421
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,16,7168,8,256,32,balanced,0.07633600011467934
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,32,7168,8,256,32,balanced,0.06069760136306286
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,32,7168,8,256,32,balanced,0.053241600096225736
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,32,7168,8,256,32,balanced,0.07560319975018501
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,32,7168,8,256,32,balanced,0.06618879921734333
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,32,7168,8,256,32,balanced,0.050207999721169475
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,32,7168,8,256,32,balanced,0.07387199923396111
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,32,7168,8,256,32,balanced,0.059411199763417244
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,32,7168,8,256,32,balanced,0.05230080001056194
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,32,7168,8,256,32,balanced,0.07786240056157112
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,32,7168,8,256,32,balanced,0.07600319981575013
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,48,7168,8,256,32,balanced,0.058134399354457855
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,48,7168,8,256,32,balanced,0.06989119909703731
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,48,7168,8,256,32,balanced,0.09468799978494644
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,48,7168,8,256,32,balanced,0.07195520028471947
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,48,7168,8,256,32,balanced,0.06272320002317429
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,48,7168,8,256,32,balanced,0.08672319948673249
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,48,7168,8,256,32,balanced,0.0682399995625019
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,48,7168,8,256,32,balanced,0.060256000608205795
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,48,7168,8,256,32,balanced,0.0781151995062828
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,48,7168,8,256,32,balanced,0.0805759996175766
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,64,7168,8,256,32,balanced,0.06716160029172898
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,64,7168,8,256,32,balanced,0.05990400016307831
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,64,7168,8,256,32,balanced,0.07404799908399581
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,64,7168,8,256,32,balanced,0.06028800085186958
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,64,7168,8,256,32,balanced,0.05241919979453087
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,64,7168,8,256,32,balanced,0.07802879959344863
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,64,7168,8,256,32,balanced,0.06451200135052204
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,64,7168,8,256,32,balanced,0.05535999946296215
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,64,7168,8,256,32,balanced,0.07839039936661721
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,64,7168,8,256,32,balanced,0.0745888002216816
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,80,7168,8,256,32,balanced,0.061424000561237334
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,80,7168,8,256,32,balanced,0.05772159956395626
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,80,7168,8,256,32,balanced,0.08367360010743141
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,80,7168,8,256,32,balanced,0.06215040013194084
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,80,7168,8,256,32,balanced,0.05299520008265972
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,80,7168,8,256,32,balanced,0.08095039948821067
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,80,7168,8,256,32,balanced,0.06266879998147487
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,80,7168,8,256,32,balanced,0.054393600672483444
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,80,7168,8,256,32,balanced,0.0939071998000145
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,80,7168,8,256,32,balanced,0.07554559931159019
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,96,7168,8,256,32,balanced,0.06303680054843426
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,96,7168,8,256,32,balanced,0.05432640053331852
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,96,7168,8,256,32,balanced,0.08112959861755371
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,96,7168,8,256,32,balanced,0.05902079902589321
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,96,7168,8,256,32,balanced,0.04878400005400181
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,96,7168,8,256,32,balanced,0.07953279986977577
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,96,7168,8,256,32,balanced,0.060841599479317665
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,96,7168,8,256,32,balanced,0.05141440033912659
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,96,7168,8,256,32,balanced,0.07958400025963783
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,96,7168,8,256,32,balanced,0.07073599994182586
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,128,7168,8,256,32,balanced,0.06174080036580563
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,128,7168,8,256,32,balanced,0.05532479956746102
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,128,7168,8,256,32,balanced,0.08074239939451218
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,128,7168,8,256,32,balanced,0.058326400071382525
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,128,7168,8,256,32,balanced,0.04649599976837635
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,128,7168,8,256,32,balanced,0.07193279974162578
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,128,7168,8,256,32,balanced,0.059030399471521375
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,128,7168,8,256,32,balanced,0.04877759926021099
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,128,7168,8,256,32,balanced,0.08095040023326874
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,128,7168,8,256,32,balanced,0.07999679967761039
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,160,7168,8,256,32,balanced,0.060208000242710114
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,160,7168,8,256,32,balanced,0.05682239979505539
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,160,7168,8,256,32,balanced,0.07354559898376464
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,160,7168,8,256,32,balanced,0.05562559999525547
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,160,7168,8,256,32,balanced,0.049049599841237065
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,160,7168,8,256,32,balanced,0.07510399967432022
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,160,7168,8,256,32,balanced,0.06070400029420853
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,160,7168,8,256,32,balanced,0.05087039992213249
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,160,7168,8,256,32,balanced,0.07699519917368888
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,160,7168,8,256,32,balanced,0.07464319914579391
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,192,7168,8,256,32,balanced,0.061283199861645696
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,192,7168,8,256,32,balanced,0.05868799947202206
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,192,7168,8,256,32,balanced,0.07520639896392822
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,192,7168,8,256,32,balanced,0.05776959918439388
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,192,7168,8,256,32,balanced,0.0513215996325016
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,192,7168,8,256,32,balanced,0.0788256011903286
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,192,7168,8,256,32,balanced,0.057996800914406774
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,192,7168,8,256,32,balanced,0.0543007992208004
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,192,7168,8,256,32,balanced,0.07656960040330887
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,192,7168,8,256,32,balanced,0.0765952005982399
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,256,7168,8,256,32,balanced,0.05783680006861687
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,256,7168,8,256,32,balanced,0.05774080008268356
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,256,7168,8,256,32,balanced,0.07763200029730796
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,256,7168,8,256,32,balanced,0.05838080123066902
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,256,7168,8,256,32,balanced,0.048864000290632245
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,256,7168,8,256,32,balanced,0.07837439998984337
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,256,7168,8,256,32,balanced,0.05924479961395264
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,256,7168,8,256,32,balanced,0.05165759958326817
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,256,7168,8,256,32,balanced,0.07352639958262444
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,256,7168,8,256,32,balanced,0.06969279870390892
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,320,7168,8,256,32,balanced,0.0611519992351532
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,320,7168,8,256,32,balanced,0.05681279934942722
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,320,7168,8,256,32,balanced,0.07940160073339939
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,320,7168,8,256,32,balanced,0.05921599939465523
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,320,7168,8,256,32,balanced,0.051286400854587556
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,320,7168,8,256,32,balanced,0.07640640065073967
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,320,7168,8,256,32,balanced,0.05895039960741997
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,320,7168,8,256,32,balanced,0.053007999807596205
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,320,7168,8,256,32,balanced,0.07677119970321655
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,320,7168,8,256,32,balanced,0.07732800021767616
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,384,7168,8,256,32,balanced,0.060329600051045415
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,384,7168,8,256,32,balanced,0.05602560006082058
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,384,7168,8,256,32,balanced,0.07863040044903755
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,384,7168,8,256,32,balanced,0.06050239950418472
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,384,7168,8,256,32,balanced,0.049027199670672414
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,384,7168,8,256,32,balanced,0.0787648007273674
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,384,7168,8,256,32,balanced,0.05681600011885166
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,384,7168,8,256,32,balanced,0.04997440092265606
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,384,7168,8,256,32,balanced,0.07416960150003434
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,384,7168,8,256,32,balanced,0.07998720034956933
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,512,7168,8,256,32,balanced,0.06052160039544106
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,512,7168,8,256,32,balanced,0.05857600010931492
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,512,7168,8,256,32,balanced,0.08232000097632408
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,512,7168,8,256,32,balanced,0.05718400031328201
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,512,7168,8,256,32,balanced,0.05009919963777065
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,512,7168,8,256,32,balanced,0.07911679968237877
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,512,7168,8,256,32,balanced,0.05786880031228066
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,512,7168,8,256,32,balanced,0.05328320041298866
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,512,7168,8,256,32,balanced,0.0778559997677803
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,512,7168,8,256,32,balanced,0.07874240055680275
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,768,7168,8,256,32,balanced,0.05838719941675663
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,768,7168,8,256,32,balanced,0.06390400007367134
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,768,7168,8,256,32,balanced,0.0976415991783142
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,768,7168,8,256,32,balanced,0.060416000336408614
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,768,7168,8,256,32,balanced,0.05498879961669445
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,768,7168,8,256,32,balanced,0.09081280007958412
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,768,7168,8,256,32,balanced,0.06601599976420403
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,768,7168,8,256,32,balanced,0.05674239993095398
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,768,7168,8,256,32,balanced,0.08998400121927261
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,768,7168,8,256,32,balanced,0.0890175998210907
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1024,7168,8,256,32,balanced,0.061161600053310394
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1024,7168,8,256,32,balanced,0.07218240015208721
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1024,7168,8,256,32,balanced,0.10722880065441132
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1024,7168,8,256,32,balanced,0.06258880086243153
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1024,7168,8,256,32,balanced,0.05506879985332489
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1024,7168,8,256,32,balanced,0.10732159912586212
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1024,7168,8,256,32,balanced,0.06117439977824688
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1024,7168,8,256,32,balanced,0.05342720039188862
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1024,7168,8,256,32,balanced,0.10404479950666427
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1024,7168,8,256,32,balanced,0.10383360087871552
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1536,7168,8,256,32,balanced,0.06360639967024326
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1536,7168,8,256,32,balanced,0.08531840071082115
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1536,7168,8,256,32,balanced,0.13363839983940123
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1536,7168,8,256,32,balanced,0.06784000061452389
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1536,7168,8,256,32,balanced,0.0649632003158331
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1536,7168,8,256,32,balanced,0.13180799782276154
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1536,7168,8,256,32,balanced,0.06138239987194538
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1536,7168,8,256,32,balanced,0.055587200075387956
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1536,7168,8,256,32,balanced,0.13158720284700393
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1536,7168,8,256,32,balanced,0.12887999936938285
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,2048,7168,8,256,32,balanced,0.06862720064818859
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,2048,7168,8,256,32,balanced,0.0962016001343727
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,2048,7168,8,256,32,balanced,0.16120960116386412
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,2048,7168,8,256,32,balanced,0.06656960062682629
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,2048,7168,8,256,32,balanced,0.0703520007431507
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,2048,7168,8,256,32,balanced,0.16121279895305635
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,2048,7168,8,256,32,balanced,0.0630816001445055
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,2048,7168,8,256,32,balanced,0.06283199936151504
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,2048,7168,8,256,32,balanced,0.15755840092897416
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,2048,7168,8,256,32,balanced,0.15342080146074294
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,3072,7168,8,256,32,balanced,0.07074880003929138
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,3072,7168,8,256,32,balanced,0.11977280229330063
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,3072,7168,8,256,32,balanced,0.21382399797439575
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,3072,7168,8,256,32,balanced,0.07115840055048465
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,3072,7168,8,256,32,balanced,0.08104640021920204
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,3072,7168,8,256,32,balanced,0.21386879980564116
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,3072,7168,8,256,32,balanced,0.06933760046958923
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,3072,7168,8,256,32,balanced,0.06863360032439232
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,3072,7168,8,256,32,balanced,0.2145919993519783
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,3072,7168,8,256,32,balanced,0.19973440021276473
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,65536,7168,8,256,16,balanced,0.4424864023923874
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,65536,7168,8,256,16,balanced,0.5342592179775238
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,65536,7168,8,256,16,balanced,3.483689570426941
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,65536,7168,8,256,16,balanced,3.0586528062820433
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,4096,7168,8,256,32,balanced,0.07683199942111969
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,4096,7168,8,256,32,balanced,0.14737920016050338
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,4096,7168,8,256,32,balanced,0.2685952067375183
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,4096,7168,8,256,32,balanced,0.07649919986724854
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,4096,7168,8,256,32,balanced,0.09556159898638725
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,4096,7168,8,256,32,balanced,0.2692479968070984
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,4096,7168,8,256,32,balanced,0.0725856002420187
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,4096,7168,8,256,32,balanced,0.0767584003508091
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,4096,7168,8,256,32,balanced,0.2692127972841263
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,4096,7168,8,256,32,balanced,0.24850240349769592
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,6144,7168,8,256,32,balanced,0.090608000010252
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,6144,7168,8,256,32,balanced,0.1931072011590004
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,6144,7168,8,256,32,balanced,0.3759872019290924
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,6144,7168,8,256,32,balanced,0.08830080032348633
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,6144,7168,8,256,32,balanced,0.11794240102171898
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,6144,7168,8,256,32,balanced,0.3742944002151489
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,6144,7168,8,256,32,balanced,0.08674559965729714
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,6144,7168,8,256,32,balanced,0.09549760073423386
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,6144,7168,8,256,32,balanced,0.3741311997175217
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,6144,7168,8,256,32,balanced,0.3410176008939743
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,8192,7168,8,256,32,balanced,0.10086399912834168
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,8192,7168,8,256,32,balanced,0.23945279866456987
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,8192,7168,8,256,32,balanced,0.4813887983560562
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,8192,7168,8,256,32,balanced,0.10032959952950478
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,8192,7168,8,256,32,balanced,0.14741120040416716
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,8192,7168,8,256,32,balanced,0.48135679960250854
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,8192,7168,8,256,32,balanced,0.09929599985480309
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,8192,7168,8,256,32,balanced,0.11029760017991067
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,8192,7168,8,256,32,balanced,0.4843103975057602
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,8192,7168,8,256,32,balanced,0.4344736009836197
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,12288,7168,8,256,32,balanced,0.12653119936585427
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,12288,7168,8,256,32,balanced,0.336243200302124
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,12288,7168,8,256,32,balanced,0.7011967957019806
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,12288,7168,8,256,32,balanced,0.12761599943041801
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,12288,7168,8,256,32,balanced,0.19695999920368196
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,12288,7168,8,256,32,balanced,0.698668795824051
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,12288,7168,8,256,32,balanced,0.12399040013551713
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,12288,7168,8,256,32,balanced,0.14032959938049316
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,12288,7168,8,256,32,balanced,0.7004479944705964
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,12288,7168,8,256,32,balanced,0.6254016041755677
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,16384,7168,8,256,32,balanced,0.14929599761962892
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,16384,7168,8,256,32,balanced,0.4342752009630203
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,16384,7168,8,256,32,balanced,0.9151711940765381
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,16384,7168,8,256,32,balanced,0.14845119863748552
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,16384,7168,8,256,32,balanced,0.25008959919214246
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,16384,7168,8,256,32,balanced,0.9131808042526245
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,16384,7168,8,256,32,balanced,0.15069119930267333
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,16384,7168,8,256,32,balanced,0.17137279957532883
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,16384,7168,8,256,32,balanced,0.9136959969997406
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,16384,7168,8,256,32,balanced,0.8086208045482636
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,20480,7168,8,256,32,balanced,0.1727999985218048
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,20480,7168,8,256,32,balanced,0.5312800109386444
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,20480,7168,8,256,32,balanced,1.1405440092086792
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,20480,7168,8,256,32,balanced,0.173055998980999
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,20480,7168,8,256,32,balanced,0.30020159780979155
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,20480,7168,8,256,32,balanced,1.1621471881866454
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,20480,7168,8,256,32,balanced,0.17496320158243178
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,20480,7168,8,256,32,balanced,0.20312319844961166
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,20480,7168,8,256,32,balanced,1.1389055848121643
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,20480,7168,8,256,32,balanced,1.0134528160095215
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,32768,7168,8,256,32,balanced,0.2449087992310524
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,32768,7168,8,256,32,balanced,0.8209791898727417
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,32768,7168,8,256,32,balanced,1.7846304059028626
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,32768,7168,8,256,32,balanced,0.24297920018434524
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,32768,7168,8,256,32,balanced,0.46104000210762025
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,32768,7168,8,256,32,balanced,1.7862176060676576
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,32768,7168,8,256,32,balanced,0.24562559723854066
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,32768,7168,8,256,32,balanced,0.2997247964143753
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,32768,7168,8,256,32,balanced,1.7871935844421387
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,32768,7168,8,256,32,balanced,1.5972832202911378
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,65536,7168,8,256,32,balanced,0.4371904045343399
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,65536,7168,8,256,32,balanced,1.5946687936782837
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,65536,7168,8,256,32,balanced,3.533692789077759
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,65536,7168,8,256,32,balanced,0.4364799976348877
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,65536,7168,8,256,32,balanced,0.8715743958950043
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,65536,7168,8,256,32,balanced,3.535228800773621
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,65536,7168,8,256,32,balanced,0.4347392052412033
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,65536,7168,8,256,32,balanced,0.554780799150467
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,65536,7168,8,256,32,balanced,3.527119994163513
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,65536,7168,8,256,32,balanced,3.090720009803772
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1,7168,8,256,48,balanced,0.07154239974915981
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1,7168,8,256,48,balanced,0.060092799738049506
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1,7168,8,256,48,balanced,0.07612800002098083
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1,7168,8,256,48,balanced,0.07189119942486286
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1,7168,8,256,48,balanced,0.051571199670434
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1,7168,8,256,48,balanced,0.0781856007874012
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1,7168,8,256,48,balanced,0.06824960000813007
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1,7168,8,256,48,balanced,0.05744640044867992
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1,7168,8,256,48,balanced,0.0775616005063057
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1,7168,8,256,48,balanced,0.07293440103530884
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,2,7168,8,256,48,balanced,0.06283840015530587
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,2,7168,8,256,48,balanced,0.05718720033764839
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,2,7168,8,256,48,balanced,0.07271039932966232
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,2,7168,8,256,48,balanced,0.06783679947257042
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,2,7168,8,256,48,balanced,0.05028159990906715
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,2,7168,8,256,48,balanced,0.07055680006742478
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,2,7168,8,256,48,balanced,0.06827200017869473
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,2,7168,8,256,48,balanced,0.04943680018186569
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,2,7168,8,256,48,balanced,0.07366719990968704
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,2,7168,8,256,48,balanced,0.07328320071101188
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,4,7168,8,256,48,balanced,0.059664000570774076
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,4,7168,8,256,48,balanced,0.051187200099229814
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,4,7168,8,256,48,balanced,0.0713344007730484
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,4,7168,8,256,48,balanced,0.05971520058810711
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,4,7168,8,256,48,balanced,0.04931840039789677
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,4,7168,8,256,48,balanced,0.0680736020207405
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,4,7168,8,256,48,balanced,0.061369599774479866
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,4,7168,8,256,48,balanced,0.049107200279831884
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,4,7168,8,256,48,balanced,0.06948480010032654
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,4,7168,8,256,48,balanced,0.06813440062105655
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,8,7168,8,256,48,balanced,0.0635232001543045
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,8,7168,8,256,48,balanced,0.05105919949710369
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,8,7168,8,256,48,balanced,0.07608640044927598
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,8,7168,8,256,48,balanced,0.0656640000641346
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,8,7168,8,256,48,balanced,0.04968640021979809
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,8,7168,8,256,48,balanced,0.07244479954242707
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,8,7168,8,256,48,balanced,0.06282880008220673
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,8,7168,8,256,48,balanced,0.05138239972293377
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,8,7168,8,256,48,balanced,0.07469439879059792
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,8,7168,8,256,48,balanced,0.07223999910056592
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,16,7168,8,256,48,balanced,0.06546240001916885
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,16,7168,8,256,48,balanced,0.05391039997339249
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,16,7168,8,256,48,balanced,0.07677759975194931
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,16,7168,8,256,48,balanced,0.06680960059165955
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,16,7168,8,256,48,balanced,0.04900799989700318
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,16,7168,8,256,48,balanced,0.07565759941935539
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,16,7168,8,256,48,balanced,0.060911998897790906
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,16,7168,8,256,48,balanced,0.05111040025949478
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,16,7168,8,256,48,balanced,0.07457920089364052
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,16,7168,8,256,48,balanced,0.07352960109710693
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,32,7168,8,256,48,balanced,0.061369599774479866
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,32,7168,8,256,48,balanced,0.059225599840283395
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,32,7168,8,256,48,balanced,0.08060160055756568
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,32,7168,8,256,48,balanced,0.05941439978778362
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,32,7168,8,256,48,balanced,0.05151679962873459
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,32,7168,8,256,48,balanced,0.07636800035834312
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,32,7168,8,256,48,balanced,0.05884159952402115
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,32,7168,8,256,48,balanced,0.055958399549126625
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,32,7168,8,256,48,balanced,0.07971519902348519
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,32,7168,8,256,48,balanced,0.07725440114736556
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,48,7168,8,256,48,balanced,0.05726079940795899
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,48,7168,8,256,48,balanced,0.05653440020978451
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,48,7168,8,256,48,balanced,0.0972927987575531
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,48,7168,8,256,48,balanced,0.05941120013594627
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,48,7168,8,256,48,balanced,0.050112000480294225
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,48,7168,8,256,48,balanced,0.07661760002374648
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,48,7168,8,256,48,balanced,0.05591359995305538
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,48,7168,8,256,48,balanced,0.0559008002281189
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,48,7168,8,256,48,balanced,0.07854399904608726
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,48,7168,8,256,48,balanced,0.08212160021066665
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,64,7168,8,256,48,balanced,0.06077760048210621
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,64,7168,8,256,48,balanced,0.05404479913413525
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,64,7168,8,256,48,balanced,0.07511040046811104
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,64,7168,8,256,48,balanced,0.06050560027360916
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,64,7168,8,256,48,balanced,0.04948159977793694
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,64,7168,8,256,48,balanced,0.07220159992575645
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,64,7168,8,256,48,balanced,0.058105599507689475
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,64,7168,8,256,48,balanced,0.04951999932527542
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,64,7168,8,256,48,balanced,0.07600639909505844
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,64,7168,8,256,48,balanced,0.07620479986071586
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,80,7168,8,256,48,balanced,0.05723199956119061
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,80,7168,8,256,48,balanced,0.06091520003974438
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,80,7168,8,256,48,balanced,0.08147839978337287
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,80,7168,8,256,48,balanced,0.0592704001814127
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,80,7168,8,256,48,balanced,0.05334079973399639
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,80,7168,8,256,48,balanced,0.0740607999265194
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,80,7168,8,256,48,balanced,0.05983999967575073
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,80,7168,8,256,48,balanced,0.05141120031476021
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,80,7168,8,256,48,balanced,0.12765440046787263
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,80,7168,8,256,48,balanced,0.07711680009961128
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,96,7168,8,256,48,balanced,0.059356800839304925
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,96,7168,8,256,48,balanced,0.059161600098013876
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,96,7168,8,256,48,balanced,0.07498239949345589
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,96,7168,8,256,48,balanced,0.06266880072653294
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,96,7168,8,256,48,balanced,0.050566400215029714
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,96,7168,8,256,48,balanced,0.07982399985194207
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,96,7168,8,256,48,balanced,0.06500159986317158
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,96,7168,8,256,48,balanced,0.05445760004222393
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,96,7168,8,256,48,balanced,0.07362880036234856
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,96,7168,8,256,48,balanced,0.07310719974339008
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,128,7168,8,256,48,balanced,0.06497599966824055
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,128,7168,8,256,48,balanced,0.05814399980008602
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,128,7168,8,256,48,balanced,0.08051839992403984
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,128,7168,8,256,48,balanced,0.05940159969031811
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,128,7168,8,256,48,balanced,0.048774399980902675
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,128,7168,8,256,48,balanced,0.0777536004781723
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,128,7168,8,256,48,balanced,0.066796800121665
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,128,7168,8,256,48,balanced,0.05165439993143082
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,128,7168,8,256,48,balanced,0.07196160107851028
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,128,7168,8,256,48,balanced,0.07064639925956726
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,160,7168,8,256,48,balanced,0.058876799419522285
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,160,7168,8,256,48,balanced,0.05765760019421577
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,160,7168,8,256,48,balanced,0.0747744008898735
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,160,7168,8,256,48,balanced,0.06832960061728954
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,160,7168,8,256,48,balanced,0.05308159999549389
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,160,7168,8,256,48,balanced,0.07357759922742843
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,160,7168,8,256,48,balanced,0.056665600091218946
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,160,7168,8,256,48,balanced,0.04917439967393875
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,160,7168,8,256,48,balanced,0.07657279968261718
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,160,7168,8,256,48,balanced,0.07455359995365143
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,192,7168,8,256,48,balanced,0.05740479975938797
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,192,7168,8,256,48,balanced,0.059651199728250504
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,192,7168,8,256,48,balanced,0.07496640011668206
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,192,7168,8,256,48,balanced,0.059904000535607335
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,192,7168,8,256,48,balanced,0.05050559975206852
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,192,7168,8,256,48,balanced,0.07831359952688217
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,192,7168,8,256,48,balanced,0.0645951997488737
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,192,7168,8,256,48,balanced,0.05769920088350773
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,192,7168,8,256,48,balanced,0.07503360062837601
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,192,7168,8,256,48,balanced,0.0727680005133152
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,256,7168,8,256,48,balanced,0.0591072004288435
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,256,7168,8,256,48,balanced,0.06352319978177548
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,256,7168,8,256,48,balanced,0.07740800008177758
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,256,7168,8,256,48,balanced,0.05535680018365383
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,256,7168,8,256,48,balanced,0.05708800107240677
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,256,7168,8,256,48,balanced,0.0785375989973545
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,256,7168,8,256,48,balanced,0.05751040019094944
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,256,7168,8,256,48,balanced,0.05546240024268627
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,256,7168,8,256,48,balanced,0.07816959992051124
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,256,7168,8,256,48,balanced,0.08000000044703484
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,320,7168,8,256,48,balanced,0.05557760037481785
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,320,7168,8,256,48,balanced,0.07172480002045631
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,320,7168,8,256,48,balanced,0.08475840091705322
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,320,7168,8,256,48,balanced,0.057238399609923364
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,320,7168,8,256,48,balanced,0.05588159933686256
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,320,7168,8,256,48,balanced,0.08397760018706321
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,320,7168,8,256,48,balanced,0.06093759983778
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,320,7168,8,256,48,balanced,0.05170879997313023
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,320,7168,8,256,48,balanced,0.0868767999112606
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,320,7168,8,256,48,balanced,0.08104320093989373
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,384,7168,8,256,48,balanced,0.057263999804854394
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,384,7168,8,256,48,balanced,0.07671359963715077
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,384,7168,8,256,48,balanced,0.0883520007133484
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,384,7168,8,256,48,balanced,0.060419199988245964
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,384,7168,8,256,48,balanced,0.056748800352215766
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,384,7168,8,256,48,balanced,0.09025280177593231
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,384,7168,8,256,48,balanced,0.06117759980261326
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,384,7168,8,256,48,balanced,0.05477120019495487
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,384,7168,8,256,48,balanced,0.08744320049881935
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,384,7168,8,256,48,balanced,0.0819520004093647
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,512,7168,8,256,48,balanced,0.06221120022237301
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,512,7168,8,256,48,balanced,0.08086400032043457
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,512,7168,8,256,48,balanced,0.1022943988442421
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,512,7168,8,256,48,balanced,0.057971199974417686
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,512,7168,8,256,48,balanced,0.06085759960114956
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,512,7168,8,256,48,balanced,0.10146559923887252
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,512,7168,8,256,48,balanced,0.0649312000721693
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,512,7168,8,256,48,balanced,0.056147200241684916
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,512,7168,8,256,48,balanced,0.10264320001006126
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,512,7168,8,256,48,balanced,0.09667199850082397
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,768,7168,8,256,48,balanced,0.06075839996337891
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,768,7168,8,256,48,balanced,0.09482240080833435
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,768,7168,8,256,48,balanced,0.13135039806365967
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,768,7168,8,256,48,balanced,0.060364799574017525
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,768,7168,8,256,48,balanced,0.07093439996242523
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,768,7168,8,256,48,balanced,0.12628160044550896
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,768,7168,8,256,48,balanced,0.061999999731779096
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,768,7168,8,256,48,balanced,0.06179199926555157
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,768,7168,8,256,48,balanced,0.12665920034050943
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,768,7168,8,256,48,balanced,0.11622079983353614
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1024,7168,8,256,48,balanced,0.05979840010404587
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1024,7168,8,256,48,balanced,0.10886399894952774
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1024,7168,8,256,48,balanced,0.15147520154714583
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1024,7168,8,256,48,balanced,0.061612799391150476
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1024,7168,8,256,48,balanced,0.07553600072860718
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1024,7168,8,256,48,balanced,0.1551615998148918
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1024,7168,8,256,48,balanced,0.06378560028970241
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1024,7168,8,256,48,balanced,0.06728000044822693
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1024,7168,8,256,48,balanced,0.15038720071315764
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1024,7168,8,256,48,balanced,0.14148799926042557
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1536,7168,8,256,48,balanced,0.06207040026783943
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1536,7168,8,256,48,balanced,0.14160319715738295
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1536,7168,8,256,48,balanced,0.20271679908037185
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1536,7168,8,256,48,balanced,0.0637024000287056
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1536,7168,8,256,48,balanced,0.09487039968371391
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1536,7168,8,256,48,balanced,0.2047104001045227
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1536,7168,8,256,48,balanced,0.0648960005491972
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1536,7168,8,256,48,balanced,0.07915199920535088
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1536,7168,8,256,48,balanced,0.20302720069885255
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1536,7168,8,256,48,balanced,0.17949760109186172
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,2048,7168,8,256,48,balanced,0.06761279962956905
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,2048,7168,8,256,48,balanced,0.17268159836530686
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,2048,7168,8,256,48,balanced,0.2509632006287575
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,2048,7168,8,256,48,balanced,0.06729600057005883
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,2048,7168,8,256,48,balanced,0.10688960179686546
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,2048,7168,8,256,48,balanced,0.25220480263233186
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,2048,7168,8,256,48,balanced,0.06594879999756813
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,2048,7168,8,256,48,balanced,0.0875776007771492
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,2048,7168,8,256,48,balanced,0.2506271988153458
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,2048,7168,8,256,48,balanced,0.22127360105514526
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,3072,7168,8,256,48,balanced,0.0739487998187542
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,3072,7168,8,256,48,balanced,0.24047680348157882
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,3072,7168,8,256,48,balanced,0.3489823997020721
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,3072,7168,8,256,48,balanced,0.07219839915633201
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,3072,7168,8,256,48,balanced,0.14022080451250077
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,3072,7168,8,256,48,balanced,0.347759997844696
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,3072,7168,8,256,48,balanced,0.06864960119128227
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,3072,7168,8,256,48,balanced,0.10757760033011436
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,3072,7168,8,256,48,balanced,0.349523201584816
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,3072,7168,8,256,48,balanced,0.298742401599884
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,4096,7168,8,256,48,balanced,0.0761023998260498
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,4096,7168,8,256,48,balanced,0.32002879977226256
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,4096,7168,8,256,48,balanced,0.4506720006465912
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,4096,7168,8,256,48,balanced,0.07966400012373924
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,4096,7168,8,256,48,balanced,0.1775072008371353
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,4096,7168,8,256,48,balanced,0.4564863979816437
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,4096,7168,8,256,48,balanced,0.0785280004143715
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,4096,7168,8,256,48,balanced,0.1290559984743595
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,4096,7168,8,256,48,balanced,0.45607039928436277
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,4096,7168,8,256,48,balanced,0.3779103964567184
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,6144,7168,8,256,48,balanced,0.08965440019965172
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,6144,7168,8,256,48,balanced,0.4430592030286789
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,6144,7168,8,256,48,balanced,0.6480575978755951
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,6144,7168,8,256,48,balanced,0.08860159888863564
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,6144,7168,8,256,48,balanced,0.2445376008749008
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,6144,7168,8,256,48,balanced,0.649590402841568
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,6144,7168,8,256,48,balanced,0.08931519985198974
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,6144,7168,8,256,48,balanced,0.17218879908323287
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,6144,7168,8,256,48,balanced,0.6571423947811127
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,6144,7168,8,256,48,balanced,0.543337607383728
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,8192,7168,8,256,48,balanced,0.10022719949483871
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,8192,7168,8,256,48,balanced,0.5776799917221069
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,8192,7168,8,256,48,balanced,0.8524543881416321
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,8192,7168,8,256,48,balanced,0.10271039977669716
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,8192,7168,8,256,48,balanced,0.31279039978981016
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,8192,7168,8,256,48,balanced,0.8608864068984985
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,8192,7168,8,256,48,balanced,0.09944640025496483
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,8192,7168,8,256,48,balanced,0.2122303992509842
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,8192,7168,8,256,48,balanced,0.852316802740097
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,8192,7168,8,256,48,balanced,0.7003360092639923
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,12288,7168,8,256,48,balanced,0.12485760226845741
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,12288,7168,8,256,48,balanced,0.8549120008945466
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,12288,7168,8,256,48,balanced,1.2520095825195312
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,12288,7168,8,256,48,balanced,0.12471039965748787
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,12288,7168,8,256,48,balanced,0.45717760026454923
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,12288,7168,8,256,48,balanced,1.2533408045768737
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,12288,7168,8,256,48,balanced,0.12641920074820517
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,12288,7168,8,256,48,balanced,0.2963487982749939
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,12288,7168,8,256,48,balanced,1.265497589111328
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,12288,7168,8,256,48,balanced,1.049132800102234
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,16384,7168,8,256,48,balanced,0.14923840016126633
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,16384,7168,8,256,48,balanced,1.1553632020950317
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,16384,7168,8,256,48,balanced,1.670959997177124
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,16384,7168,8,256,48,balanced,0.14907200038433074
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,16384,7168,8,256,48,balanced,0.6094143986701965
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,16384,7168,8,256,48,balanced,1.6768064022064209
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,16384,7168,8,256,48,balanced,0.14901439994573593
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,16384,7168,8,256,48,balanced,0.3839487999677658
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,16384,7168,8,256,48,balanced,1.6683520078659058
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,16384,7168,8,256,48,balanced,1.3996384143829346
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,20480,7168,8,256,48,balanced,0.17210559993982316
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,20480,7168,8,256,48,balanced,1.4407519817352294
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,20480,7168,8,256,48,balanced,2.085142397880554
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,20480,7168,8,256,48,balanced,0.17627200186252595
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,20480,7168,8,256,48,balanced,0.757779198884964
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,20480,7168,8,256,48,balanced,2.0891168117523193
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,20480,7168,8,256,48,balanced,0.17147519886493684
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,20480,7168,8,256,48,balanced,0.46706879436969756
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,20480,7168,8,256,48,balanced,2.0877120018005373
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,20480,7168,8,256,48,balanced,1.7237087965011597
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,32768,7168,8,256,48,balanced,0.24744319915771484
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,32768,7168,8,256,48,balanced,2.3483808040618896
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,32768,7168,8,256,48,balanced,3.3279263973236084
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,32768,7168,8,256,48,balanced,0.2523488014936447
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,32768,7168,8,256,48,balanced,1.2166208148002624
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,32768,7168,8,256,48,balanced,3.331222414970398
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,32768,7168,8,256,48,balanced,0.244854399561882
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,32768,7168,8,256,48,balanced,0.737391996383667
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,32768,7168,8,256,48,balanced,3.329964804649353
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,32768,7168,8,256,48,balanced,2.627942371368408
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,65536,7168,8,256,48,balanced,0.4361407995223999
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,65536,7168,8,256,48,balanced,4.843750381469727
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,65536,7168,8,256,48,balanced,6.64136323928833
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,65536,7168,8,256,48,balanced,0.4372895985841751
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,65536,7168,8,256,48,balanced,2.490227222442627
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,65536,7168,8,256,48,balanced,6.642566394805908
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,65536,7168,8,256,48,balanced,0.43616960048675535
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,65536,7168,8,256,48,balanced,1.4609439969062805
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,65536,7168,8,256,48,balanced,6.637513589859009
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,65536,7168,8,256,48,balanced,5.174719953536988
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1,7168,8,256,72,balanced,0.07623680010437965
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1,7168,8,256,72,balanced,0.055875199660658834
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1,7168,8,256,72,balanced,0.08272639885544777
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1,7168,8,256,72,balanced,0.06848959922790528
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1,7168,8,256,72,balanced,0.057132800295948985
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1,7168,8,256,72,balanced,0.07848000004887581
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1,7168,8,256,72,balanced,0.0693599995225668
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1,7168,8,256,72,balanced,0.05710079967975616
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1,7168,8,256,72,balanced,0.07930240035057068
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1,7168,8,256,72,balanced,0.082566399872303
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,2,7168,8,256,72,balanced,0.0652127992361784
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,2,7168,8,256,72,balanced,0.0741887990385294
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,2,7168,8,256,72,balanced,0.07928639948368073
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,2,7168,8,256,72,balanced,0.079203200340271
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,2,7168,8,256,72,balanced,0.06361600048840046
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,2,7168,8,256,72,balanced,0.07702079862356186
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,2,7168,8,256,72,balanced,0.07953600063920022
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,2,7168,8,256,72,balanced,0.0630016002804041
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,2,7168,8,256,72,balanced,0.08104640021920204
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,2,7168,8,256,72,balanced,0.07266560047864914
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,4,7168,8,256,72,balanced,0.07671359926462173
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,4,7168,8,256,72,balanced,0.06265279948711396
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,4,7168,8,256,72,balanced,0.07805120050907136
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,4,7168,8,256,72,balanced,0.07358399853110313
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,4,7168,8,256,72,balanced,0.061792000010609624
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,4,7168,8,256,72,balanced,0.07275520041584968
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,4,7168,8,256,72,balanced,0.08064639940857887
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,4,7168,8,256,72,balanced,0.058687999844551086
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,4,7168,8,256,72,balanced,0.07597760036587715
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,4,7168,8,256,72,balanced,0.07961279898881912
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,8,7168,8,256,72,balanced,0.06953280083835126
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,8,7168,8,256,72,balanced,0.05797760039567947
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,8,7168,8,256,72,balanced,0.0988640010356903
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,8,7168,8,256,72,balanced,0.0694816004484892
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,8,7168,8,256,72,balanced,0.05717119984328747
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,8,7168,8,256,72,balanced,0.09634559899568558
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,8,7168,8,256,72,balanced,0.0718207996338606
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,8,7168,8,256,72,balanced,0.0601087998598814
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,8,7168,8,256,72,balanced,0.10302720069885254
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,8,7168,8,256,72,balanced,0.08484800085425377
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,16,7168,8,256,72,balanced,0.06874239966273307
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,16,7168,8,256,72,balanced,0.06142719946801663
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,16,7168,8,256,72,balanced,0.09746879935264588
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,16,7168,8,256,72,balanced,0.06862719990313053
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,16,7168,8,256,72,balanced,0.06244799979031086
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,16,7168,8,256,72,balanced,0.09402879998087883
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,16,7168,8,256,72,balanced,0.07405760027468204
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,16,7168,8,256,72,balanced,0.05762879997491836
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,16,7168,8,256,72,balanced,0.09832959994673729
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,16,7168,8,256,72,balanced,0.09010560065507889
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,32,7168,8,256,72,balanced,0.07380160018801689
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,32,7168,8,256,72,balanced,0.061219199746847155
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,32,7168,8,256,72,balanced,0.09639999940991402
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,32,7168,8,256,72,balanced,0.07169919982552528
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,32,7168,8,256,72,balanced,0.05674239993095398
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,32,7168,8,256,72,balanced,0.10992959961295128
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,32,7168,8,256,72,balanced,0.06957760080695152
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,32,7168,8,256,72,balanced,0.06538879983127117
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,32,7168,8,256,72,balanced,0.0993184007704258
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,32,7168,8,256,72,balanced,0.10189760029315949
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,48,7168,8,256,72,balanced,0.07177599966526031
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,48,7168,8,256,72,balanced,0.06465280055999756
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,48,7168,8,256,72,balanced,0.09938879981637001
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,48,7168,8,256,72,balanced,0.06934719905257225
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,48,7168,8,256,72,balanced,0.05742719992995262
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,48,7168,8,256,72,balanced,0.10239360108971596
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,48,7168,8,256,72,balanced,0.06674879938364028
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,48,7168,8,256,72,balanced,0.05959679968655109
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,48,7168,8,256,72,balanced,0.09803839847445488
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,48,7168,8,256,72,balanced,0.09247999861836434
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,64,7168,8,256,72,balanced,0.06682240031659603
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,64,7168,8,256,72,balanced,0.0767295990139246
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,64,7168,8,256,72,balanced,0.10184319913387299
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,64,7168,8,256,72,balanced,0.06852159984409809
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,64,7168,8,256,72,balanced,0.06194560006260872
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,64,7168,8,256,72,balanced,0.10344000011682511
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,64,7168,8,256,72,balanced,0.06978879906237126
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,64,7168,8,256,72,balanced,0.05931840017437935
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,64,7168,8,256,72,balanced,0.10274240002036095
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,64,7168,8,256,72,balanced,0.0975903995335102
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,80,7168,8,256,72,balanced,0.07531519941985607
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,80,7168,8,256,72,balanced,0.06767360009253025
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,80,7168,8,256,72,balanced,0.10279359966516495
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,80,7168,8,256,72,balanced,0.06954879947006702
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,80,7168,8,256,72,balanced,0.08647040128707886
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,80,7168,8,256,72,balanced,0.09485119953751564
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,80,7168,8,256,72,balanced,0.06891519911587238
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,80,7168,8,256,72,balanced,0.05849600024521351
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,80,7168,8,256,72,balanced,0.10300160050392151
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,80,7168,8,256,72,balanced,0.09779840037226677
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,96,7168,8,256,72,balanced,0.06987840011715889
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,96,7168,8,256,72,balanced,0.06478079929947853
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,96,7168,8,256,72,balanced,0.10106560066342354
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,96,7168,8,256,72,balanced,0.07060800045728684
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,96,7168,8,256,72,balanced,0.05881279967725277
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,96,7168,8,256,72,balanced,0.10382080078125
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,96,7168,8,256,72,balanced,0.07081600055098533
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,96,7168,8,256,72,balanced,0.05760320015251637
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,96,7168,8,256,72,balanced,0.10428159981966019
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,96,7168,8,256,72,balanced,0.09305280074477196
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,128,7168,8,256,72,balanced,0.06746239960193634
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,128,7168,8,256,72,balanced,0.06955840066075325
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,128,7168,8,256,72,balanced,0.09910719990730285
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,128,7168,8,256,72,balanced,0.0736991997808218
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,128,7168,8,256,72,balanced,0.06215999945998192
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,128,7168,8,256,72,balanced,0.0993984006345272
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,128,7168,8,256,72,balanced,0.07261120155453682
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,128,7168,8,256,72,balanced,0.058793599531054494
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,128,7168,8,256,72,balanced,0.1027872011065483
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,128,7168,8,256,72,balanced,0.09311359971761704
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,160,7168,8,256,72,balanced,0.07368640005588531
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,160,7168,8,256,72,balanced,0.0796671997755766
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,160,7168,8,256,72,balanced,0.10013439804315567
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,160,7168,8,256,72,balanced,0.06894720010459424
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,160,7168,8,256,72,balanced,0.06712960042059421
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,160,7168,8,256,72,balanced,0.10097279995679856
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,160,7168,8,256,72,balanced,0.07323519960045814
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,160,7168,8,256,72,balanced,0.06306880041956901
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,160,7168,8,256,72,balanced,0.10141120031476021
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,160,7168,8,256,72,balanced,0.09667840003967285
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,192,7168,8,256,72,balanced,0.07094720005989075
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,192,7168,8,256,72,balanced,0.0763264000415802
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,192,7168,8,256,72,balanced,0.10158080011606216
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,192,7168,8,256,72,balanced,0.07354880049824715
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,192,7168,8,256,72,balanced,0.062323200330138204
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,192,7168,8,256,72,balanced,0.10192319899797439
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,192,7168,8,256,72,balanced,0.080137600004673
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,192,7168,8,256,72,balanced,0.06253440007567405
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,192,7168,8,256,72,balanced,0.10366719886660576
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,192,7168,8,256,72,balanced,0.1080671988427639
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,256,7168,8,256,72,balanced,0.07513919994235038
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,256,7168,8,256,72,balanced,0.07451200038194657
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,256,7168,8,256,72,balanced,0.10118400081992149
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,256,7168,8,256,72,balanced,0.0704352006316185
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,256,7168,8,256,72,balanced,0.06424319967627526
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,256,7168,8,256,72,balanced,0.10875199958682061
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,256,7168,8,256,72,balanced,0.07246719971299172
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,256,7168,8,256,72,balanced,0.06004800014197827
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,256,7168,8,256,72,balanced,0.10127040147781372
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,256,7168,8,256,72,balanced,0.09603200107812881
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,320,7168,8,256,72,balanced,0.07531520016491414
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,320,7168,8,256,72,balanced,0.079721599817276
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,320,7168,8,256,72,balanced,0.10224959924817086
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,320,7168,8,256,72,balanced,0.06796800121665
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,320,7168,8,256,72,balanced,0.06084479987621307
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,320,7168,8,256,72,balanced,0.09935679882764817
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,320,7168,8,256,72,balanced,0.0676864005625248
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,320,7168,8,256,72,balanced,0.0657920002937317
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,320,7168,8,256,72,balanced,0.10554880127310753
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,320,7168,8,256,72,balanced,0.09610880017280579
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,384,7168,8,256,72,balanced,0.07080960050225257
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,384,7168,8,256,72,balanced,0.07959999963641166
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,384,7168,8,256,72,balanced,0.11362880021333695
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,384,7168,8,256,72,balanced,0.0723872009664774
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,384,7168,8,256,72,balanced,0.06413120031356812
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,384,7168,8,256,72,balanced,0.11022400110960007
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,384,7168,8,256,72,balanced,0.06390400007367134
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,384,7168,8,256,72,balanced,0.06604159958660602
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,384,7168,8,256,72,balanced,0.11055999994277954
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,384,7168,8,256,72,balanced,0.10358720049262046
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,512,7168,8,256,72,balanced,0.06844159848988056
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,512,7168,8,256,72,balanced,0.09531519860029221
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,512,7168,8,256,72,balanced,0.13104960173368455
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,512,7168,8,256,72,balanced,0.07499519959092141
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,512,7168,8,256,72,balanced,0.07010559961199761
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,512,7168,8,256,72,balanced,0.13231039866805078
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,512,7168,8,256,72,balanced,0.0742976002395153
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,512,7168,8,256,72,balanced,0.07241280004382133
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,512,7168,8,256,72,balanced,0.12798719927668573
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,512,7168,8,256,72,balanced,0.11342400088906288
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,768,7168,8,256,72,balanced,0.07471679970622062
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,768,7168,8,256,72,balanced,0.11649919971823693
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,768,7168,8,256,72,balanced,0.16281920075416564
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,768,7168,8,256,72,balanced,0.07488960064947606
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,768,7168,8,256,72,balanced,0.08275519981980324
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,768,7168,8,256,72,balanced,0.1652319997549057
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,768,7168,8,256,72,balanced,0.0732352003455162
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,768,7168,8,256,72,balanced,0.07297280058264732
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,768,7168,8,256,72,balanced,0.16295679956674575
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,768,7168,8,256,72,balanced,0.1434688001871109
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1024,7168,8,256,72,balanced,0.07409600019454957
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1024,7168,8,256,72,balanced,0.14327040016651155
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1024,7168,8,256,72,balanced,0.20195519924163818
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1024,7168,8,256,72,balanced,0.07481920048594475
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1024,7168,8,256,72,balanced,0.09036480039358138
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1024,7168,8,256,72,balanced,0.19570880085229875
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1024,7168,8,256,72,balanced,0.07587200030684471
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1024,7168,8,256,72,balanced,0.07956800013780593
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1024,7168,8,256,72,balanced,0.20078079998493195
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1024,7168,8,256,72,balanced,0.17196160107851027
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1536,7168,8,256,72,balanced,0.07663039974868298
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1536,7168,8,256,72,balanced,0.19339200109243393
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1536,7168,8,256,72,balanced,0.2680351972579956
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1536,7168,8,256,72,balanced,0.07653759866952896
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1536,7168,8,256,72,balanced,0.11584319844841957
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1536,7168,8,256,72,balanced,0.26994240283966064
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1536,7168,8,256,72,balanced,0.0794144008308649
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1536,7168,8,256,72,balanced,0.09376959949731827
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1536,7168,8,256,72,balanced,0.27045120000839235
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1536,7168,8,256,72,balanced,0.22490560114383698
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,2048,7168,8,256,72,balanced,0.07961600050330161
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,2048,7168,8,256,72,balanced,0.24887040108442307
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,2048,7168,8,256,72,balanced,0.3409695982933044
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,2048,7168,8,256,72,balanced,0.07665600031614303
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,2048,7168,8,256,72,balanced,0.14377280026674272
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,2048,7168,8,256,72,balanced,0.3389376014471054
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,2048,7168,8,256,72,balanced,0.07556800022721291
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,2048,7168,8,256,72,balanced,0.1118367999792099
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,2048,7168,8,256,72,balanced,0.3820479989051819
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,2048,7168,8,256,72,balanced,0.28406719863414764
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,3072,7168,8,256,72,balanced,0.08442239910364151
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,3072,7168,8,256,72,balanced,0.3556448042392731
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,3072,7168,8,256,72,balanced,0.4794624000787735
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,3072,7168,8,256,72,balanced,0.0885728009045124
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,3072,7168,8,256,72,balanced,0.19441599994897843
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,3072,7168,8,256,72,balanced,0.4779839992523193
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,3072,7168,8,256,72,balanced,0.0921343993395567
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,3072,7168,8,256,72,balanced,0.14232960045337678
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,3072,7168,8,256,72,balanced,0.4751967966556549
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,3072,7168,8,256,72,balanced,0.3810399979352951
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,4096,7168,8,256,72,balanced,0.0853887990117073
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,4096,7168,8,256,72,balanced,0.4728927999734879
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,4096,7168,8,256,72,balanced,0.6184767961502076
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,4096,7168,8,256,72,balanced,0.0819232001900673
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,4096,7168,8,256,72,balanced,0.24622079730033875
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,4096,7168,8,256,72,balanced,0.6220031976699829
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,4096,7168,8,256,72,balanced,0.08496959954500198
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,4096,7168,8,256,72,balanced,0.17639039903879167
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,4096,7168,8,256,72,balanced,0.6204351961612702
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,4096,7168,8,256,72,balanced,0.4881568044424057
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,6144,7168,8,256,72,balanced,0.10021760165691376
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,6144,7168,8,256,72,balanced,0.708025598526001
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,6144,7168,8,256,72,balanced,0.9118528068065643
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,6144,7168,8,256,72,balanced,0.10114559978246689
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,6144,7168,8,256,72,balanced,0.34864960312843324
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,6144,7168,8,256,72,balanced,0.9125952005386353
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,6144,7168,8,256,72,balanced,0.09788480177521705
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,6144,7168,8,256,72,balanced,0.24038719832897187
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,6144,7168,8,256,72,balanced,0.9050303936004639
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,6144,7168,8,256,72,balanced,0.7003327965736389
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,8192,7168,8,256,72,balanced,0.11206720098853111
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,8192,7168,8,256,72,balanced,0.9583615958690643
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,8192,7168,8,256,72,balanced,1.2014975905418397
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,8192,7168,8,256,72,balanced,0.10927359983325005
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,8192,7168,8,256,72,balanced,0.4628064066171646
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,8192,7168,8,256,72,balanced,1.1980160236358643
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,8192,7168,8,256,72,balanced,0.1176576018333435
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,8192,7168,8,256,72,balanced,0.30736319720745087
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,8192,7168,8,256,72,balanced,1.205235195159912
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,8192,7168,8,256,72,balanced,0.9078656017780304
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,12288,7168,8,256,72,balanced,0.13623359948396682
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,12288,7168,8,256,72,balanced,1.456064021587372
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,12288,7168,8,256,72,balanced,1.773308789730072
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,12288,7168,8,256,72,balanced,0.13395199850201606
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,12288,7168,8,256,72,balanced,0.6814560055732727
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,12288,7168,8,256,72,balanced,1.7722559928894044
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,12288,7168,8,256,72,balanced,0.13601919934153556
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,12288,7168,8,256,72,balanced,0.4351199984550476
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,12288,7168,8,256,72,balanced,1.7653696179389953
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,12288,7168,8,256,72,balanced,1.3682175874710083
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,16384,7168,8,256,72,balanced,0.1589599996805191
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,16384,7168,8,256,72,balanced,1.9788576126098634
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,16384,7168,8,256,72,balanced,2.3538559913635253
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,16384,7168,8,256,72,balanced,0.15932160019874572
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,16384,7168,8,256,72,balanced,0.9060288071632385
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,16384,7168,8,256,72,balanced,2.349884796142578
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,16384,7168,8,256,72,balanced,0.18014719635248183
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,16384,7168,8,256,72,balanced,0.563344007730484
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,16384,7168,8,256,72,balanced,2.348912000656128
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,16384,7168,8,256,72,balanced,1.7720800042152405
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,20480,7168,8,256,72,balanced,0.18157760053873062
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,20480,7168,8,256,72,balanced,2.5248735666275026
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,20480,7168,8,256,72,balanced,2.941769599914551
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,20480,7168,8,256,72,balanced,0.18543680012226105
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,20480,7168,8,256,72,balanced,1.1451359868049622
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,20480,7168,8,256,72,balanced,2.947798418998718
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,20480,7168,8,256,72,balanced,0.1876479998230934
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,20480,7168,8,256,72,balanced,0.7076895952224731
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,20480,7168,8,256,72,balanced,2.9441855907440186
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,20480,7168,8,256,72,balanced,2.20230085849762
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,32768,7168,8,256,72,balanced,0.2609599992632866
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,32768,7168,8,256,72,balanced,4.229142379760742
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,32768,7168,8,256,72,balanced,4.701219177246093
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,32768,7168,8,256,72,balanced,0.25882239937782286
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,32768,7168,8,256,72,balanced,2.022857630252838
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,32768,7168,8,256,72,balanced,4.696908807754516
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,32768,7168,8,256,72,balanced,0.26176639795303347
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,32768,7168,8,256,72,balanced,1.1211999893188476
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,32768,7168,8,256,72,balanced,4.694073581695557
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,32768,7168,8,256,72,balanced,3.421766424179077
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,65536,7168,8,256,72,balanced,0.45503680109977723
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,65536,7168,8,256,72,balanced,8.815062427520752
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,65536,7168,8,256,72,balanced,9.374054622650146
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,65536,7168,8,256,72,balanced,0.45753919780254365
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,65536,7168,8,256,72,balanced,4.374367904663086
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,65536,7168,8,256,72,balanced,9.37174711227417
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,65536,7168,8,256,72,balanced,0.45125439763069153
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,65536,7168,8,256,72,balanced,2.244537615776062
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,65536,7168,8,256,72,balanced,9.382166385650635
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,65536,7168,8,256,72,balanced,6.720047950744629
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1,7168,8,256,64,balanced,0.07420160099864007
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1,7168,8,256,64,balanced,0.05947519950568676
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1,7168,8,256,64,balanced,0.07439040020108223
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1,7168,8,256,64,balanced,0.0683104008436203
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1,7168,8,256,64,balanced,0.06029119938611984
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1,7168,8,256,64,balanced,0.07852799966931343
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1,7168,8,256,64,balanced,0.07038719914853572
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1,7168,8,256,64,balanced,0.06285760030150414
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1,7168,8,256,64,balanced,0.07820479944348335
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1,7168,8,256,64,balanced,0.08264959901571274
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,2,7168,8,256,64,balanced,0.06471359990537166
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,2,7168,8,256,64,balanced,0.06154560036957264
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,2,7168,8,256,64,balanced,0.07645759880542755
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,2,7168,8,256,64,balanced,0.06586559899151326
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,2,7168,8,256,64,balanced,0.06315519958734513
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,2,7168,8,256,64,balanced,0.07512960061430932
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,2,7168,8,256,64,balanced,0.06465919949114322
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,2,7168,8,256,64,balanced,0.06181440092623234
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,2,7168,8,256,64,balanced,0.07493119984865189
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,2,7168,8,256,64,balanced,0.07515520006418228
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,4,7168,8,256,64,balanced,0.06996480077505111
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,4,7168,8,256,64,balanced,0.060742399841547015
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,4,7168,8,256,64,balanced,0.07468799948692321
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,4,7168,8,256,64,balanced,0.0679296001791954
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,4,7168,8,256,64,balanced,0.06135679967701435
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,4,7168,8,256,64,balanced,0.0732863999903202
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,4,7168,8,256,64,balanced,0.07077759988605976
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,4,7168,8,256,64,balanced,0.09243519976735115
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,4,7168,8,256,64,balanced,0.07895359992980958
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,4,7168,8,256,64,balanced,0.08318720012903214
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,8,7168,8,256,64,balanced,0.08139199949800968
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,8,7168,8,256,64,balanced,0.06618879996240139
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,8,7168,8,256,64,balanced,0.08015680089592933
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,8,7168,8,256,64,balanced,0.07426880002021789
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,8,7168,8,256,64,balanced,0.06786239966750145
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,8,7168,8,256,64,balanced,0.07093119993805885
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,8,7168,8,256,64,balanced,0.08000320009887218
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,8,7168,8,256,64,balanced,0.07028160020709037
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,8,7168,8,256,64,balanced,0.07587519958615303
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,8,7168,8,256,64,balanced,0.08357119932770729
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,16,7168,8,256,64,balanced,0.07900159843266011
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,16,7168,8,256,64,balanced,0.071951999142766
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,16,7168,8,256,64,balanced,0.10594239979982376
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,16,7168,8,256,64,balanced,0.08010879978537559
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,16,7168,8,256,64,balanced,0.06576960012316704
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,16,7168,8,256,64,balanced,0.09912000000476837
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,16,7168,8,256,64,balanced,0.07088639996945859
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,16,7168,8,256,64,balanced,0.06748159974813461
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,16,7168,8,256,64,balanced,0.11350079849362374
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,16,7168,8,256,64,balanced,0.09277440011501312
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,32,7168,8,256,64,balanced,0.07105600014328957
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,32,7168,8,256,64,balanced,0.0716352004557848
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,32,7168,8,256,64,balanced,0.10190079957246781
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,32,7168,8,256,64,balanced,0.07334399968385696
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,32,7168,8,256,64,balanced,0.06471359878778457
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,32,7168,8,256,64,balanced,0.10960959941148758
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,32,7168,8,256,64,balanced,0.0749088004231453
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,32,7168,8,256,64,balanced,0.0639679990708828
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,32,7168,8,256,64,balanced,0.1092448003590107
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,32,7168,8,256,64,balanced,0.10258879885077477
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,48,7168,8,256,64,balanced,0.06900159940123558
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,48,7168,8,256,64,balanced,0.07127039954066276
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,48,7168,8,256,64,balanced,0.10166080072522163
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,48,7168,8,256,64,balanced,0.0664512000977993
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,48,7168,8,256,64,balanced,0.06104640066623688
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,48,7168,8,256,64,balanced,0.09577280059456825
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,48,7168,8,256,64,balanced,0.07218240015208721
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,48,7168,8,256,64,balanced,0.0644799992442131
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,48,7168,8,256,64,balanced,0.10887359976768493
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,48,7168,8,256,64,balanced,0.10156799927353859
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,64,7168,8,256,64,balanced,0.06777920015156269
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,64,7168,8,256,64,balanced,0.08017599806189538
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,64,7168,8,256,64,balanced,0.11527359932661056
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,64,7168,8,256,64,balanced,0.06853759996592998
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,64,7168,8,256,64,balanced,0.05858239978551864
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,64,7168,8,256,64,balanced,0.0974752001464367
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,64,7168,8,256,64,balanced,0.07033280022442341
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,64,7168,8,256,64,balanced,0.06461759991943836
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,64,7168,8,256,64,balanced,0.094428800791502
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,64,7168,8,256,64,balanced,0.11479999870061874
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,80,7168,8,256,64,balanced,0.07345280051231384
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,80,7168,8,256,64,balanced,0.09675519876182079
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,80,7168,8,256,64,balanced,0.09830400049686432
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,80,7168,8,256,64,balanced,0.06873279884457588
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,80,7168,8,256,64,balanced,0.06478400006890297
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,80,7168,8,256,64,balanced,0.11025599911808967
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,80,7168,8,256,64,balanced,0.06922559961676597
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,80,7168,8,256,64,balanced,0.06292799897491932
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,80,7168,8,256,64,balanced,0.10101759880781173
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,80,7168,8,256,64,balanced,0.1026528000831604
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,96,7168,8,256,64,balanced,0.06821119971573353
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,96,7168,8,256,64,balanced,0.0629248008131981
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,96,7168,8,256,64,balanced,0.09292799979448318
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,96,7168,8,256,64,balanced,0.06837760023772717
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,96,7168,8,256,64,balanced,0.06133440099656582
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,96,7168,8,256,64,balanced,0.103404800593853
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,96,7168,8,256,64,balanced,0.0691904004663229
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,96,7168,8,256,64,balanced,0.06360960043966771
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,96,7168,8,256,64,balanced,0.10278400033712387
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,96,7168,8,256,64,balanced,0.09844800010323525
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,128,7168,8,256,64,balanced,0.06971519961953163
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,128,7168,8,256,64,balanced,0.06627199985086918
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,128,7168,8,256,64,balanced,0.09491519927978516
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,128,7168,8,256,64,balanced,0.07089919969439507
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,128,7168,8,256,64,balanced,0.06283520050346851
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,128,7168,8,256,64,balanced,0.10157439783215523
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,128,7168,8,256,64,balanced,0.07557440102100373
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,128,7168,8,256,64,balanced,0.06059199944138527
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,128,7168,8,256,64,balanced,0.10094399973750115
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,128,7168,8,256,64,balanced,0.10420799925923348
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,160,7168,8,256,64,balanced,0.0681247990578413
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,160,7168,8,256,64,balanced,0.06890880018472671
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,160,7168,8,256,64,balanced,0.09953600019216538
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,160,7168,8,256,64,balanced,0.07396799996495247
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,160,7168,8,256,64,balanced,0.06091520003974438
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,160,7168,8,256,64,balanced,0.09405440017580986
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,160,7168,8,256,64,balanced,0.06464639976620674
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,160,7168,8,256,64,balanced,0.06146559938788414
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,160,7168,8,256,64,balanced,0.10179199948906899
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,160,7168,8,256,64,balanced,0.10169920027256012
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,192,7168,8,256,64,balanced,0.07420480065047741
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,192,7168,8,256,64,balanced,0.07441920042037964
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,192,7168,8,256,64,balanced,0.10270079970359802
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,192,7168,8,256,64,balanced,0.06585280038416386
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,192,7168,8,256,64,balanced,0.05882879979908466
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,192,7168,8,256,64,balanced,0.09602559953927994
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,192,7168,8,256,64,balanced,0.06879999935626983
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,192,7168,8,256,64,balanced,0.061500801146030425
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,192,7168,8,256,64,balanced,0.09592959880828858
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,192,7168,8,256,64,balanced,0.09792959988117218
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,256,7168,8,256,64,balanced,0.06996160037815571
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,256,7168,8,256,64,balanced,0.07031679973006248
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,256,7168,8,256,64,balanced,0.09487359896302223
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,256,7168,8,256,64,balanced,0.06970559991896152
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,256,7168,8,256,64,balanced,0.05742399953305721
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,256,7168,8,256,64,balanced,0.10136960074305534
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,256,7168,8,256,64,balanced,0.07466240078210831
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,256,7168,8,256,64,balanced,0.06303999982774258
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,256,7168,8,256,64,balanced,0.1015104003250599
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,256,7168,8,256,64,balanced,0.09523200020194053
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,320,7168,8,256,64,balanced,0.06948159895837307
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,320,7168,8,256,64,balanced,0.07449600100517273
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,320,7168,8,256,64,balanced,0.10009279996156692
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,320,7168,8,256,64,balanced,0.07146240063011647
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,320,7168,8,256,64,balanced,0.062201600149273874
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,320,7168,8,256,64,balanced,0.10308480039238929
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,320,7168,8,256,64,balanced,0.07500160075724124
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,320,7168,8,256,64,balanced,0.06243520081043243
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,320,7168,8,256,64,balanced,0.09617599993944168
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,320,7168,8,256,64,balanced,0.10054080039262772
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,384,7168,8,256,64,balanced,0.07147519923746586
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,384,7168,8,256,64,balanced,0.07850239984691143
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,384,7168,8,256,64,balanced,0.1071615993976593
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,384,7168,8,256,64,balanced,0.06797759979963303
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,384,7168,8,256,64,balanced,0.06306560002267361
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,384,7168,8,256,64,balanced,0.1070080004632473
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,384,7168,8,256,64,balanced,0.07433920055627823
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,384,7168,8,256,64,balanced,0.06970239989459515
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,384,7168,8,256,64,balanced,0.09653440043330193
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,384,7168,8,256,64,balanced,0.0977183997631073
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,512,7168,8,256,64,balanced,0.07835520058870316
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,512,7168,8,256,64,balanced,0.08243839889764785
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,512,7168,8,256,64,balanced,0.10730240121483803
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,512,7168,8,256,64,balanced,0.0719904001802206
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,512,7168,8,256,64,balanced,0.0639552004635334
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,512,7168,8,256,64,balanced,0.1043040007352829
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,512,7168,8,256,64,balanced,0.07061439976096154
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,512,7168,8,256,64,balanced,0.07017600052058696
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,512,7168,8,256,64,balanced,0.10526399835944175
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,512,7168,8,256,64,balanced,0.09853760153055191
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,768,7168,8,256,64,balanced,0.07294080033898354
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,768,7168,8,256,64,balanced,0.09669440016150474
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,768,7168,8,256,64,balanced,0.12235200256109238
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,768,7168,8,256,64,balanced,0.07431359849870205
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,768,7168,8,256,64,balanced,0.07129920087754726
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,768,7168,8,256,64,balanced,0.12343039810657501
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,768,7168,8,256,64,balanced,0.07027839943766594
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,768,7168,8,256,64,balanced,0.06547839939594269
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,768,7168,8,256,64,balanced,0.12167039886116982
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,768,7168,8,256,64,balanced,0.11365439966320992
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1024,7168,8,256,64,balanced,0.07332159988582135
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1024,7168,8,256,64,balanced,0.10175360068678856
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1024,7168,8,256,64,balanced,0.14289599955081939
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1024,7168,8,256,64,balanced,0.08078720085322857
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1024,7168,8,256,64,balanced,0.07742399945855141
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1024,7168,8,256,64,balanced,0.13836159855127333
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1024,7168,8,256,64,balanced,0.07093120031058789
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1024,7168,8,256,64,balanced,0.06983359977602958
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1024,7168,8,256,64,balanced,0.14085760116577148
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1024,7168,8,256,64,balanced,0.1340768001973629
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1536,7168,8,256,64,balanced,0.07627199962735176
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1536,7168,8,256,64,balanced,0.13137919828295708
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1536,7168,8,256,64,balanced,0.17999999821186066
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1536,7168,8,256,64,balanced,0.07823359966278076
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1536,7168,8,256,64,balanced,0.09042240008711815
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1536,7168,8,256,64,balanced,0.1809440016746521
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1536,7168,8,256,64,balanced,0.0732575997710228
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1536,7168,8,256,64,balanced,0.07608960047364235
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1536,7168,8,256,64,balanced,0.17912640273571015
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1536,7168,8,256,64,balanced,0.16787839978933333
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,2048,7168,8,256,64,balanced,0.07487679906189441
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,2048,7168,8,256,64,balanced,0.15593919903039932
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,2048,7168,8,256,64,balanced,0.22103039920330048
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,2048,7168,8,256,64,balanced,0.07648640051484108
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,2048,7168,8,256,64,balanced,0.10453439950942993
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,2048,7168,8,256,64,balanced,0.22984319925308228
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,2048,7168,8,256,64,balanced,0.09535040073096752
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,2048,7168,8,256,64,balanced,0.08641600012779235
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,2048,7168,8,256,64,balanced,0.22378239929676055
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,2048,7168,8,256,64,balanced,0.20217280089855194
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,3072,7168,8,256,64,balanced,0.08016640022397041
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,3072,7168,8,256,64,balanced,0.20643199980258942
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,3072,7168,8,256,64,balanced,0.3020832002162933
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,3072,7168,8,256,64,balanced,0.0797824002802372
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,3072,7168,8,256,64,balanced,0.1295551985502243
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,3072,7168,8,256,64,balanced,0.3253119945526123
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,3072,7168,8,256,64,balanced,0.08743359968066215
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,3072,7168,8,256,64,balanced,0.10324799865484238
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,3072,7168,8,256,64,balanced,0.3021952033042908
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,3072,7168,8,256,64,balanced,0.2692319989204407
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,4096,7168,8,256,64,balanced,0.08228479996323586
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,4096,7168,8,256,64,balanced,0.26133120357990264
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,4096,7168,8,256,64,balanced,0.3827711999416351
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,4096,7168,8,256,64,balanced,0.08574720174074173
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,4096,7168,8,256,64,balanced,0.1538335993885994
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,4096,7168,8,256,64,balanced,0.38486399948596955
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,4096,7168,8,256,64,balanced,0.08413120061159134
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,4096,7168,8,256,64,balanced,0.11988160014152527
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,4096,7168,8,256,64,balanced,0.3830751985311508
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,4096,7168,8,256,64,balanced,0.3398463994264603
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,6144,7168,8,256,64,balanced,0.09549439996480942
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,6144,7168,8,256,64,balanced,0.3697119981050491
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,6144,7168,8,256,64,balanced,0.548118394613266
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,6144,7168,8,256,64,balanced,0.09721919894218445
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,6144,7168,8,256,64,balanced,0.2106911987066269
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,6144,7168,8,256,64,balanced,0.5473920047283173
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,6144,7168,8,256,64,balanced,0.09697600081562996
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,6144,7168,8,256,64,balanced,0.15581759959459304
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,6144,7168,8,256,64,balanced,0.5491968035697937
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,6144,7168,8,256,64,balanced,0.47659520208835604
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,8192,7168,8,256,64,balanced,0.11079039946198463
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,8192,7168,8,256,64,balanced,0.4809056013822556
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,8192,7168,8,256,64,balanced,0.7216159880161286
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,8192,7168,8,256,64,balanced,0.10751679912209511
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,8192,7168,8,256,64,balanced,0.26813119649887085
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,8192,7168,8,256,64,balanced,0.7262911975383759
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,8192,7168,8,256,64,balanced,0.10609599947929382
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,8192,7168,8,256,64,balanced,0.18905919939279556
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,8192,7168,8,256,64,balanced,0.7214175939559937
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,8192,7168,8,256,64,balanced,0.6143679916858673
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,12288,7168,8,256,64,balanced,0.1318880006670952
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,12288,7168,8,256,64,balanced,0.7179039895534516
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,12288,7168,8,256,64,balanced,1.0759008169174193
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,12288,7168,8,256,64,balanced,0.13351679891347884
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,12288,7168,8,256,64,balanced,0.3783199995756149
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,12288,7168,8,256,64,balanced,1.077894401550293
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,12288,7168,8,256,64,balanced,0.13117119818925857
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,12288,7168,8,256,64,balanced,0.2596640005707741
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,12288,7168,8,256,64,balanced,1.075164806842804
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,12288,7168,8,256,64,balanced,0.891951996088028
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,16384,7168,8,256,64,balanced,0.22636799961328508
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,16384,7168,8,256,64,balanced,0.9630944013595581
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,16384,7168,8,256,64,balanced,1.4395680189132691
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,16384,7168,8,256,64,balanced,0.15637760013341903
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,16384,7168,8,256,64,balanced,0.4955487996339798
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,16384,7168,8,256,64,balanced,1.4340768098831176
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,16384,7168,8,256,64,balanced,0.15562240034341812
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,16384,7168,8,256,64,balanced,0.3278112024068832
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,16384,7168,8,256,64,balanced,1.4364895939826965
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,16384,7168,8,256,64,balanced,1.1853983998298645
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,20480,7168,8,256,64,balanced,0.1797439992427826
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,20480,7168,8,256,64,balanced,1.1995839953422547
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,20480,7168,8,256,64,balanced,1.7963200092315674
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,20480,7168,8,256,64,balanced,0.18178240060806275
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,20480,7168,8,256,64,balanced,0.6275071978569031
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,20480,7168,8,256,64,balanced,1.8003743886947632
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,20480,7168,8,256,64,balanced,0.18190400004386903
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,20480,7168,8,256,64,balanced,0.39454720318317416
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,20480,7168,8,256,64,balanced,1.8011168003082276
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,20480,7168,8,256,64,balanced,1.4776831984519958
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,32768,7168,8,256,64,balanced,0.25633599907159804
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,32768,7168,8,256,64,balanced,1.9314688086509704
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,32768,7168,8,256,64,balanced,2.8690752267837523
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,32768,7168,8,256,64,balanced,0.2528063997626305
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,32768,7168,8,256,64,balanced,0.9977344036102295
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,32768,7168,8,256,64,balanced,2.8761055707931518
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,32768,7168,8,256,64,balanced,0.25208639949560163
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,32768,7168,8,256,64,balanced,0.5994368016719818
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,32768,7168,8,256,64,balanced,2.873699188232422
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,32768,7168,8,256,64,balanced,2.3058112144470213
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,65536,7168,8,256,64,balanced,0.452115198969841
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,65536,7168,8,256,64,balanced,3.8675327777862547
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,65536,7168,8,256,64,balanced,5.769625568389893
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,65536,7168,8,256,64,balanced,0.4537696003913879
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,65536,7168,8,256,64,balanced,2.0042912006378173
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,65536,7168,8,256,64,balanced,5.76692795753479
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,65536,7168,8,256,64,balanced,0.4490336000919342
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,65536,7168,8,256,64,balanced,1.1365056037902832
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,65536,7168,8,256,64,balanced,5.777731227874756
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,65536,7168,8,256,64,balanced,4.41145281791687
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1,7168,8,256,4,balanced,0.06011519990861416
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1,7168,8,256,4,balanced,0.050063999369740486
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1,7168,8,256,4,balanced,0.07439680024981499
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1,7168,8,256,4,balanced,0.07881279923021793
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1,7168,8,256,4,balanced,0.04588479958474636
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1,7168,8,256,4,balanced,0.07285119965672493
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1,7168,8,256,4,balanced,0.05745600052177906
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1,7168,8,256,4,balanced,0.05479360073804855
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1,7168,8,256,4,balanced,0.08026240095496177
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1,7168,8,256,4,balanced,0.07418880015611648
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,2,7168,8,256,4,balanced,0.058742400258779526
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,2,7168,8,256,4,balanced,0.0482783991843462
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,2,7168,8,256,4,balanced,0.07819200083613395
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,2,7168,8,256,4,balanced,0.0566943995654583
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,2,7168,8,256,4,balanced,0.04649280048906803
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,2,7168,8,256,4,balanced,0.07593920007348061
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,2,7168,8,256,4,balanced,0.05711039938032627
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,2,7168,8,256,4,balanced,0.04890559986233711
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,2,7168,8,256,4,balanced,0.07276160046458244
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,2,7168,8,256,4,balanced,0.06883200109004975
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,4,7168,8,256,4,balanced,0.05599039979279041
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,4,7168,8,256,4,balanced,0.04764159992337227
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,4,7168,8,256,4,balanced,0.06770879924297332
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,4,7168,8,256,4,balanced,0.05353279933333397
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,4,7168,8,256,4,balanced,0.04678080044686794
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,4,7168,8,256,4,balanced,0.0677920013666153
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,4,7168,8,256,4,balanced,0.057062399387359616
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,4,7168,8,256,4,balanced,0.048444800078868866
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,4,7168,8,256,4,balanced,0.07311040088534355
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,4,7168,8,256,4,balanced,0.07234239876270294
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,8,7168,8,256,4,balanced,0.052422399446368215
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,8,7168,8,256,4,balanced,0.046521599590778354
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,8,7168,8,256,4,balanced,0.07095679976046085
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,8,7168,8,256,4,balanced,0.054518400132656096
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,8,7168,8,256,4,balanced,0.04629119969904423
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,8,7168,8,256,4,balanced,0.06799679957330226
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,8,7168,8,256,4,balanced,0.054896000400185586
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,8,7168,8,256,4,balanced,0.04903039969503879
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,8,7168,8,256,4,balanced,0.06804799996316432
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,8,7168,8,256,4,balanced,0.07003519982099533
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,16,7168,8,256,4,balanced,0.05925439894199371
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,16,7168,8,256,4,balanced,0.047673600167036055
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,16,7168,8,256,4,balanced,0.07553279995918274
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,16,7168,8,256,4,balanced,0.057702400162816046
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,16,7168,8,256,4,balanced,0.0439008004963398
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,16,7168,8,256,4,balanced,0.07156159952282906
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,16,7168,8,256,4,balanced,0.051782400161027906
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,16,7168,8,256,4,balanced,0.047523199766874316
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,16,7168,8,256,4,balanced,0.06743040010333061
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,16,7168,8,256,4,balanced,0.06883839927613736
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,32,7168,8,256,4,balanced,0.05495999976992607
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,32,7168,8,256,4,balanced,0.04685439988970756
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,32,7168,8,256,4,balanced,0.06638720110058785
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,32,7168,8,256,4,balanced,0.05292480029165745
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,32,7168,8,256,4,balanced,0.051276800036430356
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,32,7168,8,256,4,balanced,0.06723200082778931
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,32,7168,8,256,4,balanced,0.05216000005602837
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,32,7168,8,256,4,balanced,0.04485120065510273
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,32,7168,8,256,4,balanced,0.06430080011487008
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,32,7168,8,256,4,balanced,0.06835200041532516
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,48,7168,8,256,4,balanced,0.05686080008745194
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,48,7168,8,256,4,balanced,0.042643199861049655
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,48,7168,8,256,4,balanced,0.06570239998400211
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,48,7168,8,256,4,balanced,0.05509439930319786
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,48,7168,8,256,4,balanced,0.04190080054104328
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,48,7168,8,256,4,balanced,0.06511039994657039
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,48,7168,8,256,4,balanced,0.056547199934720994
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,48,7168,8,256,4,balanced,0.04608000032603741
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,48,7168,8,256,4,balanced,0.06673920080065728
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,48,7168,8,256,4,balanced,0.06815039999783039
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,64,7168,8,256,4,balanced,0.0530239999294281
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,64,7168,8,256,4,balanced,0.04493119977414608
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,64,7168,8,256,4,balanced,0.06435199901461601
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,64,7168,8,256,4,balanced,0.053331200033426285
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,64,7168,8,256,4,balanced,0.04402240067720413
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,64,7168,8,256,4,balanced,0.06749439910054207
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,64,7168,8,256,4,balanced,0.05307199992239475
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,64,7168,8,256,4,balanced,0.04532799981534481
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,64,7168,8,256,4,balanced,0.06486399918794632
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,64,7168,8,256,4,balanced,0.06855679973959923
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,80,7168,8,256,4,balanced,0.05093759968876839
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,80,7168,8,256,4,balanced,0.04382079988718033
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,80,7168,8,256,4,balanced,0.06778240129351616
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,80,7168,8,256,4,balanced,0.05108160004019737
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,80,7168,8,256,4,balanced,0.04169279970228672
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,80,7168,8,256,4,balanced,0.06631359979510307
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,80,7168,8,256,4,balanced,0.05487999953329563
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,80,7168,8,256,4,balanced,0.04421439990401268
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,80,7168,8,256,4,balanced,0.06615039929747582
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,80,7168,8,256,4,balanced,0.06621759980916977
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,96,7168,8,256,4,balanced,0.05000960007309914
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,96,7168,8,256,4,balanced,0.043961599841713904
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,96,7168,8,256,4,balanced,0.06671360023319721
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,96,7168,8,256,4,balanced,0.051660800352692604
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,96,7168,8,256,4,balanced,0.04135359972715378
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,96,7168,8,256,4,balanced,0.0659680001437664
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,96,7168,8,256,4,balanced,0.05124800018966198
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,96,7168,8,256,4,balanced,0.04446719996631145
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,96,7168,8,256,4,balanced,0.0643839992582798
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,96,7168,8,256,4,balanced,0.06438400000333785
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,128,7168,8,256,4,balanced,0.05074560008943081
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,128,7168,8,256,4,balanced,0.04317119941115379
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,128,7168,8,256,4,balanced,0.06556159928441048
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,128,7168,8,256,4,balanced,0.051903999969363214
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,128,7168,8,256,4,balanced,0.04163520000874996
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,128,7168,8,256,4,balanced,0.0661504004150629
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,128,7168,8,256,4,balanced,0.050944000482559204
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,128,7168,8,256,4,balanced,0.04623360000550747
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,128,7168,8,256,4,balanced,0.06823360063135624
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,128,7168,8,256,4,balanced,0.06454720050096512
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,160,7168,8,256,4,balanced,0.05420800000429153
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,160,7168,8,256,4,balanced,0.04492800012230873
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,160,7168,8,256,4,balanced,0.0659616008400917
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,160,7168,8,256,4,balanced,0.0536991998553276
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,160,7168,8,256,4,balanced,0.04219520054757595
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,160,7168,8,256,4,balanced,0.0670623991638422
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,160,7168,8,256,4,balanced,0.04897280000150204
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,160,7168,8,256,4,balanced,0.04687360040843487
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,160,7168,8,256,4,balanced,0.06800960004329681
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,160,7168,8,256,4,balanced,0.06977600045502186
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,192,7168,8,256,4,balanced,0.05099200047552586
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,192,7168,8,256,4,balanced,0.043209600076079366
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,192,7168,8,256,4,balanced,0.06895360052585602
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,192,7168,8,256,4,balanced,0.054252799972891806
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,192,7168,8,256,4,balanced,0.04343359991908073
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,192,7168,8,256,4,balanced,0.06889600083231925
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,192,7168,8,256,4,balanced,0.050630399584770204
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,192,7168,8,256,4,balanced,0.04631040021777153
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,192,7168,8,256,4,balanced,0.06655999943614006
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,192,7168,8,256,4,balanced,0.06618560031056404
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,256,7168,8,256,4,balanced,0.051638399437069896
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,256,7168,8,256,4,balanced,0.04352319985628128
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,256,7168,8,256,4,balanced,0.06703360080718994
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,256,7168,8,256,4,balanced,0.05283519998192787
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,256,7168,8,256,4,balanced,0.04170240052044392
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,256,7168,8,256,4,balanced,0.06661440022289752
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,256,7168,8,256,4,balanced,0.050832000374794004
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,256,7168,8,256,4,balanced,0.044012800231575964
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,256,7168,8,256,4,balanced,0.06564799956977367
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,256,7168,8,256,4,balanced,0.06831040009856224
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,320,7168,8,256,4,balanced,0.05189119949936867
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,320,7168,8,256,4,balanced,0.04505920000374317
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,320,7168,8,256,4,balanced,0.06718080006539821
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,320,7168,8,256,4,balanced,0.052102400362491606
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,320,7168,8,256,4,balanced,0.043091199919581415
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,320,7168,8,256,4,balanced,0.06829760000109672
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,320,7168,8,256,4,balanced,0.05562239922583103
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,320,7168,8,256,4,balanced,0.044553599879145624
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,320,7168,8,256,4,balanced,0.06070400029420853
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,320,7168,8,256,4,balanced,0.06490880027413368
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,384,7168,8,256,4,balanced,0.05105919949710369
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,384,7168,8,256,4,balanced,0.044838400185108186
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,384,7168,8,256,4,balanced,0.06623679995536805
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,384,7168,8,256,4,balanced,0.05194879993796349
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,384,7168,8,256,4,balanced,0.04389119930565357
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,384,7168,8,256,4,balanced,0.06386879943311215
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,384,7168,8,256,4,balanced,0.048470400646328925
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,384,7168,8,256,4,balanced,0.04506879970431328
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,384,7168,8,256,4,balanced,0.06434560045599938
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,384,7168,8,256,4,balanced,0.06669440008699894
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,512,7168,8,256,4,balanced,0.05276160016655922
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,512,7168,8,256,4,balanced,0.04668480008840561
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,512,7168,8,256,4,balanced,0.0702463999390602
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,512,7168,8,256,4,balanced,0.050144000351428984
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,512,7168,8,256,4,balanced,0.0406272005289793
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,512,7168,8,256,4,balanced,0.06573120020329952
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,512,7168,8,256,4,balanced,0.05250240117311478
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,512,7168,8,256,4,balanced,0.045087999850511554
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,512,7168,8,256,4,balanced,0.0672640010714531
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,512,7168,8,256,4,balanced,0.07344959974288941
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,768,7168,8,256,4,balanced,0.053619199246168134
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,768,7168,8,256,4,balanced,0.0512959998100996
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,768,7168,8,256,4,balanced,0.07952639982104301
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,768,7168,8,256,4,balanced,0.04945919960737229
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,768,7168,8,256,4,balanced,0.042703999951481816
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,768,7168,8,256,4,balanced,0.08077759817242622
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,768,7168,8,256,4,balanced,0.050387199595570566
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,768,7168,8,256,4,balanced,0.045017600432038306
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,768,7168,8,256,4,balanced,0.0786816008388996
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,768,7168,8,256,4,balanced,0.09167679920792579
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1024,7168,8,256,4,balanced,0.05239039994776249
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1024,7168,8,256,4,balanced,0.05716799981892109
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1024,7168,8,256,4,balanced,0.09208640083670616
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1024,7168,8,256,4,balanced,0.051446400210261346
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1024,7168,8,256,4,balanced,0.045151999592781066
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1024,7168,8,256,4,balanced,0.09133120104670525
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1024,7168,8,256,4,balanced,0.0505791999399662
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1024,7168,8,256,4,balanced,0.04525440037250519
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1024,7168,8,256,4,balanced,0.0939071998000145
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1024,7168,8,256,4,balanced,0.1037823997437954
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1536,7168,8,256,4,balanced,0.05463680066168308
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1536,7168,8,256,4,balanced,0.06696959882974625
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1536,7168,8,256,4,balanced,0.12138240039348602
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1536,7168,8,256,4,balanced,0.0547712005674839
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1536,7168,8,256,4,balanced,0.05107520036399364
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1536,7168,8,256,4,balanced,0.11780799999833107
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1536,7168,8,256,4,balanced,0.05344960018992424
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1536,7168,8,256,4,balanced,0.05297280065715313
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1536,7168,8,256,4,balanced,0.11920320019125938
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1536,7168,8,256,4,balanced,0.13530239909887315
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,2048,7168,8,256,4,balanced,0.053254399448633194
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,2048,7168,8,256,4,balanced,0.07813440039753913
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,2048,7168,8,256,4,balanced,0.14556800127029418
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,2048,7168,8,256,4,balanced,0.05486719980835915
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,2048,7168,8,256,4,balanced,0.05610559955239296
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,2048,7168,8,256,4,balanced,0.1424800008535385
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1,7168,8,256,8,balanced,0.06721919998526574
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1,7168,8,256,8,balanced,0.05628479979932308
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1,7168,8,256,8,balanced,0.07381120100617408
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1,7168,8,256,8,balanced,0.06251840069890022
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1,7168,8,256,8,balanced,0.05039360001683235
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1,7168,8,256,8,balanced,0.06705279983580112
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1,7168,8,256,8,balanced,0.06251199916005135
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1,7168,8,256,8,balanced,0.0709087997674942
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1,7168,8,256,8,balanced,0.07328960001468658
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1,7168,8,256,8,balanced,0.10025919899344445
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,2048,7168,8,256,4,balanced,0.053359999880194665
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,2048,7168,8,256,4,balanced,0.05682240054011345
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,2,7168,8,256,8,balanced,0.06842240095138549
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,2048,7168,8,256,4,balanced,0.14444479942321778
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,2,7168,8,256,8,balanced,0.05691520050168038
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,2048,7168,8,256,4,balanced,0.16667519956827165
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,2,7168,8,256,8,balanced,0.07560960054397584
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,2,7168,8,256,8,balanced,0.06912000104784966
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,2,7168,8,256,8,balanced,0.05728319957852364
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,2,7168,8,256,8,balanced,0.0700096007436514
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,2,7168,8,256,8,balanced,0.06359040066599846
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,2,7168,8,256,8,balanced,0.05609600059688091
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,2,7168,8,256,8,balanced,0.0727743998169899
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,2,7168,8,256,8,balanced,0.07660800032317638
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,4,7168,8,256,8,balanced,0.06589439883828163
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,4,7168,8,256,8,balanced,0.05620800070464611
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,4,7168,8,256,8,balanced,0.07120640054345131
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,4,7168,8,256,8,balanced,0.0696735993027687
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,4,7168,8,256,8,balanced,0.05949440039694309
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,4,7168,8,256,8,balanced,0.06517120078206062
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,4,7168,8,256,8,balanced,0.07121280059218407
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,4,7168,8,256,8,balanced,0.05871039927005768
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,4,7168,8,256,8,balanced,0.06575040034949779
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,4,7168,8,256,8,balanced,0.07888640016317368
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,8,7168,8,256,8,balanced,0.06618879996240139
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,8,7168,8,256,8,balanced,0.0591839998960495
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,8,7168,8,256,8,balanced,0.09350719824433326
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,8,7168,8,256,8,balanced,0.07400640025734902
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,8,7168,8,256,8,balanced,0.05804160051047802
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,8,7168,8,256,8,balanced,0.09355520009994507
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,8,7168,8,256,8,balanced,0.07224320061504841
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,8,7168,8,256,8,balanced,0.057145599648356436
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,8,7168,8,256,8,balanced,0.0946528010070324
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,8,7168,8,256,8,balanced,0.08976000100374222
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,16,7168,8,256,8,balanced,0.06939519941806793
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,16,7168,8,256,8,balanced,0.06038399934768677
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,16,7168,8,256,8,balanced,0.09848320111632347
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,16,7168,8,256,8,balanced,0.06499840021133423
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,16,7168,8,256,8,balanced,0.05696960054337978
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,16,7168,8,256,8,balanced,0.09517440050840378
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,16,7168,8,256,8,balanced,0.0672704003751278
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,16,7168,8,256,8,balanced,0.06024000011384487
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,16,7168,8,256,8,balanced,0.10106880068778992
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,16,7168,8,256,8,balanced,0.09587199985980988
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,32,7168,8,256,8,balanced,0.06556159928441048
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,32,7168,8,256,8,balanced,0.05846400074660778
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,32,7168,8,256,8,balanced,0.10002880021929741
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,32,7168,8,256,8,balanced,0.06222079955041408
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,32,7168,8,256,8,balanced,0.05610879994928837
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,32,7168,8,256,8,balanced,0.09621120095252991
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,32,7168,8,256,8,balanced,0.05786879993975162
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,32,7168,8,256,8,balanced,0.058099199831485746
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,32,7168,8,256,8,balanced,0.1025056004524231
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,32,7168,8,256,8,balanced,0.09734400063753128
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,48,7168,8,256,8,balanced,0.07132480032742024
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,48,7168,8,256,8,balanced,0.05734720043838024
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,48,7168,8,256,8,balanced,0.10220160111784934
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,48,7168,8,256,8,balanced,0.061792000010609624
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,48,7168,8,256,8,balanced,0.05819519981741905
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,48,7168,8,256,8,balanced,0.1011039987206459
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,48,7168,8,256,8,balanced,0.056854399666190145
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,48,7168,8,256,8,balanced,0.05677120052278042
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,48,7168,8,256,8,balanced,0.0949632003903389
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,48,7168,8,256,8,balanced,0.0840735986828804
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,64,7168,8,256,8,balanced,0.0637472003698349
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,64,7168,8,256,8,balanced,0.05641600005328655
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,64,7168,8,256,8,balanced,0.09933760017156601
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,64,7168,8,256,8,balanced,0.06552640050649643
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,64,7168,8,256,8,balanced,0.055264000222086905
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,64,7168,8,256,8,balanced,0.09112000018358231
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,64,7168,8,256,8,balanced,0.06344319991767407
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,64,7168,8,256,8,balanced,0.06183680035173893
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,64,7168,8,256,8,balanced,0.09793919995427132
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,64,7168,8,256,8,balanced,0.09329920038580894
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,80,7168,8,256,8,balanced,0.06450879946351051
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,80,7168,8,256,8,balanced,0.05804799981415272
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,80,7168,8,256,8,balanced,0.10246719866991043
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1,7168,8,256,2,balanced,0.0608704000711441
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1,7168,8,256,2,balanced,0.054851199686527255
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1,7168,8,256,2,balanced,0.07798720076680184
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,80,7168,8,256,8,balanced,0.0651232000440359
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,80,7168,8,256,8,balanced,0.06069759875535965
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,80,7168,8,256,8,balanced,0.09514240026474
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1,7168,8,256,2,balanced,0.060681600496172904
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1,7168,8,256,2,balanced,0.05068800002336502
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1,7168,8,256,2,balanced,0.08329920023679734
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1,7168,8,256,2,balanced,0.06390080004930496
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1,7168,8,256,2,balanced,0.05981439985334873
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1,7168,8,256,2,balanced,0.07783359959721566
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1,7168,8,256,2,balanced,0.08386879935860633
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,2,7168,8,256,2,balanced,0.06462399996817111
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,2,7168,8,256,2,balanced,0.052745600789785387
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,2,7168,8,256,2,balanced,0.07803840041160584
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,2,7168,8,256,2,balanced,0.06007999964058399
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,3072,7168,8,256,4,balanced,0.06111040003597736
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,3072,7168,8,256,4,balanced,0.09669759944081306
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,3072,7168,8,256,4,balanced,0.19390079975128174
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,3072,7168,8,256,4,balanced,0.05957439988851547
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,3072,7168,8,256,4,balanced,0.06788800060749053
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,3072,7168,8,256,4,balanced,0.19765760004520416
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,3072,7168,8,256,4,balanced,0.05875199921429157
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,3072,7168,8,256,4,balanced,0.06824959889054298
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,3072,7168,8,256,4,balanced,0.19316799938678741
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,3072,7168,8,256,4,balanced,0.2223711982369423
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,2,7168,8,256,2,balanced,0.049347199872136115
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,2,7168,8,256,2,balanced,0.0829984001815319
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,80,7168,8,256,8,balanced,0.06114559955894947
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,80,7168,8,256,8,balanced,0.057744001224637034
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,2,7168,8,256,2,balanced,0.07246079966425896
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,2,7168,8,256,2,balanced,0.06379200108349323
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,2,7168,8,256,2,balanced,0.08440639972686767
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,2,7168,8,256,2,balanced,0.0956287994980812
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,4,7168,8,256,2,balanced,0.07051519863307476
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,4,7168,8,256,2,balanced,0.05685760006308556
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,4,7168,8,256,2,balanced,0.08360960036516189
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,4,7168,8,256,2,balanced,0.07128320038318633
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,4,7168,8,256,2,balanced,0.0561247993260622
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,4,7168,8,256,2,balanced,0.0917055994272232
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,4,7168,8,256,2,balanced,0.06586559973657131
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,4,7168,8,256,2,balanced,0.0594591997563839
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,4,7168,8,256,2,balanced,0.09051200076937675
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,4,7168,8,256,2,balanced,0.08733120113611222
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,8,7168,8,256,2,balanced,0.06609919928014278
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,8,7168,8,256,2,balanced,0.054700800031423566
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,8,7168,8,256,2,balanced,0.08660160005092621
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,8,7168,8,256,2,balanced,0.068886399269104
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,8,7168,8,256,2,balanced,0.05405759960412979
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,8,7168,8,256,2,balanced,0.08501759991049766
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,8,7168,8,256,2,balanced,0.06577599942684173
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,8,7168,8,256,2,balanced,0.05599039979279041
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,8,7168,8,256,2,balanced,0.08387519940733909
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,8,7168,8,256,2,balanced,0.0845440000295639
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,16,7168,8,256,2,balanced,0.058185600116848944
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,16,7168,8,256,2,balanced,0.05355519950389862
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,16,7168,8,256,2,balanced,0.08383679986000062
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,16,7168,8,256,2,balanced,0.06410240083932876
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,16,7168,8,256,2,balanced,0.051782400161027906
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,16,7168,8,256,2,balanced,0.08209600001573562
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,16,7168,8,256,2,balanced,0.06426560021936893
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,16,7168,8,256,2,balanced,0.05342399999499321
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,16,7168,8,256,2,balanced,0.08855359926819802
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,16,7168,8,256,2,balanced,0.0830047994852066
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,32,7168,8,256,2,balanced,0.062441599741578105
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,32,7168,8,256,2,balanced,0.05772799998521805
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,32,7168,8,256,2,balanced,0.08261759951710701
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,32,7168,8,256,2,balanced,0.06058880016207695
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,32,7168,8,256,2,balanced,0.05373760014772415
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,32,7168,8,256,2,balanced,0.0796032004058361
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,32,7168,8,256,2,balanced,0.05873280055820942
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,32,7168,8,256,2,balanced,0.05373759940266609
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,32,7168,8,256,2,balanced,0.07762239947915077
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,32,7168,8,256,2,balanced,0.07742080017924309
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,48,7168,8,256,2,balanced,0.06207679994404316
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,48,7168,8,256,2,balanced,0.049667200073599815
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,48,7168,8,256,2,balanced,0.08858559876680375
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,48,7168,8,256,2,balanced,0.06384640000760555
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,48,7168,8,256,2,balanced,0.050102400779724124
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,48,7168,8,256,2,balanced,0.08221440017223358
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,48,7168,8,256,2,balanced,0.059996800124645235
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,48,7168,8,256,2,balanced,0.051875199750065805
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,48,7168,8,256,2,balanced,0.08008640110492707
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,48,7168,8,256,2,balanced,0.07988160029053688
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,64,7168,8,256,2,balanced,0.0626624010503292
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,64,7168,8,256,2,balanced,0.04796480014920235
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,64,7168,8,256,2,balanced,0.07687360048294067
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,64,7168,8,256,2,balanced,0.06178240031003952
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,64,7168,8,256,2,balanced,0.04739199988543987
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,64,7168,8,256,2,balanced,0.08151040002703666
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,64,7168,8,256,2,balanced,0.05904639959335327
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,64,7168,8,256,2,balanced,0.050832000374794004
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,64,7168,8,256,2,balanced,0.07517440021038055
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,64,7168,8,256,2,balanced,0.07651519998908043
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,4096,7168,8,256,4,balanced,0.06573440060019493
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,4096,7168,8,256,4,balanced,0.11863679960370063
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,4096,7168,8,256,4,balanced,0.24450880140066147
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,4096,7168,8,256,4,balanced,0.06467840112745762
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,4096,7168,8,256,4,balanced,0.07885119989514351
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,4096,7168,8,256,4,balanced,0.24336320012807847
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,80,7168,8,256,8,balanced,0.09503680020570755
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,80,7168,8,256,8,balanced,0.09092479944229126
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,96,7168,8,256,8,balanced,0.06496639996767044
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,96,7168,8,256,8,balanced,0.060435198992490766
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,96,7168,8,256,8,balanced,0.09862079992890357
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,96,7168,8,256,8,balanced,0.06721279993653298
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,96,7168,8,256,8,balanced,0.06162239946424961
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,96,7168,8,256,8,balanced,0.0930527999997139
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,96,7168,8,256,8,balanced,0.061011200025677684
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,96,7168,8,256,8,balanced,0.05899519994854927
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,96,7168,8,256,8,balanced,0.08924480006098748
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,96,7168,8,256,8,balanced,0.09321920052170754
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,128,7168,8,256,8,balanced,0.06312960013747215
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,128,7168,8,256,8,balanced,0.05795519985258579
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,128,7168,8,256,8,balanced,0.10150719955563545
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,128,7168,8,256,8,balanced,0.060326400026679036
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,128,7168,8,256,8,balanced,0.055532800406217574
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,128,7168,8,256,8,balanced,0.09497600048780441
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,128,7168,8,256,8,balanced,0.059676801040768626
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,128,7168,8,256,8,balanced,0.05562880039215088
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,128,7168,8,256,8,balanced,0.0915743999183178
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,128,7168,8,256,8,balanced,0.09607679918408393
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,160,7168,8,256,8,balanced,0.06551040001213551
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,160,7168,8,256,8,balanced,0.055635199695825574
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,160,7168,8,256,8,balanced,0.09801600053906441
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,160,7168,8,256,8,balanced,0.06256639920175075
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,160,7168,8,256,8,balanced,0.055350400134921075
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,160,7168,8,256,8,balanced,0.08887040019035339
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,160,7168,8,256,8,balanced,0.06302719935774803
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,160,7168,8,256,8,balanced,0.07138880118727683
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,160,7168,8,256,8,balanced,0.09562239944934844
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,160,7168,8,256,8,balanced,0.09568639919161796
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,192,7168,8,256,8,balanced,0.06570560075342655
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,192,7168,8,256,8,balanced,0.06590079925954342
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,192,7168,8,256,8,balanced,0.09716799855232239
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,80,7168,8,256,2,balanced,0.06045759990811348
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,80,7168,8,256,2,balanced,0.04903039969503879
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,80,7168,8,256,2,balanced,0.07260480001568795
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,80,7168,8,256,2,balanced,0.060368000715970996
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,192,7168,8,256,8,balanced,0.06409600004553795
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,192,7168,8,256,8,balanced,0.05601920001208782
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,192,7168,8,256,8,balanced,0.09670720025897026
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,4096,7168,8,256,4,balanced,0.06551360003650189
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,4096,7168,8,256,4,balanced,0.07627200037240982
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,4096,7168,8,256,4,balanced,0.24410560131072997
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,4096,7168,8,256,4,balanced,0.2790719985961914
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,192,7168,8,256,8,balanced,0.061510400474071504
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,192,7168,8,256,8,balanced,0.05663359984755516
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,192,7168,8,256,8,balanced,0.09630079865455628
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,192,7168,8,256,8,balanced,0.09100480079650879
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,256,7168,8,256,8,balanced,0.06278080008924007
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,256,7168,8,256,8,balanced,0.060745599865913394
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,256,7168,8,256,8,balanced,0.09101440012454987
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,256,7168,8,256,8,balanced,0.0632352001965046
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,256,7168,8,256,8,balanced,0.05319680012762547
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,256,7168,8,256,8,balanced,0.0887327991425991
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,256,7168,8,256,8,balanced,0.06297600083053112
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,256,7168,8,256,8,balanced,0.0573247991502285
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,256,7168,8,256,8,balanced,0.0977792002260685
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,256,7168,8,256,8,balanced,0.08944640085101127
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,320,7168,8,256,8,balanced,0.06016319952905178
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,320,7168,8,256,8,balanced,0.058819199725985524
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,320,7168,8,256,8,balanced,0.09130560085177422
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,320,7168,8,256,8,balanced,0.0641024000942707
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,320,7168,8,256,8,balanced,0.05421759970486164
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,320,7168,8,256,8,balanced,0.09632319957017899
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,320,7168,8,256,8,balanced,0.06247039921581745
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,320,7168,8,256,8,balanced,0.06082880049943924
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,320,7168,8,256,8,balanced,0.08037759959697724
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,320,7168,8,256,8,balanced,0.09225280061364174
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,384,7168,8,256,8,balanced,0.06348799988627433
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,384,7168,8,256,8,balanced,0.06154239997267723
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,384,7168,8,256,8,balanced,0.10041279941797257
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,384,7168,8,256,8,balanced,0.060355200245976445
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,384,7168,8,256,8,balanced,0.05851200073957443
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,384,7168,8,256,8,balanced,0.0920575998723507
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,384,7168,8,256,8,balanced,0.06479040011763573
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,384,7168,8,256,8,balanced,0.05828160010278225
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,384,7168,8,256,8,balanced,0.09618880078196526
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,384,7168,8,256,8,balanced,0.09105280041694641
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,512,7168,8,256,8,balanced,0.06266240030527115
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,512,7168,8,256,8,balanced,0.06395199969410896
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,512,7168,8,256,8,balanced,0.10109439939260483
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,512,7168,8,256,8,balanced,0.06498240083456039
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,512,7168,8,256,8,balanced,0.05572800040245056
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,512,7168,8,256,8,balanced,0.09369279891252517
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,512,7168,8,256,8,balanced,0.064131198823452
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,512,7168,8,256,8,balanced,0.059158400446176526
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,512,7168,8,256,8,balanced,0.09947520196437835
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,512,7168,8,256,8,balanced,0.09182719886302948
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,6144,7168,8,256,4,balanced,0.07824319899082184
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,6144,7168,8,256,4,balanced,0.16206399947404862
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,6144,7168,8,256,4,balanced,0.34714879989624026
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,768,7168,8,256,8,balanced,0.06248959936201572
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,768,7168,8,256,8,balanced,0.0655456006526947
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,768,7168,8,256,8,balanced,0.09489920064806938
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,768,7168,8,256,8,balanced,0.06545280106365681
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,768,7168,8,256,8,balanced,0.05932799987494945
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,768,7168,8,256,8,balanced,0.10088320076465607
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,80,7168,8,256,2,balanced,0.046662400290369985
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,80,7168,8,256,2,balanced,0.07590079978108406
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,80,7168,8,256,2,balanced,0.0644863996654749
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,80,7168,8,256,2,balanced,0.05373760014772415
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,80,7168,8,256,2,balanced,0.07783680036664009
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,80,7168,8,256,2,balanced,0.0800416000187397
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,768,7168,8,256,8,balanced,0.06814080066978931
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,96,7168,8,256,2,balanced,0.060118400305509564
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,96,7168,8,256,2,balanced,0.04824319928884506
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,96,7168,8,256,2,balanced,0.07682239934802056
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,96,7168,8,256,2,balanced,0.05818240083754063
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,96,7168,8,256,2,balanced,0.04714560061693192
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,96,7168,8,256,2,balanced,0.0747808001935482
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,96,7168,8,256,2,balanced,0.06069119907915592
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,96,7168,8,256,2,balanced,0.050918400287628174
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,96,7168,8,256,2,balanced,0.07404799908399581
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,96,7168,8,256,2,balanced,0.07733760103583336
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,128,7168,8,256,2,balanced,0.05646400041878223
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,128,7168,8,256,2,balanced,0.04885440021753311
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,128,7168,8,256,2,balanced,0.07615040019154548
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,128,7168,8,256,2,balanced,0.06017599925398827
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,128,7168,8,256,2,balanced,0.046934400126338004
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,128,7168,8,256,2,balanced,0.07483839988708496
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,128,7168,8,256,2,balanced,0.0568448007106781
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,128,7168,8,256,2,balanced,0.05037439987063408
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,128,7168,8,256,2,balanced,0.07176319882273674
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,128,7168,8,256,2,balanced,0.0782431997358799
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,160,7168,8,256,2,balanced,0.05803840011358261
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,160,7168,8,256,2,balanced,0.046751999855041505
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,160,7168,8,256,2,balanced,0.07657919973134994
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,160,7168,8,256,2,balanced,0.05539840012788773
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,160,7168,8,256,2,balanced,0.047017600387334824
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,160,7168,8,256,2,balanced,0.07826559916138649
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,160,7168,8,256,2,balanced,0.055616000294685365
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,160,7168,8,256,2,balanced,0.049113599210977556
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,160,7168,8,256,2,balanced,0.07457920089364052
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,160,7168,8,256,2,balanced,0.0767999991774559
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,192,7168,8,256,2,balanced,0.058870400115847585
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,192,7168,8,256,2,balanced,0.049011199921369555
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,192,7168,8,256,2,balanced,0.07860160022974014
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,192,7168,8,256,2,balanced,0.05938239991664886
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,192,7168,8,256,2,balanced,0.04776640012860298
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,192,7168,8,256,2,balanced,0.07078079953789711
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,192,7168,8,256,2,balanced,0.05432959981262684
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,192,7168,8,256,2,balanced,0.05006719976663589
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,192,7168,8,256,2,balanced,0.08207679986953735
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,192,7168,8,256,2,balanced,0.0704479992389679
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,256,7168,8,256,2,balanced,0.05556800030171871
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,256,7168,8,256,2,balanced,0.046783999726176265
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,256,7168,8,256,2,balanced,0.07823359966278076
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,256,7168,8,256,2,balanced,0.05754879973828793
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,256,7168,8,256,2,balanced,0.04806080050766468
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,256,7168,8,256,2,balanced,0.08049280047416688
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,256,7168,8,256,2,balanced,0.05910399965941906
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,256,7168,8,256,2,balanced,0.050918399915099145
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,256,7168,8,256,2,balanced,0.07774720042943954
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,256,7168,8,256,2,balanced,0.07583040073513984
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,6144,7168,8,256,4,balanced,0.07997119948267936
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,6144,7168,8,256,4,balanced,0.10606399923563004
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,6144,7168,8,256,4,balanced,0.3456768006086349
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,320,7168,8,256,2,balanced,0.056515200063586235
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,320,7168,8,256,2,balanced,0.04601600021123886
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,320,7168,8,256,2,balanced,0.07761920094490052
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,320,7168,8,256,2,balanced,0.05470719933509827
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,320,7168,8,256,2,balanced,0.04677439965307713
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,320,7168,8,256,2,balanced,0.0710719995200634
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,320,7168,8,256,2,balanced,0.05804159939289093
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,320,7168,8,256,2,balanced,0.05066239982843399
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,320,7168,8,256,2,balanced,0.07631680071353912
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,320,7168,8,256,2,balanced,0.07537920027971268
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,384,7168,8,256,2,balanced,0.06016639992594719
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,384,7168,8,256,2,balanced,0.04885439947247505
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,384,7168,8,256,2,balanced,0.07954560071229935
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,384,7168,8,256,2,balanced,0.0582495991140604
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,384,7168,8,256,2,balanced,0.048121600225567815
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,384,7168,8,256,2,balanced,0.07855360060930253
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,384,7168,8,256,2,balanced,0.05637760013341904
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,384,7168,8,256,2,balanced,0.050713599845767024
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,384,7168,8,256,2,balanced,0.07747840061783791
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,384,7168,8,256,2,balanced,0.0801472008228302
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,512,7168,8,256,2,balanced,0.05540799982845783
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,512,7168,8,256,2,balanced,0.05118399970233441
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,512,7168,8,256,2,balanced,0.07833920046687126
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,512,7168,8,256,2,balanced,0.05705920048058033
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,512,7168,8,256,2,balanced,0.048041599616408345
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,512,7168,8,256,2,balanced,0.08001600056886674
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,512,7168,8,256,2,balanced,0.056486399844288826
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,512,7168,8,256,2,balanced,0.051279999688267706
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,512,7168,8,256,2,balanced,0.07729599997401237
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,512,7168,8,256,2,balanced,0.08575680032372475
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,768,7168,8,256,2,balanced,0.0565311998128891
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,768,7168,8,256,2,balanced,0.057705600559711454
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,768,7168,8,256,2,balanced,0.08975680023431779
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,768,7168,8,256,2,balanced,0.05884800031781197
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,768,7168,8,256,2,balanced,0.05254079960286617
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,768,7168,8,256,2,balanced,0.08978559970855712
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,6144,7168,8,256,4,balanced,0.0795039989054203
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,6144,7168,8,256,4,balanced,0.09875840097665786
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,6144,7168,8,256,4,balanced,0.34543359875679014
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,6144,7168,8,256,4,balanced,0.39844800233840943
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,768,7168,8,256,2,balanced,0.05650240033864975
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,768,7168,8,256,2,balanced,0.0548384003341198
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,768,7168,8,256,2,balanced,0.08885759860277176
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,768,7168,8,256,2,balanced,0.11302720010280609
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1024,7168,8,256,2,balanced,0.0586559996008873
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1024,7168,8,256,2,balanced,0.06367040053009987
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1024,7168,8,256,2,balanced,0.10313600078225135
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1024,7168,8,256,2,balanced,0.057734400406479834
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1024,7168,8,256,2,balanced,0.056198399513959885
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1024,7168,8,256,2,balanced,0.10081600025296211
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1024,7168,8,256,2,balanced,0.05640639998018741
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1024,7168,8,256,2,balanced,0.05894079990684986
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1024,7168,8,256,2,balanced,0.10256000086665154
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1024,7168,8,256,2,balanced,0.13477439880371095
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1536,7168,8,256,2,balanced,0.05941120013594627
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1536,7168,8,256,2,balanced,0.07747519984841347
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1536,7168,8,256,2,balanced,0.1338943988084793
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1536,7168,8,256,2,balanced,0.06095679886639118
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1536,7168,8,256,2,balanced,0.06498880013823509
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1536,7168,8,256,2,balanced,0.13029439896345138
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,8192,7168,8,256,4,balanced,0.09294399917125702
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,8192,7168,8,256,4,balanced,0.20179519951343536
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,8192,7168,8,256,4,balanced,0.4530047982931137
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1536,7168,8,256,2,balanced,0.05680640004575253
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1536,7168,8,256,2,balanced,0.06927039921283722
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1536,7168,8,256,2,balanced,0.13337599933147432
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1536,7168,8,256,2,balanced,0.18119999766349792
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,2048,7168,8,256,2,balanced,0.059731199964880945
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,2048,7168,8,256,2,balanced,0.09052479937672615
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,2048,7168,8,256,2,balanced,0.15961600095033646
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,2048,7168,8,256,2,balanced,0.05848640017211437
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,2048,7168,8,256,2,balanced,0.07467520013451576
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,2048,7168,8,256,2,balanced,0.15990399867296218
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,768,7168,8,256,8,balanced,0.05954559929668903
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,768,7168,8,256,8,balanced,0.09255039915442467
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,768,7168,8,256,8,balanced,0.09396160021424294
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1024,7168,8,256,8,balanced,0.07380799986422062
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1024,7168,8,256,8,balanced,0.07108800038695336
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1024,7168,8,256,8,balanced,0.10938559994101524
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1024,7168,8,256,8,balanced,0.06863039955496789
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1024,7168,8,256,8,balanced,0.05970880016684532
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1024,7168,8,256,8,balanced,0.10221439972519875
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,8192,7168,8,256,4,balanced,0.09359040036797524
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,8192,7168,8,256,4,balanced,0.12883200272917747
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,8192,7168,8,256,4,balanced,0.45022079944610593
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1024,7168,8,256,8,balanced,0.07225599884986877
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1024,7168,8,256,8,balanced,0.06034240014851093
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1024,7168,8,256,8,balanced,0.10497919991612434
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1024,7168,8,256,8,balanced,0.10316160023212433
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1536,7168,8,256,8,balanced,0.06585599929094314
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1536,7168,8,256,8,balanced,0.0788160003721714
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1536,7168,8,256,8,balanced,0.13500479906797408
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1536,7168,8,256,8,balanced,0.06990719996392727
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1536,7168,8,256,8,balanced,0.06082879975438118
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1536,7168,8,256,8,balanced,0.12872959971427916
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1536,7168,8,256,8,balanced,0.06692800000309944
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1536,7168,8,256,8,balanced,0.05789440013468265
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1536,7168,8,256,8,balanced,0.13350720033049585
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1536,7168,8,256,8,balanced,0.13206080049276353
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,2048,7168,8,256,2,balanced,0.0600064005702734
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,2048,7168,8,256,2,balanced,0.07992319911718368
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,2048,7168,8,256,2,balanced,0.15982720106840134
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,2048,7168,8,256,2,balanced,0.22123840004205703
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,2048,7168,8,256,8,balanced,0.06344639956951141
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,2048,7168,8,256,8,balanced,0.0869664005935192
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,2048,7168,8,256,8,balanced,0.1568479984998703
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,8192,7168,8,256,4,balanced,0.09337600022554397
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,8192,7168,8,256,4,balanced,0.12134079933166504
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,8192,7168,8,256,4,balanced,0.45040959417819976
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,8192,7168,8,256,4,balanced,0.5160255908966065
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,3072,7168,8,256,2,balanced,0.06517120040953159
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,3072,7168,8,256,2,balanced,0.11919359937310219
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,3072,7168,8,256,2,balanced,0.2161632001399994
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,2048,7168,8,256,8,balanced,0.07240640036761761
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,2048,7168,8,256,8,balanced,0.06741760149598122
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,2048,7168,8,256,8,balanced,0.15873919874429704
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,2048,7168,8,256,8,balanced,0.07088639922440051
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,2048,7168,8,256,8,balanced,0.06103039979934692
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,2048,7168,8,256,8,balanced,0.15736639946699144
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,2048,7168,8,256,8,balanced,0.15314880162477493
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,3072,7168,8,256,2,balanced,0.06288959980010986
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,3072,7168,8,256,2,balanced,0.09266239926218986
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,3072,7168,8,256,2,balanced,0.21450559943914413
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,3072,7168,8,256,8,balanced,0.0718688003718853
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,3072,7168,8,256,8,balanced,0.11019839867949485
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,3072,7168,8,256,8,balanced,0.2098495990037918
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,3072,7168,8,256,2,balanced,0.06198720075190067
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,3072,7168,8,256,2,balanced,0.10047999918460845
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,3072,7168,8,256,2,balanced,0.21483519822359085
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,3072,7168,8,256,2,balanced,0.3086112022399902
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,3072,7168,8,256,8,balanced,0.07012799978256226
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,3072,7168,8,256,8,balanced,0.07423360049724578
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,3072,7168,8,256,8,balanced,0.2052159994840622
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,12288,7168,8,256,4,balanced,0.11857279986143113
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,12288,7168,8,256,4,balanced,0.28225280046463014
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,4096,7168,8,256,2,balanced,0.06888319998979568
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,12288,7168,8,256,4,balanced,0.6569983899593353
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,4096,7168,8,256,2,balanced,0.14315839856863022
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,4096,7168,8,256,2,balanced,0.27097920179367063
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,3072,7168,8,256,8,balanced,0.06950399912893772
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,3072,7168,8,256,8,balanced,0.06565440036356449
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,3072,7168,8,256,8,balanced,0.20891520231962205
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,3072,7168,8,256,8,balanced,0.20396800041198732
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,4096,7168,8,256,2,balanced,0.06932800039649009
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,4096,7168,8,256,2,balanced,0.11840960010886192
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,4096,7168,8,256,2,balanced,0.26960960030555725
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,4096,7168,8,256,8,balanced,0.07644159942865372
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,4096,7168,8,256,8,balanced,0.12931520044803618
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,4096,7168,8,256,8,balanced,0.2637600004673004
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,4096,7168,8,256,2,balanced,0.06815999895334243
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,4096,7168,8,256,2,balanced,0.11939200013875961
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,4096,7168,8,256,2,balanced,0.27050880193710325
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1,7168,8,256,16,balanced,0.07031360045075416
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1,7168,8,256,16,balanced,0.05768640041351318
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1,7168,8,256,16,balanced,0.07144960090517997
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1,7168,8,256,16,balanced,0.069513601064682
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1,7168,8,256,16,balanced,0.055871999263763426
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1,7168,8,256,16,balanced,0.07183680012822151
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1,7168,8,256,16,balanced,0.06665599942207337
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1,7168,8,256,16,balanced,0.07243840098381042
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1,7168,8,256,16,balanced,0.08118720054626465
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1,7168,8,256,16,balanced,0.07224959991872311
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,2,7168,8,256,16,balanced,0.07338879965245723
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,2,7168,8,256,16,balanced,0.0641344003379345
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,2,7168,8,256,16,balanced,0.07291520088911056
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,2,7168,8,256,16,balanced,0.07549439892172813
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,2,7168,8,256,16,balanced,0.0665535993874073
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,2,7168,8,256,16,balanced,0.06821760013699532
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,2,7168,8,256,16,balanced,0.06701439954340457
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,2,7168,8,256,16,balanced,0.06174080036580563
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,2,7168,8,256,16,balanced,0.06973119974136352
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,2,7168,8,256,16,balanced,0.07074240073561669
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,4,7168,8,256,16,balanced,0.0702431995421648
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,4,7168,8,256,16,balanced,0.06295999996364117
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,4,7168,8,256,16,balanced,0.07278719991445541
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,4,7168,8,256,16,balanced,0.06921919956803321
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,4,7168,8,256,16,balanced,0.06007679924368858
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,4,7168,8,256,16,balanced,0.07149759978055954
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,4,7168,8,256,16,balanced,0.06781440116465091
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,4,7168,8,256,16,balanced,0.060041599348187445
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,4,7168,8,256,16,balanced,0.07214079946279525
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,4,7168,8,256,16,balanced,0.07410879954695701
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,8,7168,8,256,16,balanced,0.06611519977450371
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,8,7168,8,256,16,balanced,0.055619199946522715
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,8,7168,8,256,16,balanced,0.07415359914302826
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,8,7168,8,256,16,balanced,0.06655679903924465
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,8,7168,8,256,16,balanced,0.05677120052278042
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,8,7168,8,256,16,balanced,0.07647679969668389
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,8,7168,8,256,16,balanced,0.06539840064942837
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,8,7168,8,256,16,balanced,0.05456319972872734
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,8,7168,8,256,16,balanced,0.07249919921159745
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,8,7168,8,256,16,balanced,0.074252799898386
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,4096,7168,8,256,8,balanced,0.07374719902873039
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,4096,7168,8,256,8,balanced,0.08601600229740143
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,4096,7168,8,256,8,balanced,0.2612575963139534
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,12288,7168,8,256,4,balanced,0.11544320061802864
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,12288,7168,8,256,4,balanced,0.17661440074443818
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,12288,7168,8,256,4,balanced,0.6524959981441498
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,4096,7168,8,256,8,balanced,0.07237759940326213
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,4096,7168,8,256,8,balanced,0.07375999987125396
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,4096,7168,8,256,8,balanced,0.2602976024150848
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,4096,7168,8,256,8,balanced,0.24768640100955963
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,4096,7168,8,256,2,balanced,0.3907999962568283
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,16,7168,8,256,16,balanced,0.06249920055270195
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,16,7168,8,256,16,balanced,0.05552319996058941
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,16,7168,8,256,16,balanced,0.07944959998130799
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,16,7168,8,256,16,balanced,0.06210879981517792
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,16,7168,8,256,16,balanced,0.05319680012762547
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,16,7168,8,256,16,balanced,0.08248959854245186
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,16,7168,8,256,16,balanced,0.06451200060546398
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,16,7168,8,256,16,balanced,0.05719360001385212
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,16,7168,8,256,16,balanced,0.08040319979190827
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,16,7168,8,256,16,balanced,0.07868800014257431
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,32,7168,8,256,16,balanced,0.06401600055396557
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,32,7168,8,256,16,balanced,0.057996800914406774
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,32,7168,8,256,16,balanced,0.0861407995223999
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,32,7168,8,256,16,balanced,0.06599040180444718
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,32,7168,8,256,16,balanced,0.054633600264787675
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,32,7168,8,256,16,balanced,0.08627520054578781
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,32,7168,8,256,16,balanced,0.0633568000048399
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,32,7168,8,256,16,balanced,0.052207999676465985
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,32,7168,8,256,16,balanced,0.08020799979567528
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,32,7168,8,256,16,balanced,0.07984319999814034
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,48,7168,8,256,16,balanced,0.0616895992308855
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,48,7168,8,256,16,balanced,0.05920960009098053
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,48,7168,8,256,16,balanced,0.08063360080122947
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,48,7168,8,256,16,balanced,0.061606400460004804
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,48,7168,8,256,16,balanced,0.05110719986259937
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,48,7168,8,256,16,balanced,0.0794175997376442
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,48,7168,8,256,16,balanced,0.06073920093476772
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,48,7168,8,256,16,balanced,0.055481599643826485
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,48,7168,8,256,16,balanced,0.08056000173091889
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,48,7168,8,256,16,balanced,0.08191679939627647
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,64,7168,8,256,16,balanced,0.06003840006887913
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,64,7168,8,256,16,balanced,0.056249600276350974
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,64,7168,8,256,16,balanced,0.08057280108332635
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,64,7168,8,256,16,balanced,0.05976639948785305
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,64,7168,8,256,16,balanced,0.05162559971213341
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,64,7168,8,256,16,balanced,0.08164799958467484
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,64,7168,8,256,16,balanced,0.0603487990796566
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,64,7168,8,256,16,balanced,0.05260479971766472
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,64,7168,8,256,16,balanced,0.08668159916996956
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,64,7168,8,256,16,balanced,0.07493759989738465
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,80,7168,8,256,16,balanced,0.05908799991011619
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,80,7168,8,256,16,balanced,0.05220159962773323
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,80,7168,8,256,16,balanced,0.07312319874763488
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,80,7168,8,256,16,balanced,0.05944960042834282
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,80,7168,8,256,16,balanced,0.049644799903035164
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,80,7168,8,256,16,balanced,0.0767039991915226
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,80,7168,8,256,16,balanced,0.05662080012261868
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,80,7168,8,256,16,balanced,0.05296000018715859
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,80,7168,8,256,16,balanced,0.08539200052618981
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,80,7168,8,256,16,balanced,0.07773440107703208
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,96,7168,8,256,16,balanced,0.06173119992017746
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,96,7168,8,256,16,balanced,0.053251199796795844
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,96,7168,8,256,16,balanced,0.07853439971804618
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,96,7168,8,256,16,balanced,0.05524799972772598
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,96,7168,8,256,16,balanced,0.051583999395370485
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,6144,7168,8,256,8,balanced,0.0859264001250267
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,6144,7168,8,256,8,balanced,0.17456320077180862
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,6144,7168,8,256,8,balanced,0.36365759968757627
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,6144,7168,8,256,2,balanced,0.08245759978890418
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,6144,7168,8,256,2,balanced,0.19838720113039016
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,6144,7168,8,256,2,balanced,0.38218240439891815
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,12288,7168,8,256,4,balanced,0.11685120165348054
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,12288,7168,8,256,4,balanced,0.16379840075969695
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,12288,7168,8,256,4,balanced,0.6530144035816192
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,12288,7168,8,256,4,balanced,0.7506175935268402
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,6144,7168,8,256,8,balanced,0.0860447995364666
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,6144,7168,8,256,8,balanced,0.11287360042333602
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,6144,7168,8,256,8,balanced,0.3635232001543045
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,96,7168,8,256,16,balanced,0.07766400128602982
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,96,7168,8,256,16,balanced,0.06231039986014366
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,96,7168,8,256,16,balanced,0.04945599958300591
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,96,7168,8,256,16,balanced,0.08216319978237152
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,96,7168,8,256,16,balanced,0.07985280007123947
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,128,7168,8,256,16,balanced,0.05697920024394989
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,128,7168,8,256,16,balanced,0.05545599982142448
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,128,7168,8,256,16,balanced,0.07461760081350803
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,6144,7168,8,256,2,balanced,0.08151039853692055
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,6144,7168,8,256,2,balanced,0.16272320002317428
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,6144,7168,8,256,2,balanced,0.38144000470638273
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,128,7168,8,256,16,balanced,0.05518720000982284
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,128,7168,8,256,16,balanced,0.04811519980430603
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,128,7168,8,256,16,balanced,0.07867520079016685
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,128,7168,8,256,16,balanced,0.0565055999904871
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,128,7168,8,256,16,balanced,0.050726399570703504
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,128,7168,8,256,16,balanced,0.0857728011906147
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,128,7168,8,256,16,balanced,0.07429120019078254
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,160,7168,8,256,16,balanced,0.055667200684547426
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,160,7168,8,256,16,balanced,0.05763200037181378
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,160,7168,8,256,16,balanced,0.07419520020484924
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,160,7168,8,256,16,balanced,0.05734719969332218
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,160,7168,8,256,16,balanced,0.05240960009396076
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,160,7168,8,256,16,balanced,0.07713280022144317
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,160,7168,8,256,16,balanced,0.06212479993700981
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,160,7168,8,256,16,balanced,0.054128000512719154
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,160,7168,8,256,16,balanced,0.0797664001584053
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,160,7168,8,256,16,balanced,0.07570879980921745
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,192,7168,8,256,16,balanced,0.058751999586820605
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,192,7168,8,256,16,balanced,0.055772799998521805
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,192,7168,8,256,16,balanced,0.07742079943418503
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,192,7168,8,256,16,balanced,0.058102399483323096
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,192,7168,8,256,16,balanced,0.0508799996227026
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,192,7168,8,256,16,balanced,0.07814720049500465
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,192,7168,8,256,16,balanced,0.06024320051074028
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,192,7168,8,256,16,balanced,0.04965759962797165
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,192,7168,8,256,16,balanced,0.07410240098834038
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,192,7168,8,256,16,balanced,0.07438079863786698
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,256,7168,8,256,16,balanced,0.060659199580550195
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,256,7168,8,256,16,balanced,0.05525120012462139
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,256,7168,8,256,16,balanced,0.07546880021691323
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,256,7168,8,256,16,balanced,0.06257920116186141
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,256,7168,8,256,16,balanced,0.05106559991836548
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,256,7168,8,256,16,balanced,0.07344000115990638
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,256,7168,8,256,16,balanced,0.05891840048134327
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,256,7168,8,256,16,balanced,0.05166079923510551
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,256,7168,8,256,16,balanced,0.07742400020360947
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,256,7168,8,256,16,balanced,0.07739520072937012
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,320,7168,8,256,16,balanced,0.05940800085663796
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,320,7168,8,256,16,balanced,0.0568671990185976
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,320,7168,8,256,16,balanced,0.07864320054650306
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,6144,7168,8,256,8,balanced,0.08572160005569458
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,320,7168,8,256,16,balanced,0.05874880030751228
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,320,7168,8,256,16,balanced,0.05060800015926361
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,320,7168,8,256,16,balanced,0.07730560004711151
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,320,7168,8,256,16,balanced,0.0619839996099472
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,320,7168,8,256,16,balanced,0.051769600808620454
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,320,7168,8,256,16,balanced,0.07374080121517182
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,320,7168,8,256,16,balanced,0.07864639908075333
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,384,7168,8,256,16,balanced,0.05836159959435463
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,384,7168,8,256,16,balanced,0.05933120027184487
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,384,7168,8,256,16,balanced,0.0812063992023468
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,384,7168,8,256,16,balanced,0.062435200437903404
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,384,7168,8,256,16,balanced,0.048927999660372735
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,384,7168,8,256,16,balanced,0.07542399913072587
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,384,7168,8,256,16,balanced,0.059040000289678575
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,384,7168,8,256,16,balanced,0.05136639960110188
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,384,7168,8,256,16,balanced,0.07356800138950348
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,384,7168,8,256,16,balanced,0.07665280029177665
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,6144,7168,8,256,2,balanced,0.08233919963240624
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,6144,7168,8,256,2,balanced,0.16217279881238938
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,6144,7168,8,256,2,balanced,0.380128002166748
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,6144,7168,8,256,2,balanced,0.5654080033302307
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,512,7168,8,256,16,balanced,0.06046079955995083
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,512,7168,8,256,16,balanced,0.06120960004627705
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,512,7168,8,256,16,balanced,0.08388479948043823
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,512,7168,8,256,16,balanced,0.05757759995758534
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,512,7168,8,256,16,balanced,0.050915200263261795
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,512,7168,8,256,16,balanced,0.08000000044703484
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,512,7168,8,256,16,balanced,0.06117759980261326
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,512,7168,8,256,16,balanced,0.052111999690532686
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,512,7168,8,256,16,balanced,0.0808863990008831
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,512,7168,8,256,16,balanced,0.07794560045003891
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,768,7168,8,256,16,balanced,0.05827839933335781
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,768,7168,8,256,16,balanced,0.06775680109858513
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,768,7168,8,256,16,balanced,0.09008320048451424
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,768,7168,8,256,16,balanced,0.059542400389909746
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,768,7168,8,256,16,balanced,0.05304640047252178
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,768,7168,8,256,16,balanced,0.08951999843120576
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,768,7168,8,256,16,balanced,0.05910079926252365
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,768,7168,8,256,16,balanced,0.05299519933760166
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,768,7168,8,256,16,balanced,0.09232000038027763
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,768,7168,8,256,16,balanced,0.09029759988188743
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1024,7168,8,256,16,balanced,0.05864000022411346
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1024,7168,8,256,16,balanced,0.0713184006512165
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1024,7168,8,256,16,balanced,0.10823359936475754
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,6144,7168,8,256,8,balanced,0.08484159931540489
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,6144,7168,8,256,8,balanced,0.36144319772720335
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,6144,7168,8,256,8,balanced,0.3482015997171402
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,16384,7168,8,256,4,balanced,0.1419103965163231
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,16384,7168,8,256,4,balanced,0.36297599971294403
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,16384,7168,8,256,4,balanced,0.858953595161438
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1024,7168,8,256,16,balanced,0.06864000000059604
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1024,7168,8,256,16,balanced,0.05813440009951591
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1024,7168,8,256,16,balanced,0.10398080125451088
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1024,7168,8,256,16,balanced,0.0596063993871212
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1024,7168,8,256,16,balanced,0.05384959988296032
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1024,7168,8,256,16,balanced,0.1081887997686863
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1024,7168,8,256,16,balanced,0.10477759912610055
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,8192,7168,8,256,2,balanced,0.0956735998392105
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,8192,7168,8,256,2,balanced,0.25142719894647597
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,8192,7168,8,256,2,balanced,0.49359680712223053
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1536,7168,8,256,16,balanced,0.06726719997823238
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1536,7168,8,256,16,balanced,0.08081279993057251
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1536,7168,8,256,16,balanced,0.13345280140638352
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1536,7168,8,256,16,balanced,0.06586240045726299
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1536,7168,8,256,16,balanced,0.0616480004042387
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1536,7168,8,256,16,balanced,0.1327327996492386
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1536,7168,8,256,16,balanced,0.06055999957025051
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1536,7168,8,256,16,balanced,0.05755519941449165
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1536,7168,8,256,16,balanced,0.13434880077838898
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1536,7168,8,256,16,balanced,0.13035200014710427
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,8192,7168,8,256,8,balanced,0.0983135998249054
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,8192,7168,8,256,8,balanced,0.21305599957704544
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,8192,7168,8,256,8,balanced,0.4627967983484268
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,2048,7168,8,256,16,balanced,0.06277120001614094
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,2048,7168,8,256,16,balanced,0.09472640082240105
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,2048,7168,8,256,16,balanced,0.16013120114803314
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,8192,7168,8,256,2,balanced,0.09570559933781624
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,8192,7168,8,256,2,balanced,0.20556480139493943
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,8192,7168,8,256,2,balanced,0.4893375992774963
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,2048,7168,8,256,16,balanced,0.06040639914572239
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,2048,7168,8,256,16,balanced,0.0701983992010355
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,2048,7168,8,256,16,balanced,0.15624640136957169
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,2048,7168,8,256,16,balanced,0.06330880001187325
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,2048,7168,8,256,16,balanced,0.060185599699616434
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,2048,7168,8,256,16,balanced,0.15921920090913771
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,2048,7168,8,256,16,balanced,0.1526559993624687
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,3072,7168,8,256,16,balanced,0.06880000010132789
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,3072,7168,8,256,16,balanced,0.11564160138368607
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,3072,7168,8,256,16,balanced,0.2322880044579506
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,16384,7168,8,256,4,balanced,0.1419423997402191
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,16384,7168,8,256,4,balanced,0.2243487998843193
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,16384,7168,8,256,4,balanced,0.8571904063224792
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,8192,7168,8,256,8,balanced,0.09890239834785461
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,8192,7168,8,256,8,balanced,0.1324480004608631
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,8192,7168,8,256,8,balanced,0.4608672022819519
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,8192,7168,8,256,2,balanced,0.09709120020270348
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,8192,7168,8,256,2,balanced,0.2040479987859726
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,8192,7168,8,256,2,balanced,0.4904416024684906
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,8192,7168,8,256,2,balanced,0.7379072070121765
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,3072,7168,8,256,16,balanced,0.06775999926030636
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,3072,7168,8,256,16,balanced,0.07880640029907227
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,3072,7168,8,256,16,balanced,0.21164159923791886
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,3072,7168,8,256,16,balanced,0.0682240005582571
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,3072,7168,8,256,16,balanced,0.06899199932813645
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,3072,7168,8,256,16,balanced,0.21539520025253295
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,3072,7168,8,256,16,balanced,0.2005215987563133
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,8192,7168,8,256,8,balanced,0.09972800090909004
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,8192,7168,8,256,8,balanced,0.09882239997386932
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,8192,7168,8,256,8,balanced,0.46396480202674867
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,8192,7168,8,256,8,balanced,0.4353119969367981
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,4096,7168,8,256,16,balanced,0.07407039999961854
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,4096,7168,8,256,16,balanced,0.13699839860200883
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,4096,7168,8,256,16,balanced,0.2637791961431503
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,12288,7168,8,256,2,balanced,0.1245919980108738
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,12288,7168,8,256,2,balanced,0.3594271957874298
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,12288,7168,8,256,2,balanced,0.7121823966503144
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,4096,7168,8,256,16,balanced,0.07304639965295792
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,4096,7168,8,256,16,balanced,0.09171200022101403
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,4096,7168,8,256,16,balanced,0.26525439918041227
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,16384,7168,8,256,4,balanced,0.1401759997010231
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,16384,7168,8,256,4,balanced,0.2078687995672226
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,16384,7168,8,256,4,balanced,0.8552287995815278
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,16384,7168,8,256,4,balanced,0.9848992049694061
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,4096,7168,8,256,16,balanced,0.0723904013633728
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,4096,7168,8,256,16,balanced,0.07767039984464645
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,4096,7168,8,256,16,balanced,0.2663296014070511
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,4096,7168,8,256,16,balanced,0.2446688026189804
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,12288,7168,8,256,8,balanced,0.12451200038194657
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,12288,7168,8,256,8,balanced,0.296713599562645
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,12288,7168,8,256,8,balanced,0.668639999628067
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,6144,7168,8,256,16,balanced,0.08873919993638993
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,6144,7168,8,256,16,balanced,0.1861344024538994
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,6144,7168,8,256,16,balanced,0.37039360105991365
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,12288,7168,8,256,2,balanced,0.1242655985057354
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,12288,7168,8,256,2,balanced,0.28720320463180543
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,12288,7168,8,256,2,balanced,0.7103616058826446
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,6144,7168,8,256,16,balanced,0.0888383999466896
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,6144,7168,8,256,16,balanced,0.11724160015583038
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,6144,7168,8,256,16,balanced,0.371302404999733
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,12288,7168,8,256,8,balanced,0.12353279888629913
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,12288,7168,8,256,8,balanced,0.1785311996936798
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,12288,7168,8,256,8,balanced,0.6665695965290069
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,20480,7168,8,256,4,balanced,0.16832960098981858
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,20480,7168,8,256,4,balanced,0.4458175957202911
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,20480,7168,8,256,4,balanced,1.0681408047676086
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,6144,7168,8,256,16,balanced,0.09008959978818894
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,6144,7168,8,256,16,balanced,0.0919680006802082
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,6144,7168,8,256,16,balanced,0.3701631993055344
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,6144,7168,8,256,16,balanced,0.33951039910316466
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,12288,7168,8,256,2,balanced,0.12336319983005524
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,12288,7168,8,256,2,balanced,0.2871264010667801
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,12288,7168,8,256,2,balanced,0.711980801820755
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,12288,7168,8,256,2,balanced,1.0792255997657776
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,12288,7168,8,256,8,balanced,0.12158720120787621
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,12288,7168,8,256,8,balanced,0.11933119967579842
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,12288,7168,8,256,8,balanced,0.6644960045814514
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,12288,7168,8,256,8,balanced,0.6282080054283142
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,8192,7168,8,256,16,balanced,0.09674559906125069
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,8192,7168,8,256,16,balanced,0.23386240154504775
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,8192,7168,8,256,16,balanced,0.477852800488472
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,8192,7168,8,256,16,balanced,0.09848959892988204
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,8192,7168,8,256,16,balanced,0.14243519753217698
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,8192,7168,8,256,16,balanced,0.4776608020067215
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,16384,7168,8,256,2,balanced,0.14926400035619736
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,16384,7168,8,256,2,balanced,0.46811520159244535
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,16384,7168,8,256,2,balanced,0.9337791979312897
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,20480,7168,8,256,4,balanced,0.16762559860944748
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,20480,7168,8,256,4,balanced,0.27437119781970976
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,20480,7168,8,256,4,balanced,1.066262423992157
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,16384,7168,8,256,8,balanced,0.1461760014295578
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,16384,7168,8,256,8,balanced,0.38273279666900634
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,16384,7168,8,256,8,balanced,0.870054405927658
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,8192,7168,8,256,16,balanced,0.1011328011751175
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,8192,7168,8,256,16,balanced,0.10878399908542633
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,8192,7168,8,256,16,balanced,0.47818879783153534
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,8192,7168,8,256,16,balanced,0.43189760148525236
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,16384,7168,8,256,2,balanced,0.14853439927101136
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,16384,7168,8,256,2,balanced,0.3729440003633499
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,16384,7168,8,256,2,balanced,0.9334496021270752
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,12288,7168,8,256,16,balanced,0.12491519898176193
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,12288,7168,8,256,16,balanced,0.32549439668655394
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,12288,7168,8,256,16,balanced,0.6909152090549469
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,20480,7168,8,256,4,balanced,0.16521280109882355
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,20480,7168,8,256,4,balanced,0.24758719950914382
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,20480,7168,8,256,4,balanced,1.0659967899322509
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,20480,7168,8,256,4,balanced,1.2254527926445007
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,16384,7168,8,256,8,balanced,0.1477600023150444
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,16384,7168,8,256,8,balanced,0.2231552004814148
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,16384,7168,8,256,8,balanced,0.870416009426117
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,12288,7168,8,256,16,balanced,0.12735040113329887
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,12288,7168,8,256,16,balanced,0.19329280108213426
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,12288,7168,8,256,16,balanced,0.6887935936450958
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,16384,7168,8,256,2,balanced,0.14714880287647247
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,16384,7168,8,256,2,balanced,0.3701344013214111
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,16384,7168,8,256,2,balanced,0.9294912099838257
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,16384,7168,8,256,2,balanced,1.4216063737869262
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1,7168,8,256,48,balanced,0.07035199999809265
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1,7168,8,256,48,balanced,0.053804800286889075
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1,7168,8,256,48,balanced,0.0772256001830101
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1,7168,8,256,48,balanced,0.07529279999434949
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1,7168,8,256,48,balanced,0.05568000003695488
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1,7168,8,256,48,balanced,0.07568640038371086
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1,7168,8,256,48,balanced,0.0695519994944334
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1,7168,8,256,48,balanced,0.05751040019094944
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1,7168,8,256,48,balanced,0.07547840029001236
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1,7168,8,256,48,balanced,0.07709119990468025
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,2,7168,8,256,48,balanced,0.06544959992170334
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,2,7168,8,256,48,balanced,0.05386239960789681
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,2,7168,8,256,48,balanced,0.07786240056157112
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,2,7168,8,256,48,balanced,0.06864320039749146
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,2,7168,8,256,48,balanced,0.06950720064342022
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,2,7168,8,256,48,balanced,0.08245759978890418
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,2,7168,8,256,48,balanced,0.07002560012042522
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,2,7168,8,256,48,balanced,0.05301759988069534
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,2,7168,8,256,48,balanced,0.08087359890341758
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,2,7168,8,256,48,balanced,0.07619519978761673
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,4,7168,8,256,48,balanced,0.06983039937913418
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,4,7168,8,256,48,balanced,0.05563200004398823
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,4,7168,8,256,48,balanced,0.07343999892473221
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,4,7168,8,256,48,balanced,0.06520319879055023
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,4,7168,8,256,48,balanced,0.053702399507164954
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,4,7168,8,256,48,balanced,0.08018559962511063
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,4,7168,8,256,48,balanced,0.06363839916884899
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,4,7168,8,256,48,balanced,0.05251839980483055
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,4,7168,8,256,48,balanced,0.07593600153923034
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,4,7168,8,256,48,balanced,0.07492479979991913
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,8,7168,8,256,48,balanced,0.07128000035881996
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,8,7168,8,256,48,balanced,0.053907199576497075
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,8,7168,8,256,48,balanced,0.08712320029735565
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,8,7168,8,256,48,balanced,0.06280639916658401
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,8,7168,8,256,48,balanced,0.0566624004393816
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,8,7168,8,256,48,balanced,0.080799999833107
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,8,7168,8,256,48,balanced,0.06452160030603409
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,8,7168,8,256,48,balanced,0.05345280095934868
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,8,7168,8,256,48,balanced,0.08623999953269959
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,8,7168,8,256,48,balanced,0.07815679982304573
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,16,7168,8,256,48,balanced,0.06558080017566681
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,16,7168,8,256,48,balanced,0.08060479946434498
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,16,7168,8,256,48,balanced,0.09265279918909072
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,16,7168,8,256,48,balanced,0.087932800501585
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,16,7168,8,256,48,balanced,0.06894720047712326
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,16,7168,8,256,48,balanced,0.09368320032954217
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,16,7168,8,256,48,balanced,0.08449280112981797
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,16,7168,8,256,48,balanced,0.07572479993104934
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,16,7168,8,256,48,balanced,0.09309440031647682
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,16,7168,8,256,48,balanced,0.12122880145907403
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,32,7168,8,256,48,balanced,0.09732799977064133
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,32,7168,8,256,48,balanced,0.07970559969544411
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,32,7168,8,256,48,balanced,0.14778880029916763
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,32,7168,8,256,48,balanced,0.09161280021071434
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,32,7168,8,256,48,balanced,0.07062719948589802
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,32,7168,8,256,48,balanced,0.1495264008641243
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,32,7168,8,256,48,balanced,0.09019839987158776
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,32,7168,8,256,48,balanced,0.07731840014457703
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,32,7168,8,256,48,balanced,0.15111999958753586
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,32,7168,8,256,48,balanced,0.13653120175004005
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,48,7168,8,256,48,balanced,0.08603520020842552
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,48,7168,8,256,48,balanced,0.08634880036115647
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,48,7168,8,256,48,balanced,0.1515167996287346
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,48,7168,8,256,48,balanced,0.0886624000966549
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,48,7168,8,256,48,balanced,0.0756191998720169
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,48,7168,8,256,48,balanced,0.14951040148735045
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,48,7168,8,256,48,balanced,0.09062719941139222
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,48,7168,8,256,48,balanced,0.08033920004963875
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,48,7168,8,256,48,balanced,0.14384959936141967
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,48,7168,8,256,48,balanced,0.1448608011007309
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,64,7168,8,256,48,balanced,0.09335360080003738
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,64,7168,8,256,48,balanced,0.07881599962711335
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,64,7168,8,256,48,balanced,0.1469631977379322
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,64,7168,8,256,48,balanced,0.08521600067615509
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,64,7168,8,256,48,balanced,0.09260479882359504
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,64,7168,8,256,48,balanced,0.15460799932479857
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,64,7168,8,256,48,balanced,0.09008959904313088
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,64,7168,8,256,48,balanced,0.07662400007247924
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,64,7168,8,256,48,balanced,0.14920639991760254
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,64,7168,8,256,48,balanced,0.15048960000276565
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,80,7168,8,256,48,balanced,0.09206079989671707
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,80,7168,8,256,48,balanced,0.07542399987578392
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,80,7168,8,256,48,balanced,0.14792319983243943
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,80,7168,8,256,48,balanced,0.08329279944300652
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,80,7168,8,256,48,balanced,0.07407040074467659
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,80,7168,8,256,48,balanced,0.14082239866256713
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,16384,7168,8,256,8,balanced,0.14727360159158706
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,16384,7168,8,256,8,balanced,0.17186879962682725
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,16384,7168,8,256,8,balanced,0.8695007979869842
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,16384,7168,8,256,8,balanced,0.8085536003112793
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,12288,7168,8,256,16,balanced,0.12514240071177482
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,12288,7168,8,256,16,balanced,0.14192319959402083
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,12288,7168,8,256,16,balanced,0.6897760093212127
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,12288,7168,8,256,16,balanced,0.6194527924060822
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,80,7168,8,256,48,balanced,0.08951999992132187
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,80,7168,8,256,48,balanced,0.07370560020208358
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,80,7168,8,256,48,balanced,0.16040000170469285
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,80,7168,8,256,48,balanced,0.14752000272274018
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,96,7168,8,256,48,balanced,0.08662080094218254
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,96,7168,8,256,48,balanced,0.0836671993136406
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,96,7168,8,256,48,balanced,0.1523488000035286
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,96,7168,8,256,48,balanced,0.086684799939394
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,96,7168,8,256,48,balanced,0.07335040047764778
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,96,7168,8,256,48,balanced,0.13691200017929078
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,96,7168,8,256,48,balanced,0.09262719973921776
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,96,7168,8,256,48,balanced,0.08528639972209931
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,96,7168,8,256,48,balanced,0.15923519879579545
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,96,7168,8,256,48,balanced,0.15344959795475005
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,128,7168,8,256,48,balanced,0.08852480053901672
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,128,7168,8,256,48,balanced,0.08499839901924133
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,128,7168,8,256,48,balanced,0.1321024015545845
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,128,7168,8,256,48,balanced,0.08263039961457253
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,128,7168,8,256,48,balanced,0.070991999655962
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,128,7168,8,256,48,balanced,0.14350400269031524
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,128,7168,8,256,48,balanced,0.08145599998533726
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,128,7168,8,256,48,balanced,0.07695360034704209
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,128,7168,8,256,48,balanced,0.1501183971762657
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,128,7168,8,256,48,balanced,0.13940479829907418
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,160,7168,8,256,48,balanced,0.08897600024938583
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,160,7168,8,256,48,balanced,0.08028160035610199
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,160,7168,8,256,48,balanced,0.14218879938125611
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,160,7168,8,256,48,balanced,0.08432640135288239
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,160,7168,8,256,48,balanced,0.07650239989161492
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,160,7168,8,256,48,balanced,0.1465407982468605
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,160,7168,8,256,48,balanced,0.08051200062036515
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,160,7168,8,256,48,balanced,0.07319359928369522
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,160,7168,8,256,48,balanced,0.13820479959249496
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,160,7168,8,256,48,balanced,0.1476032018661499
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,192,7168,8,256,48,balanced,0.08329280018806458
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,192,7168,8,256,48,balanced,0.08263999968767166
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,192,7168,8,256,48,balanced,0.1349983975291252
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,192,7168,8,256,48,balanced,0.08238399997353554
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,192,7168,8,256,48,balanced,0.07346239909529687
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,192,7168,8,256,48,balanced,0.13789760023355485
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,192,7168,8,256,48,balanced,0.0792095996439457
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,192,7168,8,256,48,balanced,0.07512960061430932
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,192,7168,8,256,48,balanced,0.13363839834928512
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,192,7168,8,256,48,balanced,0.1392960011959076
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,256,7168,8,256,48,balanced,0.08033599965274334
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,256,7168,8,256,48,balanced,0.08511359989643097
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,256,7168,8,256,48,balanced,0.14363519996404647
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,256,7168,8,256,48,balanced,0.08625919967889786
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,256,7168,8,256,48,balanced,0.0746848002076149
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,256,7168,8,256,48,balanced,0.14060800075531005
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,256,7168,8,256,48,balanced,0.08885119929909706
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,256,7168,8,256,48,balanced,0.07719680145382882
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,256,7168,8,256,48,balanced,0.1409376010298729
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,256,7168,8,256,48,balanced,0.14451839849352838
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,320,7168,8,256,48,balanced,0.08596480041742324
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,320,7168,8,256,48,balanced,0.09315200075507164
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,320,7168,8,256,48,balanced,0.14578560069203378
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,20480,7168,8,256,2,balanced,0.17532479912042617
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,20480,7168,8,256,2,balanced,0.5700640082359314
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,20480,7168,8,256,2,balanced,1.157753586769104
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,320,7168,8,256,48,balanced,0.08402880057692527
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,320,7168,8,256,48,balanced,0.07699519991874695
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,320,7168,8,256,48,balanced,0.14832640066742897
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,320,7168,8,256,48,balanced,0.09188800007104873
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,320,7168,8,256,48,balanced,0.07391680032014847
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,320,7168,8,256,48,balanced,0.143539197742939
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,320,7168,8,256,48,balanced,0.14682559967041015
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,384,7168,8,256,48,balanced,0.08310720026493072
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,384,7168,8,256,48,balanced,0.08389120101928711
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,384,7168,8,256,48,balanced,0.14083520025014878
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,384,7168,8,256,48,balanced,0.07994880005717278
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,384,7168,8,256,48,balanced,0.07039679922163486
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,384,7168,8,256,48,balanced,0.1401823990046978
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,384,7168,8,256,48,balanced,0.0828064002096653
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,384,7168,8,256,48,balanced,0.06933439970016479
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,384,7168,8,256,48,balanced,0.14117120131850242
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,384,7168,8,256,48,balanced,0.13853119760751725
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,32768,7168,8,256,4,balanced,0.24469119906425477
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,32768,7168,8,256,4,balanced,0.6949568092823029
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,32768,7168,8,256,4,balanced,1.6755584001541137
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,512,7168,8,256,48,balanced,0.07944959998130799
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,512,7168,8,256,48,balanced,0.09103999957442284
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,512,7168,8,256,48,balanced,0.13635200038552284
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,512,7168,8,256,48,balanced,0.08416320085525512
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,512,7168,8,256,48,balanced,0.07109119929373264
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,512,7168,8,256,48,balanced,0.13899200037121773
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,512,7168,8,256,48,balanced,0.09115520119667053
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,512,7168,8,256,48,balanced,0.07223040089011193
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,512,7168,8,256,48,balanced,0.1369152009487152
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,512,7168,8,256,48,balanced,0.13851840049028397
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,768,7168,8,256,48,balanced,0.08530240133404732
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,768,7168,8,256,48,balanced,0.10070079937577248
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,768,7168,8,256,48,balanced,0.1563295990228653
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,768,7168,8,256,48,balanced,0.08431040048599243
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,768,7168,8,256,48,balanced,0.08468800038099289
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,768,7168,8,256,48,balanced,0.15807680189609527
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,768,7168,8,256,48,balanced,0.08745600059628486
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,768,7168,8,256,48,balanced,0.07833599895238877
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,768,7168,8,256,48,balanced,0.15755840092897416
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,768,7168,8,256,48,balanced,0.14728640168905258
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,20480,7168,8,256,8,balanced,0.17202560007572174
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,20480,7168,8,256,8,balanced,0.46546879708766936
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,20480,7168,8,256,8,balanced,1.0803200006484985
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1024,7168,8,256,48,balanced,0.08738240078091622
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1024,7168,8,256,48,balanced,0.11641919985413551
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1024,7168,8,256,48,balanced,0.18241599798202515
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,16384,7168,8,256,16,balanced,0.14775360226631165
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,16384,7168,8,256,16,balanced,0.4223679959774017
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,16384,7168,8,256,16,balanced,0.9046208083629608
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1024,7168,8,256,48,balanced,0.08576000034809113
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1024,7168,8,256,48,balanced,0.08524800017476082
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1024,7168,8,256,48,balanced,0.181734399497509
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1024,7168,8,256,48,balanced,0.0983808010816574
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1024,7168,8,256,48,balanced,0.08079040050506592
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1024,7168,8,256,48,balanced,0.18279999792575835
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1024,7168,8,256,48,balanced,0.17109760046005248
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1536,7168,8,256,48,balanced,0.08569599986076355
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1536,7168,8,256,48,balanced,0.14875199794769287
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1536,7168,8,256,48,balanced,0.2300031989812851
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1536,7168,8,256,48,balanced,0.08742720037698745
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1536,7168,8,256,48,balanced,0.10129920020699501
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1536,7168,8,256,48,balanced,0.2304671972990036
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1536,7168,8,256,48,balanced,0.08451200053095817
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1536,7168,8,256,48,balanced,0.08982720002532005
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1536,7168,8,256,48,balanced,0.2308320015668869
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1536,7168,8,256,48,balanced,0.20699840039014816
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,2048,7168,8,256,48,balanced,0.08650879934430122
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,2048,7168,8,256,48,balanced,0.18224319964647293
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,2048,7168,8,256,48,balanced,0.28101440370082853
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,20480,7168,8,256,2,balanced,0.17272959947586058
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,20480,7168,8,256,2,balanced,0.45875519812107085
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,20480,7168,8,256,2,balanced,1.1579808235168456
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,2048,7168,8,256,48,balanced,0.08904959931969643
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,2048,7168,8,256,48,balanced,0.11764800101518631
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,2048,7168,8,256,48,balanced,0.2854400008916855
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,2048,7168,8,256,48,balanced,0.0876960001885891
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,2048,7168,8,256,48,balanced,0.09833279922604561
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,2048,7168,8,256,48,balanced,0.2861952006816864
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,2048,7168,8,256,48,balanced,0.2564704015851021
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,16384,7168,8,256,16,balanced,0.14692800045013427
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,16384,7168,8,256,16,balanced,0.2423807978630066
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,16384,7168,8,256,16,balanced,0.9016287982463836
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,3072,7168,8,256,48,balanced,0.08948159962892532
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,3072,7168,8,256,48,balanced,0.2498208001255989
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,3072,7168,8,256,48,balanced,0.37608639895915985
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,20480,7168,8,256,8,balanced,0.17039679884910583
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,20480,7168,8,256,8,balanced,0.26566079556941985
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,20480,7168,8,256,8,balanced,1.0795232057571411
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,3072,7168,8,256,48,balanced,0.08631680011749268
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,3072,7168,8,256,48,balanced,0.15041920095682143
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,3072,7168,8,256,48,balanced,0.3728735983371735
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,3072,7168,8,256,48,balanced,0.08900480046868324
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,3072,7168,8,256,48,balanced,0.11864000037312508
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,3072,7168,8,256,48,balanced,0.3775808036327362
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,3072,7168,8,256,48,balanced,0.33619519472122195
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,32768,7168,8,256,4,balanced,0.24477120041847228
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,32768,7168,8,256,4,balanced,0.4119712024927139
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,32768,7168,8,256,4,balanced,1.6729248046875
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,4096,7168,8,256,48,balanced,0.0906016007065773
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,4096,7168,8,256,48,balanced,0.31834560334682466
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,4096,7168,8,256,48,balanced,0.4748863995075226
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,20480,7168,8,256,2,balanced,0.1742047995328903
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,20480,7168,8,256,2,balanced,0.4522175967693329
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,20480,7168,8,256,2,balanced,1.161033606529236
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,20480,7168,8,256,2,balanced,1.770803201198578
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,16384,7168,8,256,16,balanced,0.1518527999520302
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,16384,7168,8,256,16,balanced,0.17219839990139008
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,16384,7168,8,256,16,balanced,0.9044703900814056
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,16384,7168,8,256,16,balanced,0.8030655980110168
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,4096,7168,8,256,48,balanced,0.08955200016498566
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,4096,7168,8,256,48,balanced,0.18516159951686859
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,4096,7168,8,256,48,balanced,0.4809951990842819
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,4096,7168,8,256,48,balanced,0.09768640026450157
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,4096,7168,8,256,48,balanced,0.1396127998828888
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,4096,7168,8,256,48,balanced,0.48264639973640444
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,4096,7168,8,256,48,balanced,0.414806404709816
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,20480,7168,8,256,8,balanced,0.16887039989233016
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,20480,7168,8,256,8,balanced,0.17099200040102006
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,20480,7168,8,256,8,balanced,1.0797888040542603
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,20480,7168,8,256,8,balanced,1.003171193599701
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,6144,7168,8,256,48,balanced,0.09696319922804833
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,6144,7168,8,256,48,balanced,0.45701760053634644
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,6144,7168,8,256,48,balanced,0.6672063946723938
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,6144,7168,8,256,48,balanced,0.10069120079278945
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,6144,7168,8,256,48,balanced,0.25416320115327834
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,6144,7168,8,256,48,balanced,0.6646111905574799
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,20480,7168,8,256,16,balanced,0.17081920355558394
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,20480,7168,8,256,16,balanced,0.514105600118637
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,20480,7168,8,256,16,balanced,1.1192768096923829
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,32768,7168,8,256,4,balanced,0.24440960437059403
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,32768,7168,8,256,4,balanced,0.3716511994600296
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,32768,7168,8,256,4,balanced,1.6732416152954102
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,6144,7168,8,256,48,balanced,0.10053760036826134
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,32768,7168,8,256,4,balanced,1.9304255962371826
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,6144,7168,8,256,48,balanced,0.1783840000629425
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,6144,7168,8,256,48,balanced,0.669187194108963
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,6144,7168,8,256,48,balanced,0.5789792001247406
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,32768,7168,8,256,2,balanced,0.25472959876060486
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,32768,7168,8,256,2,balanced,0.9258016049861908
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,32768,7168,8,256,2,balanced,1.9075423955917359
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,8192,7168,8,256,48,balanced,0.10981760025024415
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,8192,7168,8,256,48,balanced,0.5977759957313538
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,8192,7168,8,256,48,balanced,0.8706368029117584
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,32768,7168,8,256,8,balanced,0.24656959921121596
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,32768,7168,8,256,8,balanced,0.7195487976074219
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,32768,7168,8,256,8,balanced,1.6915520071983337
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,20480,7168,8,256,16,balanced,0.1710656017065048
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,20480,7168,8,256,16,balanced,0.2883807957172394
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,20480,7168,8,256,16,balanced,1.1209440231323242
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,8192,7168,8,256,48,balanced,0.11169280037283898
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,8192,7168,8,256,48,balanced,0.32760640382766726
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,8192,7168,8,256,48,balanced,0.8740224063396453
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,8192,7168,8,256,48,balanced,0.11299839913845063
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,8192,7168,8,256,48,balanced,0.22252160012722016
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,8192,7168,8,256,48,balanced,0.8710911989212036
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,8192,7168,8,256,48,balanced,0.7456096053123474
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,20480,7168,8,256,16,balanced,0.17132480144500734
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,20480,7168,8,256,16,balanced,0.20459519773721696
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,20480,7168,8,256,16,balanced,1.1192928075790405
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,20480,7168,8,256,16,balanced,0.9981407999992371
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,32768,7168,8,256,2,balanced,0.2521024063229561
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,32768,7168,8,256,2,balanced,0.708624005317688
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,32768,7168,8,256,2,balanced,1.822275197505951
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,12288,7168,8,256,48,balanced,0.13654720038175583
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,12288,7168,8,256,48,balanced,0.8857119858264924
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,12288,7168,8,256,48,balanced,1.2737344026565551
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,32768,7168,8,256,8,balanced,0.24467839896678925
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,32768,7168,8,256,8,balanced,0.4018656015396118
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,32768,7168,8,256,8,balanced,1.6897536039352417
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,12288,7168,8,256,48,balanced,0.13944960087537767
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,12288,7168,8,256,48,balanced,0.4753887981176376
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,12288,7168,8,256,48,balanced,1.2750752091407775
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,65536,7168,8,256,4,balanced,0.4477791965007782
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,65536,7168,8,256,4,balanced,1.3503040075302124
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,65536,7168,8,256,4,balanced,3.309609627723694
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,12288,7168,8,256,48,balanced,0.13347200006246568
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,12288,7168,8,256,48,balanced,0.30776640176773074
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,12288,7168,8,256,48,balanced,1.280016005039215
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,12288,7168,8,256,48,balanced,1.0731583952903747
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,32768,7168,8,256,16,balanced,0.24381120055913924
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,32768,7168,8,256,16,balanced,0.7900543928146362
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,32768,7168,8,256,16,balanced,1.7578367829322814
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,32768,7168,8,256,2,balanced,0.25521279573440553
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,32768,7168,8,256,2,balanced,0.7014944076538085
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,32768,7168,8,256,2,balanced,1.8188127994537353
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,32768,7168,8,256,2,balanced,2.7982336044311524
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,32768,7168,8,256,8,balanced,0.2436447948217392
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,32768,7168,8,256,8,balanced,0.24760960042476654
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,32768,7168,8,256,8,balanced,1.6889952063560485
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,32768,7168,8,256,8,balanced,1.5636959791183471
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,16384,7168,8,256,48,balanced,0.15724480003118516
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,16384,7168,8,256,48,balanced,1.1751327991485596
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,16384,7168,8,256,48,balanced,1.6926144003868102
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,32768,7168,8,256,16,balanced,0.24444160014390945
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,32768,7168,8,256,16,balanced,0.4327072024345398
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,32768,7168,8,256,16,balanced,1.7564639925956727
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,16384,7168,8,256,48,balanced,0.15885120034217834
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,16384,7168,8,256,48,balanced,0.6234848022460937
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,16384,7168,8,256,48,balanced,1.687443208694458
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,16384,7168,8,256,48,balanced,0.16182079911231995
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,16384,7168,8,256,48,balanced,0.39855040311813356
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,16384,7168,8,256,48,balanced,1.6890464186668397
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,16384,7168,8,256,48,balanced,1.3957471966743469
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,65536,7168,8,256,4,balanced,0.44406079649925234
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,65536,7168,8,256,4,balanced,0.7866207957267761
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,65536,7168,8,256,4,balanced,3.3090240240097044
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,32768,7168,8,256,16,balanced,0.24495360106229783
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,32768,7168,8,256,16,balanced,0.29346559941768646
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,32768,7168,8,256,16,balanced,1.75550719499588
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,32768,7168,8,256,16,balanced,1.5481824040412904
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,65536,7168,8,256,2,balanced,0.4603616029024124
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,65536,7168,8,256,2,balanced,1.875276803970337
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,65536,7168,8,256,2,balanced,3.7555840015411377
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,20480,7168,8,256,48,balanced,0.18270719945430755
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,20480,7168,8,256,48,balanced,1.4652384161949157
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,20480,7168,8,256,48,balanced,2.126208019256592
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,65536,7168,8,256,8,balanced,0.442086398601532
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,65536,7168,8,256,8,balanced,1.3930559873580932
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,65536,7168,8,256,8,balanced,3.3308512210845946
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,20480,7168,8,256,48,balanced,0.18133120089769364
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,20480,7168,8,256,48,balanced,0.773174399137497
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,20480,7168,8,256,48,balanced,2.119935965538025
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,20480,7168,8,256,48,balanced,0.18514879941940307
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,20480,7168,8,256,48,balanced,0.48351040184497834
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,20480,7168,8,256,48,balanced,2.132102394104004
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,20480,7168,8,256,48,balanced,1.7484127879142761
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,65536,7168,8,256,4,balanced,0.4422080010175705
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,65536,7168,8,256,4,balanced,0.7073344111442565
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,65536,7168,8,256,4,balanced,3.3067999601364138
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,65536,7168,8,256,4,balanced,3.8110463619232178
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,65536,7168,8,256,2,balanced,0.46054079830646516
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,65536,7168,8,256,2,balanced,1.3800639986991883
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,65536,7168,8,256,2,balanced,3.582710385322571
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,65536,7168,8,256,16,balanced,0.4359104007482529
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,65536,7168,8,256,16,balanced,1.5280895948410034
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,65536,7168,8,256,16,balanced,3.452726411819458
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,65536,7168,8,256,8,balanced,0.4360607981681824
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,65536,7168,8,256,8,balanced,0.7686655998229981
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,65536,7168,8,256,8,balanced,3.3266495943069456
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,32768,7168,8,256,48,balanced,0.25478400141000745
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,32768,7168,8,256,48,balanced,2.350918412208557
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,32768,7168,8,256,48,balanced,3.3928704023361207
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,32768,7168,8,256,48,balanced,0.2576575994491577
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,32768,7168,8,256,48,balanced,1.2382495880126954
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,32768,7168,8,256,48,balanced,3.3952224254608154
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,65536,7168,8,256,2,balanced,0.46201920211315156
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,65536,7168,8,256,2,balanced,1.3607392072677613
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,65536,7168,8,256,2,balanced,3.761462378501892
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,65536,7168,8,256,2,balanced,5.541731119155884
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,65536,7168,8,256,16,balanced,0.4378111958503723
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,65536,7168,8,256,16,balanced,0.8201503932476044
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,65536,7168,8,256,16,balanced,3.4573696136474608
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,65536,7168,8,256,8,balanced,0.43783040046691896
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,65536,7168,8,256,8,balanced,0.46284799873828886
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,65536,7168,8,256,8,balanced,3.328857588768005
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,65536,7168,8,256,8,balanced,3.036992001533508
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,32768,7168,8,256,48,balanced,0.254236800968647
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,32768,7168,8,256,48,balanced,0.7579935967922211
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,32768,7168,8,256,48,balanced,3.393372821807861
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,32768,7168,8,256,48,balanced,2.7391775846481323
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,65536,7168,8,256,16,balanced,0.43735680282115935
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,65536,7168,8,256,16,balanced,0.5450751960277558
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,65536,7168,8,256,16,balanced,3.457606387138367
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,65536,7168,8,256,16,balanced,3.049785614013672
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,65536,7168,8,256,48,balanced,0.46021759808063506
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,65536,7168,8,256,48,balanced,4.812975978851318
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,65536,7168,8,256,48,balanced,6.777001619338989
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,65536,7168,8,256,48,balanced,0.4573248028755188
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,65536,7168,8,256,48,balanced,2.506876802444458
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,65536,7168,8,256,48,balanced,6.781046390533447
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,65536,7168,8,256,48,balanced,0.455049604177475
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,65536,7168,8,256,48,balanced,1.4930591940879823
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,65536,7168,8,256,48,balanced,6.775823974609375
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,65536,7168,8,256,48,balanced,5.3131519794464115
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1,7168,8,256,32,balanced,0.07150400057435036
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1,7168,8,256,32,balanced,0.05531200021505356
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1,7168,8,256,32,balanced,0.07301439866423606
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1,7168,8,256,32,balanced,0.0683008000254631
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1,7168,8,256,32,balanced,0.05923840031027794
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1,7168,8,256,32,balanced,0.069548799097538
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1,7168,8,256,32,balanced,0.06910719983279705
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1,7168,8,256,32,balanced,0.057919999584555626
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1,7168,8,256,32,balanced,0.07564800009131431
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1,7168,8,256,32,balanced,0.06807680055499077
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,2,7168,8,256,32,balanced,0.06489919982850552
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,2,7168,8,256,32,balanced,0.057465600222349165
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,2,7168,8,256,32,balanced,0.0702239990234375
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,2,7168,8,256,32,balanced,0.07419199980795384
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,2,7168,8,256,32,balanced,0.060678400099277496
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,2,7168,8,256,32,balanced,0.07163199856877327
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,2,7168,8,256,32,balanced,0.069225599989295
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,2,7168,8,256,32,balanced,0.05885759964585304
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,2,7168,8,256,32,balanced,0.06924160048365594
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,2,7168,8,256,32,balanced,0.06623679921030998
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,4,7168,8,256,32,balanced,0.06423999965190888
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,4,7168,8,256,32,balanced,0.07721599973738194
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,4,7168,8,256,32,balanced,0.07455039992928505
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,4,7168,8,256,32,balanced,0.0706591997295618
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,4,7168,8,256,32,balanced,0.05826879926025867
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,4,7168,8,256,32,balanced,0.08079679980874062
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,4,7168,8,256,32,balanced,0.06956159882247448
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,4,7168,8,256,32,balanced,0.06084479987621307
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,4,7168,8,256,32,balanced,0.07713280022144317
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,4,7168,8,256,32,balanced,0.07743040025234223
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,8,7168,8,256,32,balanced,0.07323519997298718
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,8,7168,8,256,32,balanced,0.06041279956698418
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,8,7168,8,256,32,balanced,0.07749759927392005
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,8,7168,8,256,32,balanced,0.06544959992170334
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,8,7168,8,256,32,balanced,0.05813120044767857
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,8,7168,8,256,32,balanced,0.0739936001598835
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,8,7168,8,256,32,balanced,0.07265280075371265
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,8,7168,8,256,32,balanced,0.060950399935245515
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,8,7168,8,256,32,balanced,0.07416319996118545
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,8,7168,8,256,32,balanced,0.07835199981927872
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,16,7168,8,256,32,balanced,0.07224959954619407
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,16,7168,8,256,32,balanced,0.061471999064087866
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,16,7168,8,256,32,balanced,0.07855039983987808
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,16,7168,8,256,32,balanced,0.07801920026540757
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,16,7168,8,256,32,balanced,0.06015040054917335
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,16,7168,8,256,32,balanced,0.07992640063166619
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,16,7168,8,256,32,balanced,0.07264639995992184
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,16,7168,8,256,32,balanced,0.06069440022110939
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,16,7168,8,256,32,balanced,0.07782080173492431
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,16,7168,8,256,32,balanced,0.08473280072212219
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,32,7168,8,256,32,balanced,0.07095359936356545
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,32,7168,8,256,32,balanced,0.07212159931659698
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,32,7168,8,256,32,balanced,0.09491840153932571
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,32,7168,8,256,32,balanced,0.06551359929144382
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,32,7168,8,256,32,balanced,0.0579328004270792
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,32,7168,8,256,32,balanced,0.10178879946470261
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,32,7168,8,256,32,balanced,0.07714880034327506
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,32,7168,8,256,32,balanced,0.06097280010581017
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,32,7168,8,256,32,balanced,0.09622079953551292
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,32,7168,8,256,32,balanced,0.10061440169811249
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,48,7168,8,256,32,balanced,0.06893440000712872
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,48,7168,8,256,32,balanced,0.065251200273633
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,48,7168,8,256,32,balanced,0.10156799927353859
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,48,7168,8,256,32,balanced,0.06944319978356361
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,48,7168,8,256,32,balanced,0.058185600116848944
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,48,7168,8,256,32,balanced,0.09697920009493828
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,48,7168,8,256,32,balanced,0.06934079863131046
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,48,7168,8,256,32,balanced,0.06678080074489116
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,48,7168,8,256,32,balanced,0.09722559973597526
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,48,7168,8,256,32,balanced,0.088179200142622
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,64,7168,8,256,32,balanced,0.06862399950623513
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,64,7168,8,256,32,balanced,0.06616320013999939
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,64,7168,8,256,32,balanced,0.10404480025172233
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,64,7168,8,256,32,balanced,0.07143039964139461
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,64,7168,8,256,32,balanced,0.08100480064749718
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,64,7168,8,256,32,balanced,0.0996576003730297
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,64,7168,8,256,32,balanced,0.06807680018246173
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,64,7168,8,256,32,balanced,0.06150719970464706
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,64,7168,8,256,32,balanced,0.09657280072569847
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,64,7168,8,256,32,balanced,0.09503680169582367
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,80,7168,8,256,32,balanced,0.07370559982955456
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,80,7168,8,256,32,balanced,0.06150080002844334
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,80,7168,8,256,32,balanced,0.08184320032596588
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,80,7168,8,256,32,balanced,0.06574079915881156
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,80,7168,8,256,32,balanced,0.05441919974982738
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,80,7168,8,256,32,balanced,0.0831712007522583
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,80,7168,8,256,32,balanced,0.07608960121870041
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,80,7168,8,256,32,balanced,0.05175360068678856
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,80,7168,8,256,32,balanced,0.0830880008637905
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,80,7168,8,256,32,balanced,0.07849279940128326
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,96,7168,8,256,32,balanced,0.05901760086417198
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,96,7168,8,256,32,balanced,0.05777600035071373
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,96,7168,8,256,32,balanced,0.08047999963164329
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,96,7168,8,256,32,balanced,0.06451520062983036
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,96,7168,8,256,32,balanced,0.05491200014948845
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,96,7168,8,256,32,balanced,0.08336000069975853
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,96,7168,8,256,32,balanced,0.06498239934444427
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,96,7168,8,256,32,balanced,0.05356800071895122
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,96,7168,8,256,32,balanced,0.083839999884367
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,96,7168,8,256,32,balanced,0.07944639921188354
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,128,7168,8,256,32,balanced,0.05998400002717972
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,128,7168,8,256,32,balanced,0.07086079865694046
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,128,7168,8,256,32,balanced,0.10037440061569214
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,128,7168,8,256,32,balanced,0.06689600013196469
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,128,7168,8,256,32,balanced,0.0626848004758358
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,128,7168,8,256,32,balanced,0.09220480024814606
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,128,7168,8,256,32,balanced,0.06965440064668656
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,128,7168,8,256,32,balanced,0.06148800104856491
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,128,7168,8,256,32,balanced,0.09548800066113472
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,128,7168,8,256,32,balanced,0.09302719980478287
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,160,7168,8,256,32,balanced,0.07299520038068294
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,160,7168,8,256,32,balanced,0.057558399066329
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,160,7168,8,256,32,balanced,0.08476799950003625
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,160,7168,8,256,32,balanced,0.06611199975013733
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,160,7168,8,256,32,balanced,0.052537599578499794
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,160,7168,8,256,32,balanced,0.08020159974694252
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,160,7168,8,256,32,balanced,0.06325759999454021
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,160,7168,8,256,32,balanced,0.05376639999449253
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,160,7168,8,256,32,balanced,0.08467520028352737
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,160,7168,8,256,32,balanced,0.08602880090475082
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,192,7168,8,256,32,balanced,0.05947839953005314
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,192,7168,8,256,32,balanced,0.06933439932763577
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,192,7168,8,256,32,balanced,0.08841920047998428
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,192,7168,8,256,32,balanced,0.06615360043942928
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,192,7168,8,256,32,balanced,0.06224639974534511
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,192,7168,8,256,32,balanced,0.09452160000801087
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,192,7168,8,256,32,balanced,0.07272320017218589
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,192,7168,8,256,32,balanced,0.0617407999932766
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,192,7168,8,256,32,balanced,0.09969279989600181
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,192,7168,8,256,32,balanced,0.09341439977288246
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,256,7168,8,256,32,balanced,0.07373440004885197
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,256,7168,8,256,32,balanced,0.060259200260043146
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,256,7168,8,256,32,balanced,0.08543040007352828
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,256,7168,8,256,32,balanced,0.06720320023596286
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,256,7168,8,256,32,balanced,0.057295999675989154
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,256,7168,8,256,32,balanced,0.08081599920988083
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,256,7168,8,256,32,balanced,0.06276800036430359
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,256,7168,8,256,32,balanced,0.05427199937403202
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,256,7168,8,256,32,balanced,0.08158399984240532
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,256,7168,8,256,32,balanced,0.08345280066132546
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,320,7168,8,256,32,balanced,0.06807359978556633
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,320,7168,8,256,32,balanced,0.06953280083835126
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,320,7168,8,256,32,balanced,0.0806207999587059
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,320,7168,8,256,32,balanced,0.0701695989817381
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,320,7168,8,256,32,balanced,0.059577600285410884
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,320,7168,8,256,32,balanced,0.08322559967637062
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,320,7168,8,256,32,balanced,0.06445119976997375
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,320,7168,8,256,32,balanced,0.057411200180649755
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,320,7168,8,256,32,balanced,0.08559679985046387
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,320,7168,8,256,32,balanced,0.091219200193882
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,384,7168,8,256,32,balanced,0.06923840008676052
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,384,7168,8,256,32,balanced,0.0712032001465559
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,384,7168,8,256,32,balanced,0.0958560012280941
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,384,7168,8,256,32,balanced,0.07176319994032383
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,384,7168,8,256,32,balanced,0.06331519968807697
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,384,7168,8,256,32,balanced,0.10141440033912659
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,384,7168,8,256,32,balanced,0.0734592005610466
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,384,7168,8,256,32,balanced,0.06162239983677864
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,384,7168,8,256,32,balanced,0.09548480063676834
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,384,7168,8,256,32,balanced,0.09447679966688156
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,512,7168,8,256,32,balanced,0.06981439925730229
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,512,7168,8,256,32,balanced,0.06870079971849918
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,512,7168,8,256,32,balanced,0.09489280059933662
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,512,7168,8,256,32,balanced,0.07036800123751163
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,512,7168,8,256,32,balanced,0.060204800218343735
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,512,7168,8,256,32,balanced,0.09075200036168099
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,512,7168,8,256,32,balanced,0.07200639955699444
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,512,7168,8,256,32,balanced,0.057971199974417686
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,512,7168,8,256,32,balanced,0.09178880080580712
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,512,7168,8,256,32,balanced,0.09832959920167923
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,768,7168,8,256,32,balanced,0.06895040050148964
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,768,7168,8,256,32,balanced,0.07100160010159015
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,768,7168,8,256,32,balanced,0.09386880025267601
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,768,7168,8,256,32,balanced,0.06538239978253842
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,768,7168,8,256,32,balanced,0.05772479996085167
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,768,7168,8,256,32,balanced,0.09652799963951111
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,768,7168,8,256,32,balanced,0.06784639954566955
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,768,7168,8,256,32,balanced,0.058911999315023424
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,768,7168,8,256,32,balanced,0.09530239999294281
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,768,7168,8,256,32,balanced,0.09423680007457733
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1024,7168,8,256,32,balanced,0.0683231994509697
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1024,7168,8,256,32,balanced,0.07945920005440713
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1024,7168,8,256,32,balanced,0.1102207988500595
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1024,7168,8,256,32,balanced,0.06736640073359013
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1024,7168,8,256,32,balanced,0.05835199989378452
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1024,7168,8,256,32,balanced,0.11433919966220855
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1024,7168,8,256,32,balanced,0.06403840072453022
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1024,7168,8,256,32,balanced,0.061184000223875046
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1024,7168,8,256,32,balanced,0.11005119904875756
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1024,7168,8,256,32,balanced,0.10863680019974709
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1536,7168,8,256,32,balanced,0.06622719950973988
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1536,7168,8,256,32,balanced,0.09383040145039559
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1536,7168,8,256,32,balanced,0.14136960059404374
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1536,7168,8,256,32,balanced,0.07188480086624623
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1536,7168,8,256,32,balanced,0.06613119952380657
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1536,7168,8,256,32,balanced,0.14381760060787202
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1536,7168,8,256,32,balanced,0.06896959953010082
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1536,7168,8,256,32,balanced,0.0642815999686718
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1536,7168,8,256,32,balanced,0.1401760011911392
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1536,7168,8,256,32,balanced,0.1358367994427681
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,2048,7168,8,256,32,balanced,0.07132480032742024
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,2048,7168,8,256,32,balanced,0.10260480046272277
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,2048,7168,8,256,32,balanced,0.170694400370121
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,2048,7168,8,256,32,balanced,0.07588480003178119
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,2048,7168,8,256,32,balanced,0.07573440000414848
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,2048,7168,8,256,32,balanced,0.17090560048818587
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,2048,7168,8,256,32,balanced,0.0770528007298708
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,2048,7168,8,256,32,balanced,0.07189119979739189
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,2048,7168,8,256,32,balanced,0.16667520254850388
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,2048,7168,8,256,32,balanced,0.16370559781789779
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,3072,7168,8,256,32,balanced,0.08114880099892616
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,3072,7168,8,256,32,balanced,0.13941440135240554
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,3072,7168,8,256,32,balanced,0.22373440116643906
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,3072,7168,8,256,32,balanced,0.08042239919304847
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,3072,7168,8,256,32,balanced,0.08857280164957046
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,3072,7168,8,256,32,balanced,0.22043840140104293
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,3072,7168,8,256,32,balanced,0.07763839960098266
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,3072,7168,8,256,32,balanced,0.07590719908475876
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,3072,7168,8,256,32,balanced,0.22509759962558745
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,3072,7168,8,256,32,balanced,0.21047039926052094
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,4096,7168,8,256,32,balanced,0.0863680012524128
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,4096,7168,8,256,32,balanced,0.15276799947023392
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,4096,7168,8,256,32,balanced,0.2772256016731262
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,4096,7168,8,256,32,balanced,0.08116800040006637
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,4096,7168,8,256,32,balanced,0.09961920082569123
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,4096,7168,8,256,32,balanced,0.27314239740371704
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,4096,7168,8,256,32,balanced,0.0825215995311737
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,4096,7168,8,256,32,balanced,0.0841056004166603
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,4096,7168,8,256,32,balanced,0.27580160200595855
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,4096,7168,8,256,32,balanced,0.2548031985759735
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,6144,7168,8,256,32,balanced,0.0950367994606495
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,6144,7168,8,256,32,balanced,0.20294079929590225
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,6144,7168,8,256,32,balanced,0.38109439611434937
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,6144,7168,8,256,32,balanced,0.09585599899291992
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,6144,7168,8,256,32,balanced,0.124252799898386
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,6144,7168,8,256,32,balanced,0.3825504004955292
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,6144,7168,8,256,32,balanced,0.10051200017333031
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,6144,7168,8,256,32,balanced,0.09784639999270439
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,6144,7168,8,256,32,balanced,0.38602879643440247
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,6144,7168,8,256,32,balanced,0.35638400316238406
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,8192,7168,8,256,32,balanced,0.10596480071544648
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,8192,7168,8,256,32,balanced,0.24371520280838013
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,8192,7168,8,256,32,balanced,0.4881087988615036
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,8192,7168,8,256,32,balanced,0.10742719992995262
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,8192,7168,8,256,32,balanced,0.15081599801778794
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,8192,7168,8,256,32,balanced,0.48908800184726714
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,8192,7168,8,256,32,balanced,0.10570560097694397
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,8192,7168,8,256,32,balanced,0.1136064000427723
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,8192,7168,8,256,32,balanced,0.4890848010778427
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,8192,7168,8,256,32,balanced,0.4434719979763031
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,12288,7168,8,256,32,balanced,0.12815360054373742
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,12288,7168,8,256,32,balanced,0.34912959933280946
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,12288,7168,8,256,32,balanced,0.70384960770607
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,12288,7168,8,256,32,balanced,0.13323839902877807
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,12288,7168,8,256,32,balanced,0.20284159928560258
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,12288,7168,8,256,32,balanced,0.7053791999816894
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,12288,7168,8,256,32,balanced,0.13108799904584884
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,12288,7168,8,256,32,balanced,0.14755520075559617
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,12288,7168,8,256,32,balanced,0.7041376054286956
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,12288,7168,8,256,32,balanced,0.6561119973659515
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,16384,7168,8,256,32,balanced,0.15483199805021286
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,16384,7168,8,256,32,balanced,0.44043839871883395
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,16384,7168,8,256,32,balanced,0.9212383925914764
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,16384,7168,8,256,32,balanced,0.1539327993988991
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,16384,7168,8,256,32,balanced,0.2540415972471237
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,16384,7168,8,256,32,balanced,0.9214399993419647
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,16384,7168,8,256,32,balanced,0.15275200009346007
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,16384,7168,8,256,32,balanced,0.18084160089492798
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,16384,7168,8,256,32,balanced,0.9237567961215973
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,16384,7168,8,256,32,balanced,0.8219071984291076
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,20480,7168,8,256,32,balanced,0.17803199887275695
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,20480,7168,8,256,32,balanced,0.5372255980968476
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,20480,7168,8,256,32,balanced,1.1487904071807862
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,20480,7168,8,256,32,balanced,0.17546239495277405
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,20480,7168,8,256,32,balanced,0.3040864050388336
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,20480,7168,8,256,32,balanced,1.1459743976593018
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,20480,7168,8,256,32,balanced,0.1777312010526657
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,20480,7168,8,256,32,balanced,0.21210239827632904
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,20480,7168,8,256,32,balanced,1.1458847880363465
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,20480,7168,8,256,32,balanced,1.0222464203834534
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,32768,7168,8,256,32,balanced,0.2506687998771667
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,32768,7168,8,256,32,balanced,0.8376031935214996
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,32768,7168,8,256,32,balanced,1.801155197620392
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,32768,7168,8,256,32,balanced,0.2501824006438255
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,32768,7168,8,256,32,balanced,0.4581919968128204
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,32768,7168,8,256,32,balanced,1.7969663739204407
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,32768,7168,8,256,32,balanced,0.24853439927101134
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,32768,7168,8,256,32,balanced,0.3163807958364487
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,32768,7168,8,256,32,balanced,1.797974419593811
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,32768,7168,8,256,32,balanced,1.602230393886566
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,65536,7168,8,256,32,balanced,0.44158399999141695
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,65536,7168,8,256,32,balanced,1.6295615911483765
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,65536,7168,8,256,32,balanced,3.555836820602417
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,65536,7168,8,256,32,balanced,0.44465280175209043
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,65536,7168,8,256,32,balanced,0.8766623914241791
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,65536,7168,8,256,32,balanced,3.552201581001282
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,65536,7168,8,256,32,balanced,0.4453023999929428
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,65536,7168,8,256,32,balanced,0.5798207938671112
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,65536,7168,8,256,32,balanced,3.5528448104858397
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,65536,7168,8,256,32,balanced,3.0999840021133425
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1,7168,8,256,72,balanced,0.07495680004358292
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1,7168,8,256,72,balanced,0.05985600017011165
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1,7168,8,256,72,balanced,0.08000319972634315
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1,7168,8,256,72,balanced,0.08125440068542958
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1,7168,8,256,72,balanced,0.05274879969656467
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1,7168,8,256,72,balanced,0.07388159856200219
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1,7168,8,256,72,balanced,0.06488640047609806
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1,7168,8,256,72,balanced,0.05382079966366291
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1,7168,8,256,72,balanced,0.07988800033926964
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1,7168,8,256,72,balanced,0.07659519985318183
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,2,7168,8,256,72,balanced,0.08556480146944523
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,2,7168,8,256,72,balanced,0.05203839987516403
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,2,7168,8,256,72,balanced,0.07504640072584152
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,2,7168,8,256,72,balanced,0.061289599165320396
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,2,7168,8,256,72,balanced,0.05001279972493648
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,2,7168,8,256,72,balanced,0.07567999958992004
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,2,7168,8,256,72,balanced,0.06244799979031086
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,2,7168,8,256,72,balanced,0.054022399708628654
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,2,7168,8,256,72,balanced,0.07738239988684655
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,2,7168,8,256,72,balanced,0.07655359879136085
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,4,7168,8,256,72,balanced,0.06294719986617565
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,4,7168,8,256,72,balanced,0.051695999875664714
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,4,7168,8,256,72,balanced,0.0762080006301403
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,4,7168,8,256,72,balanced,0.07146560028195381
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,4,7168,8,256,72,balanced,0.05708480030298233
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,4,7168,8,256,72,balanced,0.07739520072937012
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,4,7168,8,256,72,balanced,0.06636160053312778
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,4,7168,8,256,72,balanced,0.056809600070118906
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,4,7168,8,256,72,balanced,0.075331199914217
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,4,7168,8,256,72,balanced,0.07676480039954185
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,8,7168,8,256,72,balanced,0.06367679983377457
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,8,7168,8,256,72,balanced,0.053987200185656545
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,8,7168,8,256,72,balanced,0.07303680032491684
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,8,7168,8,256,72,balanced,0.06316479928791523
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,8,7168,8,256,72,balanced,0.051311999931931494
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,8,7168,8,256,72,balanced,0.07441600039601326
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,8,7168,8,256,72,balanced,0.059436799958348274
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,8,7168,8,256,72,balanced,0.05331200025975704
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,8,7168,8,256,72,balanced,0.0722751997411251
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,8,7168,8,256,72,balanced,0.07373760044574737
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,16,7168,8,256,72,balanced,0.0662592001259327
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,16,7168,8,256,72,balanced,0.08687360025942326
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,16,7168,8,256,72,balanced,0.10936319977045059
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,16,7168,8,256,72,balanced,0.07909440100193024
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,16,7168,8,256,72,balanced,0.06604480110108853
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,16,7168,8,256,72,balanced,0.10366399958729744
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,16,7168,8,256,72,balanced,0.07076799906790257
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,16,7168,8,256,72,balanced,0.062409599870443346
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,16,7168,8,256,72,balanced,0.1051264002919197
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,16,7168,8,256,72,balanced,0.10011840090155602
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,32,7168,8,256,72,balanced,0.0708383996039629
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,32,7168,8,256,72,balanced,0.07765760011970997
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,32,7168,8,256,72,balanced,0.10604480057954788
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,32,7168,8,256,72,balanced,0.06856959946453571
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,32,7168,8,256,72,balanced,0.06301119923591614
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,32,7168,8,256,72,balanced,0.1041472002863884
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,32,7168,8,256,72,balanced,0.07302080057561397
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,32,7168,8,256,72,balanced,0.0619488000869751
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,32,7168,8,256,72,balanced,0.10630400031805039
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,32,7168,8,256,72,balanced,0.09749760031700135
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,48,7168,8,256,72,balanced,0.0668928012251854
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,48,7168,8,256,72,balanced,0.06515840031206607
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,48,7168,8,256,72,balanced,0.10080319941043854
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,48,7168,8,256,72,balanced,0.06529919989407063
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,48,7168,8,256,72,balanced,0.05808319933712482
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,48,7168,8,256,72,balanced,0.10263359993696212
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,48,7168,8,256,72,balanced,0.06246079988777638
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,48,7168,8,256,72,balanced,0.05788800045847893
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,48,7168,8,256,72,balanced,0.09670399948954582
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,48,7168,8,256,72,balanced,0.10162240043282508
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,64,7168,8,256,72,balanced,0.06844480074942112
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,64,7168,8,256,72,balanced,0.0687647994607687
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,64,7168,8,256,72,balanced,0.09682240188121796
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,64,7168,8,256,72,balanced,0.0679712001234293
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,64,7168,8,256,72,balanced,0.060518399626016614
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,64,7168,8,256,72,balanced,0.10340799987316132
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,64,7168,8,256,72,balanced,0.07125759944319725
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,64,7168,8,256,72,balanced,0.054998400062322615
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,64,7168,8,256,72,balanced,0.09545919969677925
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,64,7168,8,256,72,balanced,0.09336320087313651
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,80,7168,8,256,72,balanced,0.06429440006613732
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,80,7168,8,256,72,balanced,0.06592960059642791
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,80,7168,8,256,72,balanced,0.09228480011224746
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1,7168,8,256,64,balanced,0.07362239956855773
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,80,7168,8,256,72,balanced,0.06399999968707562
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1,7168,8,256,64,balanced,0.06250880099833012
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1,7168,8,256,64,balanced,0.07323199957609176
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1,7168,8,256,64,balanced,0.07394879907369614
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1,7168,8,256,64,balanced,0.058339199796319005
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1,7168,8,256,64,balanced,0.06731199994683265
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1,7168,8,256,64,balanced,0.07095040008425713
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1,7168,8,256,64,balanced,0.05743679963052273
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1,7168,8,256,64,balanced,0.07657599970698356
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1,7168,8,256,64,balanced,0.07884479984641075
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,2,7168,8,256,64,balanced,0.08544960059225559
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,2,7168,8,256,64,balanced,0.06424640007317066
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,2,7168,8,256,64,balanced,0.0771199993789196
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,2,7168,8,256,64,balanced,0.07644479870796203
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,2,7168,8,256,64,balanced,0.06975040100514888
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,2,7168,8,256,64,balanced,0.07320640012621879
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,2,7168,8,256,64,balanced,0.07410559952259063
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,2,7168,8,256,64,balanced,0.06752320006489754
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,2,7168,8,256,64,balanced,0.0737472005188465
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,2,7168,8,256,64,balanced,0.07547840103507042
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,4,7168,8,256,64,balanced,0.07566720023751258
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,4,7168,8,256,64,balanced,0.05992319993674755
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,4,7168,8,256,64,balanced,0.07552640065550804
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,4,7168,8,256,64,balanced,0.06803520135581494
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,4,7168,8,256,64,balanced,0.05948479995131493
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,4,7168,8,256,64,balanced,0.07481279894709587
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,4,7168,8,256,64,balanced,0.07022079899907112
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,4,7168,8,256,64,balanced,0.07159680016338825
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,4,7168,8,256,64,balanced,0.08307520076632499
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,4,7168,8,256,64,balanced,0.08212159872055054
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,8,7168,8,256,64,balanced,0.0762719988822937
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,8,7168,8,256,64,balanced,0.06403520032763481
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,8,7168,8,256,64,balanced,0.0871456004679203
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,8,7168,8,256,64,balanced,0.07244159989058971
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,8,7168,8,256,64,balanced,0.06257599964737892
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,8,7168,8,256,64,balanced,0.08235200047492981
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,8,7168,8,256,64,balanced,0.06887999922037125
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,8,7168,8,256,64,balanced,0.06083840020000934
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,8,7168,8,256,64,balanced,0.08045760020613671
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,8,7168,8,256,64,balanced,0.07596480026841164
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,16,7168,8,256,64,balanced,0.06989760026335716
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,16,7168,8,256,64,balanced,0.06269760020077228
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,16,7168,8,256,64,balanced,0.08406079933047295
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,16,7168,8,256,64,balanced,0.06842240020632744
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,16,7168,8,256,64,balanced,0.05971519947052002
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,16,7168,8,256,64,balanced,0.08063039928674698
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,16,7168,8,256,64,balanced,0.06673280000686646
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,16,7168,8,256,64,balanced,0.05659519992768765
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,16,7168,8,256,64,balanced,0.07308799996972085
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,16,7168,8,256,64,balanced,0.07669439986348152
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,32,7168,8,256,64,balanced,0.06770559921860694
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,32,7168,8,256,64,balanced,0.06584640070796013
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,32,7168,8,256,64,balanced,0.08872319981455803
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,32,7168,8,256,64,balanced,0.07475519999861717
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,32,7168,8,256,64,balanced,0.06363840028643608
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,32,7168,8,256,64,balanced,0.0877920001745224
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,32,7168,8,256,64,balanced,0.07236480042338371
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,32,7168,8,256,64,balanced,0.06122879981994629
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,32,7168,8,256,64,balanced,0.08609280064702034
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,32,7168,8,256,64,balanced,0.08222399950027466
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,48,7168,8,256,64,balanced,0.07021760009229183
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,48,7168,8,256,64,balanced,0.06192639991641045
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,48,7168,8,256,64,balanced,0.08258240073919296
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,48,7168,8,256,64,balanced,0.06456639990210533
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,48,7168,8,256,64,balanced,0.05634559988975525
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,48,7168,8,256,64,balanced,0.08087040111422539
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,48,7168,8,256,64,balanced,0.062217600643634796
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,48,7168,8,256,64,balanced,0.0541279997676611
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,48,7168,8,256,64,balanced,0.07800320014357567
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,48,7168,8,256,64,balanced,0.0814336009323597
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,64,7168,8,256,64,balanced,0.06799359992146492
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,64,7168,8,256,64,balanced,0.0617407999932766
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,64,7168,8,256,64,balanced,0.0790367990732193
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,64,7168,8,256,64,balanced,0.06228799894452095
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,64,7168,8,256,64,balanced,0.056838399916887286
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,64,7168,8,256,64,balanced,0.0774496003985405
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,64,7168,8,256,64,balanced,0.05780480057001114
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,64,7168,8,256,64,balanced,0.05219840034842491
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,64,7168,8,256,64,balanced,0.07714559957385063
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,64,7168,8,256,64,balanced,0.07842879965901375
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,80,7168,8,256,64,balanced,0.05965759977698326
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,80,7168,8,256,64,balanced,0.060179199650883676
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,80,7168,8,256,64,balanced,0.07668480053544044
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,80,7168,8,256,64,balanced,0.06141119934618473
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,80,7168,8,256,64,balanced,0.054316799715161325
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,80,7168,8,256,64,balanced,0.07564479932188987
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,80,7168,8,256,64,balanced,0.06008319966495037
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,80,7168,8,256,64,balanced,0.054467199742794035
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,80,7168,8,256,64,balanced,0.07814399972558021
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,80,7168,8,256,64,balanced,0.0751808002591133
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,96,7168,8,256,64,balanced,0.06403200030326843
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,96,7168,8,256,64,balanced,0.06216639988124371
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,96,7168,8,256,64,balanced,0.07611520066857339
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,96,7168,8,256,64,balanced,0.07936319932341576
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,96,7168,8,256,64,balanced,0.053654399886727336
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,96,7168,8,256,64,balanced,0.0785728007555008
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,96,7168,8,256,64,balanced,0.05910400003194809
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,96,7168,8,256,64,balanced,0.05049280002713204
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,96,7168,8,256,64,balanced,0.07925760075449943
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,96,7168,8,256,64,balanced,0.07854079902172088
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,128,7168,8,256,64,balanced,0.0615423996001482
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,128,7168,8,256,64,balanced,0.06011200025677681
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,128,7168,8,256,64,balanced,0.07877440005540848
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,128,7168,8,256,64,balanced,0.06120319999754429
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,128,7168,8,256,64,balanced,0.05590400025248528
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,128,7168,8,256,64,balanced,0.07611520066857339
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,128,7168,8,256,64,balanced,0.06388160064816475
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,128,7168,8,256,64,balanced,0.05653440020978451
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,128,7168,8,256,64,balanced,0.07434880062937736
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,128,7168,8,256,64,balanced,0.07859519943594932
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,160,7168,8,256,64,balanced,0.06408640034496785
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,160,7168,8,256,64,balanced,0.06384959928691387
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,160,7168,8,256,64,balanced,0.08115199953317642
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,160,7168,8,256,64,balanced,0.06022719964385033
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,160,7168,8,256,64,balanced,0.058956800028681755
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,160,7168,8,256,64,balanced,0.07575040087103843
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,160,7168,8,256,64,balanced,0.06364160031080246
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,160,7168,8,256,64,balanced,0.05349440015852451
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,160,7168,8,256,64,balanced,0.07967039942741394
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,160,7168,8,256,64,balanced,0.07907840013504028
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,192,7168,8,256,64,balanced,0.06072000004351139
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,192,7168,8,256,64,balanced,0.06207359917461872
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,192,7168,8,256,64,balanced,0.07676159963011742
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,192,7168,8,256,64,balanced,0.06055040024220944
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,192,7168,8,256,64,balanced,0.05575999990105629
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,192,7168,8,256,64,balanced,0.0781088002026081
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,192,7168,8,256,64,balanced,0.060719999670982364
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,192,7168,8,256,64,balanced,0.05318399891257286
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,192,7168,8,256,64,balanced,0.07451519966125489
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,192,7168,8,256,64,balanced,0.07756799906492233
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,256,7168,8,256,64,balanced,0.05812479965388775
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,256,7168,8,256,64,balanced,0.06498240083456039
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,256,7168,8,256,64,balanced,0.07739199995994568
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,256,7168,8,256,64,balanced,0.06424320116639137
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,256,7168,8,256,64,balanced,0.06178239993751049
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,256,7168,8,256,64,balanced,0.08007360100746155
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,256,7168,8,256,64,balanced,0.05808000005781651
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,256,7168,8,256,64,balanced,0.051068799942731856
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,256,7168,8,256,64,balanced,0.0829087994992733
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,256,7168,8,256,64,balanced,0.07230719923973083
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,80,7168,8,256,72,balanced,0.05774400047957897
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,80,7168,8,256,72,balanced,0.10090880021452904
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,80,7168,8,256,72,balanced,0.0663904007524252
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,80,7168,8,256,72,balanced,0.056332800164818765
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,80,7168,8,256,72,balanced,0.10298239961266517
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,80,7168,8,256,72,balanced,0.09464319944381713
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,320,7168,8,256,64,balanced,0.06004480011761189
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,320,7168,8,256,64,balanced,0.0695519994944334
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,320,7168,8,256,64,balanced,0.08124159947037697
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,96,7168,8,256,72,balanced,0.06310720033943654
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,96,7168,8,256,72,balanced,0.0701919998973608
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,96,7168,8,256,72,balanced,0.10590720102190972
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,96,7168,8,256,72,balanced,0.06855040080845357
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,96,7168,8,256,72,balanced,0.05790399983525276
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,96,7168,8,256,72,balanced,0.10332800000905991
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,320,7168,8,256,64,balanced,0.05897600017488003
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,320,7168,8,256,64,balanced,0.05565760061144829
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,320,7168,8,256,64,balanced,0.07909440025687217
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,96,7168,8,256,72,balanced,0.06502400003373623
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,96,7168,8,256,72,balanced,0.057699200510978696
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,96,7168,8,256,72,balanced,0.10259519964456558
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,96,7168,8,256,72,balanced,0.08800319954752922
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,128,7168,8,256,72,balanced,0.06405440010130406
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,128,7168,8,256,72,balanced,0.06956159994006157
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,128,7168,8,256,72,balanced,0.10025920197367669
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,320,7168,8,256,64,balanced,0.06325440034270287
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,320,7168,8,256,64,balanced,0.05609279982745648
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,320,7168,8,256,64,balanced,0.08174080029129982
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,320,7168,8,256,64,balanced,0.07986239865422248
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,128,7168,8,256,72,balanced,0.0670304000377655
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,128,7168,8,256,72,balanced,0.05441920012235642
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,128,7168,8,256,72,balanced,0.0997696004807949
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,128,7168,8,256,72,balanced,0.06638080030679702
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,128,7168,8,256,72,balanced,0.06021440029144287
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,128,7168,8,256,72,balanced,0.09492480084300041
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,128,7168,8,256,72,balanced,0.1000351995229721
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,384,7168,8,256,64,balanced,0.06029439941048622
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,384,7168,8,256,64,balanced,0.07150400057435036
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,384,7168,8,256,64,balanced,0.08334400057792664
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,160,7168,8,256,72,balanced,0.06318720057606697
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,160,7168,8,256,72,balanced,0.06634239926934242
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,160,7168,8,256,72,balanced,0.10111999958753586
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,160,7168,8,256,72,balanced,0.060182399675250055
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,160,7168,8,256,72,balanced,0.056780800223350525
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,160,7168,8,256,72,balanced,0.09602880030870438
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,384,7168,8,256,64,balanced,0.05979200080037117
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,384,7168,8,256,64,balanced,0.05804160088300705
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,384,7168,8,256,64,balanced,0.08392639979720115
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,160,7168,8,256,72,balanced,0.0656224001199007
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,160,7168,8,256,72,balanced,0.05882559977471828
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,160,7168,8,256,72,balanced,0.09992320016026497
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,160,7168,8,256,72,balanced,0.09520640075206757
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,384,7168,8,256,64,balanced,0.060092800110578534
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,384,7168,8,256,64,balanced,0.05684479922056198
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,384,7168,8,256,64,balanced,0.08112640008330345
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,192,7168,8,256,72,balanced,0.06327360086143016
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,384,7168,8,256,64,balanced,0.08180800005793572
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,192,7168,8,256,72,balanced,0.0666848000138998
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,192,7168,8,256,72,balanced,0.10085439905524254
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,192,7168,8,256,72,balanced,0.06738239973783493
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,192,7168,8,256,72,balanced,0.05882879979908466
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,192,7168,8,256,72,balanced,0.09984960108995437
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,192,7168,8,256,72,balanced,0.06489600129425525
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,512,7168,8,256,64,balanced,0.061977599188685416
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,192,7168,8,256,72,balanced,0.05879360027611256
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,512,7168,8,256,64,balanced,0.07776319831609727
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,192,7168,8,256,72,balanced,0.09877440109848976
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,512,7168,8,256,64,balanced,0.0957023985683918
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,192,7168,8,256,72,balanced,0.09080640003085136
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,256,7168,8,256,72,balanced,0.0760479986667633
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,256,7168,8,256,72,balanced,0.07103039994835854
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,256,7168,8,256,72,balanced,0.10829439908266067
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,512,7168,8,256,64,balanced,0.0602400004863739
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,512,7168,8,256,64,balanced,0.059068800136446956
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,512,7168,8,256,64,balanced,0.09650879874825477
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,256,7168,8,256,72,balanced,0.07212800011038781
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,256,7168,8,256,72,balanced,0.06500160060822964
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,256,7168,8,256,72,balanced,0.10529919937252999
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,256,7168,8,256,72,balanced,0.06852800026535988
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,256,7168,8,256,72,balanced,0.06465919986367226
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,256,7168,8,256,72,balanced,0.10379839986562729
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,256,7168,8,256,72,balanced,0.09790719971060753
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,512,7168,8,256,64,balanced,0.06139839962124825
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,512,7168,8,256,64,balanced,0.057072000578045845
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,512,7168,8,256,64,balanced,0.09455680027604103
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,512,7168,8,256,64,balanced,0.09124799966812133
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,320,7168,8,256,72,balanced,0.07206079922616482
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,320,7168,8,256,72,balanced,0.07920959889888764
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,320,7168,8,256,72,balanced,0.10440320000052453
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,320,7168,8,256,72,balanced,0.06920639984309673
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,320,7168,8,256,72,balanced,0.05926399976015091
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,320,7168,8,256,72,balanced,0.10087680071592331
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,768,7168,8,256,64,balanced,0.06242559961974621
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,768,7168,8,256,64,balanced,0.08922239989042283
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,768,7168,8,256,64,balanced,0.11430400013923644
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,320,7168,8,256,72,balanced,0.06635200008749961
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,320,7168,8,256,72,balanced,0.06521920040249825
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,320,7168,8,256,72,balanced,0.09601920023560524
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,320,7168,8,256,72,balanced,0.09773759990930557
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,768,7168,8,256,64,balanced,0.06243519932031631
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,384,7168,8,256,72,balanced,0.06597760058939457
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,384,7168,8,256,72,balanced,0.07712960094213486
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,384,7168,8,256,72,balanced,0.11075839996337891
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,384,7168,8,256,72,balanced,0.06765120066702365
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,384,7168,8,256,72,balanced,0.06209919974207878
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,384,7168,8,256,72,balanced,0.10634879991412163
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,384,7168,8,256,72,balanced,0.06531520001590252
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,384,7168,8,256,72,balanced,0.05756160020828247
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,384,7168,8,256,72,balanced,0.1059808000922203
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,384,7168,8,256,72,balanced,0.09834560006856918
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,512,7168,8,256,72,balanced,0.06579839922487736
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,512,7168,8,256,72,balanced,0.08787520006299018
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,512,7168,8,256,72,balanced,0.12160000130534172
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,512,7168,8,256,72,balanced,0.0623904000967741
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,512,7168,8,256,72,balanced,0.06541439965367317
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,512,7168,8,256,72,balanced,0.12115520164370537
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,512,7168,8,256,72,balanced,0.07391679920256138
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,512,7168,8,256,72,balanced,0.06173119992017746
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,512,7168,8,256,72,balanced,0.12440639957785607
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,512,7168,8,256,72,balanced,0.10648959875106812
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,768,7168,8,256,72,balanced,0.06729280091822147
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,768,7168,8,256,72,balanced,0.1158656008541584
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,768,7168,8,256,72,balanced,0.15862719863653182
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,768,7168,8,256,72,balanced,0.06663359999656678
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,768,7168,8,256,72,balanced,0.08088959977030755
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,768,7168,8,256,72,balanced,0.15542400181293486
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,768,7168,8,256,72,balanced,0.06870399974286556
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,768,7168,8,256,72,balanced,0.0688255999237299
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,768,7168,8,256,72,balanced,0.16011200100183487
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,768,7168,8,256,72,balanced,0.138121597468853
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1024,7168,8,256,72,balanced,0.06976320035755634
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1024,7168,8,256,72,balanced,0.14361919909715654
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1024,7168,8,256,72,balanced,0.19731200188398362
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,768,7168,8,256,64,balanced,0.06891839988529683
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,768,7168,8,256,64,balanced,0.11416960060596466
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1024,7168,8,256,72,balanced,0.06890880055725575
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1024,7168,8,256,72,balanced,0.088230399787426
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1024,7168,8,256,72,balanced,0.19494400024414063
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,768,7168,8,256,64,balanced,0.06387840025126934
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,768,7168,8,256,64,balanced,0.06147839985787869
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,768,7168,8,256,64,balanced,0.11517760157585144
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,768,7168,8,256,64,balanced,0.10906880125403404
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1024,7168,8,256,72,balanced,0.06879040002822875
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1024,7168,8,256,72,balanced,0.07470400035381317
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1024,7168,8,256,72,balanced,0.19956159889698027
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1024,7168,8,256,72,balanced,0.16450240164995195
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1024,7168,8,256,64,balanced,0.06363199949264527
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1024,7168,8,256,64,balanced,0.10242559984326363
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1024,7168,8,256,64,balanced,0.1394815981388092
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1536,7168,8,256,72,balanced,0.0679360006004572
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1536,7168,8,256,72,balanced,0.1939296007156372
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1536,7168,8,256,72,balanced,0.25826879739761355
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1024,7168,8,256,64,balanced,0.06757440008223056
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1024,7168,8,256,64,balanced,0.0761535994708538
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1024,7168,8,256,64,balanced,0.13381759971380233
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1024,7168,8,256,64,balanced,0.06398399993777275
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1024,7168,8,256,64,balanced,0.06543359942734242
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1024,7168,8,256,64,balanced,0.13639359921216965
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1024,7168,8,256,64,balanced,0.12658879905939102
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1536,7168,8,256,72,balanced,0.06802879944443703
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1536,7168,8,256,72,balanced,0.11548160091042518
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1536,7168,8,256,72,balanced,0.25811520367860796
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,1536,7168,8,256,64,balanced,0.06586880050599575
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,1536,7168,8,256,64,balanced,0.12667839974164963
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,1536,7168,8,256,64,balanced,0.17914240062236786
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1536,7168,8,256,72,balanced,0.07198400124907493
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1536,7168,8,256,72,balanced,0.08868480026721955
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1536,7168,8,256,72,balanced,0.2592256009578705
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1536,7168,8,256,72,balanced,0.21626560091972352
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,1536,7168,8,256,64,balanced,0.06674239970743656
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,1536,7168,8,256,64,balanced,0.08503040000796318
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,1536,7168,8,256,64,balanced,0.17279039919376374
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,2048,7168,8,256,72,balanced,0.06918080002069474
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,2048,7168,8,256,72,balanced,0.24545599967241288
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,2048,7168,8,256,72,balanced,0.3289055973291397
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,1536,7168,8,256,64,balanced,0.06815360188484192
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,1536,7168,8,256,64,balanced,0.07478079870343209
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,1536,7168,8,256,64,balanced,0.1785887986421585
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,1536,7168,8,256,64,balanced,0.16325440108776093
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,2048,7168,8,256,72,balanced,0.07255680114030838
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,2048,7168,8,256,72,balanced,0.14037439972162247
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,2048,7168,8,256,72,balanced,0.32841600477695465
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,2048,7168,8,256,64,balanced,0.06715520061552524
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,2048,7168,8,256,64,balanced,0.15242239981889724
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,2048,7168,8,256,64,balanced,0.21789440065622329
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,2048,7168,8,256,72,balanced,0.07263039946556091
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,2048,7168,8,256,72,balanced,0.10573119968175888
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,2048,7168,8,256,72,balanced,0.32749760150909424
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,2048,7168,8,256,72,balanced,0.2714336007833481
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,2048,7168,8,256,64,balanced,0.0701664000749588
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,2048,7168,8,256,64,balanced,0.10219519957900047
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,2048,7168,8,256,64,balanced,0.2158240020275116
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,2048,7168,8,256,64,balanced,0.06777919940650463
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,2048,7168,8,256,64,balanced,0.08477760031819344
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,2048,7168,8,256,64,balanced,0.2167263999581337
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,2048,7168,8,256,64,balanced,0.1965151995420456
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,3072,7168,8,256,72,balanced,0.07784000039100647
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,3072,7168,8,256,72,balanced,0.35788159966468813
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,3072,7168,8,256,72,balanced,0.4651712030172348
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,3072,7168,8,256,64,balanced,0.07498240023851395
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,3072,7168,8,256,64,balanced,0.20448640137910842
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,3072,7168,8,256,64,balanced,0.30026239454746245
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,3072,7168,8,256,72,balanced,0.07923519909381867
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,3072,7168,8,256,72,balanced,0.19633279889822006
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,3072,7168,8,256,72,balanced,0.47257919907569884
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,3072,7168,8,256,64,balanced,0.07275519967079162
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,3072,7168,8,256,64,balanced,0.13262079954147338
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,3072,7168,8,256,64,balanced,0.29687360227108
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,3072,7168,8,256,72,balanced,0.07500479966402054
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,3072,7168,8,256,72,balanced,0.14285119920969008
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,3072,7168,8,256,72,balanced,0.46586879789829255
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,3072,7168,8,256,72,balanced,0.3844095945358276
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,3072,7168,8,256,64,balanced,0.0775199994444847
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,3072,7168,8,256,64,balanced,0.10342399999499322
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,3072,7168,8,256,64,balanced,0.29895039200782775
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,3072,7168,8,256,64,balanced,0.2669984042644501
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,4096,7168,8,256,72,balanced,0.08168000057339668
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,4096,7168,8,256,72,balanced,0.47108160257339476
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,4096,7168,8,256,72,balanced,0.6072575926780701
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,4096,7168,8,256,64,balanced,0.07975999936461449
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,4096,7168,8,256,64,balanced,0.2581503987312317
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,4096,7168,8,256,64,balanced,0.3812991976737976
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,4096,7168,8,256,72,balanced,0.08082560077309608
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,4096,7168,8,256,72,balanced,0.2482815980911255
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,4096,7168,8,256,72,balanced,0.609932792186737
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,4096,7168,8,256,64,balanced,0.07856959924101829
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,4096,7168,8,256,64,balanced,0.15728639960289
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,4096,7168,8,256,64,balanced,0.3958335965871811
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,4096,7168,8,256,72,balanced,0.10096320137381554
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,4096,7168,8,256,72,balanced,0.1793056011199951
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,4096,7168,8,256,72,balanced,0.6147903978824616
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,4096,7168,8,256,72,balanced,0.48307200372219083
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,4096,7168,8,256,64,balanced,0.07967359945178032
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,4096,7168,8,256,64,balanced,0.11916479989886283
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,4096,7168,8,256,64,balanced,0.3779519975185394
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,4096,7168,8,256,64,balanced,0.33469759821891787
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,6144,7168,8,256,72,balanced,0.09180479943752289
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,6144,7168,8,256,72,balanced,0.7057887852191925
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,6144,7168,8,256,72,balanced,0.8940543949604034
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,6144,7168,8,256,64,balanced,0.09387520030140876
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,6144,7168,8,256,64,balanced,0.37309119701385496
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,6144,7168,8,256,64,balanced,0.5468544065952301
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,6144,7168,8,256,72,balanced,0.09519999995827674
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,6144,7168,8,256,72,balanced,0.3581472009420395
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,6144,7168,8,256,72,balanced,0.8991360068321228
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,6144,7168,8,256,64,balanced,0.09511040002107621
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,6144,7168,8,256,64,balanced,0.210172800719738
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,6144,7168,8,256,64,balanced,0.5457855999469757
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,6144,7168,8,256,72,balanced,0.09529279917478561
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,6144,7168,8,256,72,balanced,0.24583039730787276
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,6144,7168,8,256,72,balanced,0.8947839915752411
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,6144,7168,8,256,72,balanced,0.6982111930847168
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,6144,7168,8,256,64,balanced,0.08865920081734657
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,6144,7168,8,256,64,balanced,0.15483199954032897
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,6144,7168,8,256,64,balanced,0.5478496015071869
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,6144,7168,8,256,64,balanced,0.4706847995519638
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,8192,7168,8,256,72,balanced,0.10331840068101883
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,8192,7168,8,256,72,balanced,0.9435263931751251
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,8192,7168,8,256,72,balanced,1.1886431932449342
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,8192,7168,8,256,64,balanced,0.10870080068707466
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,8192,7168,8,256,64,balanced,0.48152320384979247
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,8192,7168,8,256,64,balanced,0.7273311972618103
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,8192,7168,8,256,72,balanced,0.10705919936299324
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,8192,7168,8,256,72,balanced,0.46260479688644407
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,8192,7168,8,256,72,balanced,1.1844192147254944
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,8192,7168,8,256,64,balanced,0.10250880047678948
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,8192,7168,8,256,64,balanced,0.26697279810905455
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,8192,7168,8,256,64,balanced,0.7233023941516876
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,8192,7168,8,256,72,balanced,0.11736960038542747
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,8192,7168,8,256,72,balanced,0.3112544000148773
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,8192,7168,8,256,72,balanced,1.1911615967750548
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,8192,7168,8,256,72,balanced,0.9119776010513305
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,8192,7168,8,256,64,balanced,0.11053760200738907
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,8192,7168,8,256,64,balanced,0.1885695993900299
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,8192,7168,8,256,64,balanced,0.7265248000621796
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,8192,7168,8,256,64,balanced,0.6085824012756348
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,12288,7168,8,256,72,balanced,0.12844159752130507
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,12288,7168,8,256,72,balanced,1.4281952023506164
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,12288,7168,8,256,72,balanced,1.7655840039253234
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,12288,7168,8,256,64,balanced,0.12582719922065735
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,12288,7168,8,256,64,balanced,0.722105598449707
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,12288,7168,8,256,64,balanced,1.0797887921333313
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,12288,7168,8,256,64,balanced,0.12922880053520203
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,12288,7168,8,256,64,balanced,0.37871679961681365
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,12288,7168,8,256,64,balanced,1.0828447937965393
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,12288,7168,8,256,72,balanced,0.13169920146465303
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,12288,7168,8,256,72,balanced,0.6876223981380463
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,12288,7168,8,256,72,balanced,1.767632007598877
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,12288,7168,8,256,64,balanced,0.12815679982304573
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,12288,7168,8,256,64,balanced,0.26102399826049805
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,12288,7168,8,256,64,balanced,1.0804895758628845
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,12288,7168,8,256,64,balanced,0.8920415937900543
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,12288,7168,8,256,72,balanced,0.13376000076532363
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,12288,7168,8,256,72,balanced,0.4524864017963409
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,12288,7168,8,256,72,balanced,1.7643296003341675
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,12288,7168,8,256,72,balanced,1.3835488080978393
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,16384,7168,8,256,64,balanced,0.15599039942026138
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,16384,7168,8,256,64,balanced,0.9596799969673157
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,16384,7168,8,256,64,balanced,1.4360416054725647
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,16384,7168,8,256,72,balanced,0.1520831987261772
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,16384,7168,8,256,72,balanced,1.9219615936279297
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,16384,7168,8,256,72,balanced,2.3638015747070313
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,16384,7168,8,256,64,balanced,0.15650239884853362
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,16384,7168,8,256,64,balanced,0.49882239699363706
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,16384,7168,8,256,64,balanced,1.438262414932251
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,16384,7168,8,256,72,balanced,0.15763520151376725
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,16384,7168,8,256,72,balanced,0.9212127983570099
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,16384,7168,8,256,72,balanced,2.3638495922088625
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,16384,7168,8,256,64,balanced,0.15612479895353318
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,16384,7168,8,256,64,balanced,0.329763200879097
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,16384,7168,8,256,64,balanced,1.4394559741020203
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,16384,7168,8,256,64,balanced,1.1877536058425904
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,16384,7168,8,256,72,balanced,0.15435200035572053
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,16384,7168,8,256,72,balanced,0.5903263986110687
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,16384,7168,8,256,72,balanced,2.3582880020141603
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,16384,7168,8,256,72,balanced,1.7822431921958923
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,20480,7168,8,256,64,balanced,0.179932801425457
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,20480,7168,8,256,64,balanced,1.2003199934959412
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,20480,7168,8,256,64,balanced,1.803164792060852
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,20480,7168,8,256,72,balanced,0.1817280024290085
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,20480,7168,8,256,72,balanced,2.424243211746216
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,20480,7168,8,256,72,balanced,2.9538559913635254
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,20480,7168,8,256,64,balanced,0.17951040118932723
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,20480,7168,8,256,64,balanced,0.6227648019790649
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,20480,7168,8,256,64,balanced,1.8046623945236206
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,20480,7168,8,256,72,balanced,0.17750080078840255
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,20480,7168,8,256,72,balanced,1.1643328070640564
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,20480,7168,8,256,72,balanced,2.9545088052749633
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,20480,7168,8,256,64,balanced,0.19642239958047866
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,20480,7168,8,256,64,balanced,0.39696319699287413
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,20480,7168,8,256,64,balanced,1.8065120100975036
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,20480,7168,8,256,64,balanced,1.4878464102745057
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,20480,7168,8,256,72,balanced,0.17966400235891342
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,20480,7168,8,256,72,balanced,0.7390048086643219
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,20480,7168,8,256,72,balanced,2.9572416067123415
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,20480,7168,8,256,72,balanced,2.2122463941574098
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,32768,7168,8,256,64,balanced,0.24995840191841126
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,32768,7168,8,256,64,balanced,1.9285152077674865
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,32768,7168,8,256,64,balanced,2.8841151714324953
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,32768,7168,8,256,72,balanced,0.25387840121984484
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,32768,7168,8,256,72,balanced,3.9720767974853515
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,32768,7168,8,256,72,balanced,4.7009471416473385
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,32768,7168,8,256,64,balanced,0.25074559897184373
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,32768,7168,8,256,64,balanced,1.0051807940006257
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,32768,7168,8,256,64,balanced,2.881625604629517
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,32768,7168,8,256,72,balanced,0.2534240007400513
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,32768,7168,8,256,72,balanced,1.9248288154602051
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,32768,7168,8,256,72,balanced,4.710985612869263
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,32768,7168,8,256,64,balanced,0.251740799844265
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,32768,7168,8,256,64,balanced,0.595472002029419
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,32768,7168,8,256,64,balanced,2.8946912050247193
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,32768,7168,8,256,64,balanced,2.30652801990509
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,32768,7168,8,256,72,balanced,0.2562784031033516
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,32768,7168,8,256,72,balanced,1.1537503957748414
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,32768,7168,8,256,72,balanced,4.7101696014404295
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,32768,7168,8,256,72,balanced,3.422383975982666
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,65536,7168,8,256,64,balanced,0.4467456012964249
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,65536,7168,8,256,64,balanced,3.85045440196991
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,65536,7168,8,256,64,balanced,5.760086393356323
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,float16,65536,7168,8,256,72,balanced,0.45206719636917114
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,float16,65536,7168,8,256,72,balanced,8.269187259674073
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,float16,65536,7168,8,256,72,balanced,9.372892951965332
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,65536,7168,8,256,64,balanced,0.4529151976108551
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,65536,7168,8,256,64,balanced,2.02913920879364
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,65536,7168,8,256,64,balanced,5.758777713775634
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,fp8,65536,7168,8,256,72,balanced,0.44435199797153474
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,fp8,65536,7168,8,256,72,balanced,4.07196798324585
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,fp8,65536,7168,8,256,72,balanced,9.367043018341064
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,65536,7168,8,256,64,balanced,0.4431840032339096
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,65536,7168,8,256,64,balanced,1.1331199884414673
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,65536,7168,8,256,64,balanced,5.7667967796325685
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,65536,7168,8,256,64,balanced,4.410908842086792
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_prepare,MnnvlMoe,nvfp4,65536,7168,8,256,72,balanced,0.44351679980754855
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_dispatch,MnnvlMoe,nvfp4,65536,7168,8,256,72,balanced,2.311363196372986
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine,MnnvlMoe,nvfp4,65536,7168,8,256,72,balanced,9.384902381896973
TRTLLM,1.2.0rc5,NVIDIA GB200,alltoall_combine_low_precision,MnnvlMoe,nvfp4,65536,7168,8,256,72,balanced,6.726867246627807
