framework,version,device,op_name,kernel_source,allreduce_dtype,num_gpus,message_size,latency,backend
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,128,0.006809599995613098,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,128,0.021257600784301757,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,256,0.006611199975013733,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,256,0.02129568099975586,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,512,0.006640639901161194,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,512,0.021425600051879885,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,1024,0.006607999801635742,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,1024,0.0214899206161499,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,2048,0.007009279727935791,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,2048,0.021464319229125978,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,4096,0.006989439725875854,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,4096,0.021498560905456543,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,8192,0.006992319822311402,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,8192,0.021368958950042725,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,16384,0.007516800165176392,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,16384,0.021261439323425294,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,32768,0.007765120267868042,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,32768,0.021422080993652344,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,65536,0.0077203202247619635,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,65536,0.02145024061203003,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,131072,0.007707520127296448,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,131072,0.021557440757751463,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,262144,0.009569919705390931,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,262144,0.02135999917984009,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,524288,0.013581119775772093,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,524288,0.023528640270233152,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,1048576,0.02091360092163086,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,1048576,0.03114016056060791,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,2097152,0.03410016059875488,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,2097152,0.04368415832519531,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,4194304,0.06064223766326904,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,4194304,0.07012767791748047,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,8388608,0.05695680141448974,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,8388608,0.058745598793029784,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,16777216,0.08507712364196778,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,16777216,0.09261247634887695,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,33554432,0.1526460838317871,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,33554432,0.17233215332031251,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,67108864,0.2813824081420898,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,67108864,0.32069087982177735,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,134217728,0.5236726379394531,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,134217728,0.6042963027954101,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,268435456,1.0059619140625,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,268435456,1.1696883392333983,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,536870912,1.9049824523925782,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,536870912,2.227510986328125,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,128,0.007229440212249756,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,128,0.021741759777069092,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,256,0.007024959921836852,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,256,0.03999264001846313,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,512,0.006954240202903747,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,512,0.040163521766662595,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,1024,0.006882240176200867,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,1024,0.04023583889007568,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,2048,0.007060160040855408,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,2048,0.040027518272399906,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,4096,0.007134720087051391,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,4096,0.039767999649047855,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,8192,0.00747648000717163,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,8192,0.03995968103408813,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,16384,0.007835519909858705,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,16384,0.04014560222625733,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,32768,0.007959039807319641,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,32768,0.040368962287902835,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,65536,0.00826911985874176,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,65536,0.04029664039611817,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,131072,0.008759040236473083,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,131072,0.039650559425354004,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,262144,0.017681599855422975,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,262144,0.06335552215576172,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,524288,0.01872928023338318,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,524288,0.03896224021911621,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,1048576,0.027784960269927977,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,1048576,0.041179838180541994,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,2097152,0.04376863956451416,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,2097152,0.055216321945190425,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,4194304,0.0754863977432251,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,4194304,0.08619680404663085,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,8388608,0.08112768173217774,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,8388608,0.07832704067230226,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,16777216,0.12108575820922851,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,16777216,0.12933600425720215,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,33554432,0.18485088348388673,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,33554432,0.20845632553100585,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,67108864,0.35650081634521485,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,67108864,0.3955920028686523,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,134217728,0.6743791961669923,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,134217728,0.7660655975341797,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,268435456,1.2898141479492187,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,268435456,1.4556588745117187,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,536870912,2.4982313537597656,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,536870912,2.8258331298828123,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,128,0.007490879893302918,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,128,0.021469440460205078,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,256,0.007240960001945495,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,256,0.02140320062637329,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,512,0.00709119975566864,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,512,0.021367039680480957,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,1024,0.0070800000429153435,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,1024,0.021367359161376952,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,2048,0.007446399927139283,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,2048,0.021327359676361086,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,4096,0.007545920014381409,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,4096,0.021494081020355223,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,8192,0.007830719947814941,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,8192,0.021445760726928713,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,16384,0.008326719999313354,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,16384,0.021618878841400145,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,32768,0.008632959723472595,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,32768,0.02158560037612915,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,65536,0.009416319727897644,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,65536,0.021402881145477295,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,131072,0.025581760406494142,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,131072,0.03393663883209229,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,262144,0.025872321128845216,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,262144,0.03635871887207031,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,524288,0.027150719165802,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,524288,0.038402240276336666,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,1048576,0.028927040100097657,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,1048576,0.04068511962890625,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,2097152,0.04648543834686279,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,2097152,0.05833504199981689,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,4194304,0.0804099178314209,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,4194304,0.09136672019958496,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,8388608,0.10854623794555665,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,8388608,0.10476639747619629,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,16777216,0.16996095657348634,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,16777216,0.1734761619567871,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,33554432,0.27826080322265623,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,33554432,0.2870758438110351,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,67108864,0.39740352630615233,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,67108864,0.42278369903564456,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,134217728,0.7158089447021484,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,134217728,0.7858451080322266,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,268435456,1.3374447631835937,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,268435456,1.5010450744628907,sglang_eager
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,536870912,2.6031256103515625,sglang_graph
SGLang,0.5.9,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,536870912,2.9254620361328127,sglang_eager
