framework,version,device,op_name,kernel_source,allreduce_dtype,num_gpus,message_size,latency,backend
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,128,0.005353599786758423,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,128,0.05711552143096924,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,256,0.005020800232887268,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,256,0.05427552223205566,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,512,0.005010560154914856,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,512,0.05328095912933349,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,1024,0.005345600247383118,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,1024,0.055526719093322754,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,2048,0.005464320182800293,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,2048,0.056775679588317876,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,4096,0.005396479964256287,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,4096,0.05590720176696777,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,8192,0.0054860800504684445,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,8192,0.05489664077758789,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,16384,0.005588160157203674,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,16384,0.055365438461303706,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,32768,0.005792000293731689,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,32768,0.054817280769348144,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,65536,0.005837119817733764,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,65536,0.05466911792755127,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,131072,0.006602879762649535,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,131072,0.05490784168243408,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,262144,0.007845759987831116,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,262144,0.05404128074645996,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,524288,0.010437439680099487,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,524288,0.05497983932495117,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,1048576,0.015614080429077148,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,1048576,0.05484896183013917,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,2097152,0.025040640830993655,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,2097152,0.05476160049438476,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,4194304,0.04410304069519043,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,4194304,0.05700223922729493,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,8388608,0.11506431579589844,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,8388608,0.12012031555175781,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,16777216,0.2138128089904785,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,16777216,0.2244339179992676,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,33554432,0.38552417755126955,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,33554432,0.4090099334716797,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,67108864,0.6925865936279296,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,67108864,0.7363852691650391,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,134217728,1.3585798645019531,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,134217728,1.4419171142578127,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,268435456,2.631827392578125,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,268435456,2.7991082763671877,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,536870912,5.154788208007813,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,536870912,5.48697265625,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,128,0.005491840243339539,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,128,0.04977952003479004,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,256,0.005419520139694213,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,256,0.04295167922973633,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,512,0.005415679812431335,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,512,0.04294559955596924,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,1024,0.005434560179710389,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,1024,0.043505277633666996,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,2048,0.005691840052604676,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,2048,0.04487264156341553,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,4096,0.005457280278205872,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,4096,0.04468128204345703,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,8192,0.005651519894599914,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,8192,0.04473663806915283,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,16384,0.00582144021987915,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,16384,0.04297823905944824,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,32768,0.00609503984451294,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,32768,0.04356832027435303,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,65536,0.006207360029220581,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,65536,0.04421279907226562,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,131072,0.006932479739189148,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,131072,0.043282880783081054,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,262144,0.01195039987564087,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,262144,0.04333439826965332,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,524288,0.013247040510177614,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,524288,0.04316736221313476,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,1048576,0.018024319410324098,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,1048576,0.04286335945129395,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,2097152,0.027099521160125734,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,2097152,0.04361663818359375,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,4194304,0.04539040088653564,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,4194304,0.056514558792114256,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,8388608,0.07867775917053223,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,8388608,0.10273951530456542,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,16777216,0.11175264358520506,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,16777216,0.12002176284790038,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,33554432,0.20557376861572269,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,33554432,0.22891584396362305,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,67108864,0.3956796646118164,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,67108864,0.4393459320068359,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,134217728,0.777209243774414,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,134217728,0.8620035552978516,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,268435456,1.4893193054199219,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,268435456,1.6598597717285155,sglang_eager
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,536870912,2.906771545410156,sglang_graph
SGLang,0.5.9,NVIDIA GB200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,536870912,3.2440914916992183,sglang_eager
