framework,version,device,op_name,kernel_source,allreduce_dtype,num_gpus,message_size,latency,backend
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,128,0.004191040098667145,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,128,0.020526399612426756,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,256,0.004207040071487427,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,256,0.02070336103439331,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,512,0.0042358401417732245,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,512,0.020696640014648438,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,1024,0.0042847999930381775,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,1024,0.020619521141052248,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,2048,0.004243200123310089,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,2048,0.020552000999450683,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,4096,0.004314880073070526,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,4096,0.020541119575500488,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,8192,0.004398080110549927,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,8192,0.02059135913848877,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,16384,0.0045100799202919,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,16384,0.02044735908508301,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,32768,0.00464383989572525,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,32768,0.02075615882873535,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,65536,0.004768320024013519,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,65536,0.020367040634155273,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,131072,0.005171200037002563,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,131072,0.020502080917358396,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,262144,0.006338239908218383,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,262144,0.020586559772491457,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,524288,0.008307520151138306,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,524288,0.02067039966583252,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,1048576,0.012389440536499024,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,1048576,0.020776000022888184,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,2097152,0.020411200523376465,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,2097152,0.027276799678802487,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,4194304,0.03527168035507202,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,4194304,0.04597375869750976,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,8388608,0.07632832050323486,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,8388608,0.0854588794708252,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,16777216,0.1359494400024414,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,16777216,0.15299391746520996,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,33554432,0.24675264358520507,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,33554432,0.2786816024780273,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,67108864,0.46357566833496094,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,67108864,0.5282336044311523,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,134217728,0.8880646514892578,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,134217728,1.0144998168945312,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,268435456,1.7078028869628905,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,268435456,1.94778564453125,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,536870912,3.2343832397460934,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,536870912,3.750283203125,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,128,0.004303359985351562,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,128,0.027177278995513914,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,256,0.004392960071563721,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,256,0.021197760105133058,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,512,0.004353600144386292,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,512,0.021186559200286864,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,1024,0.004396480023860931,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,1024,0.021253440380096435,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,2048,0.0044633600115776065,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,2048,0.02047391891479492,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,4096,0.004545919895172119,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,4096,0.0205679988861084,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,8192,0.00461760014295578,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,8192,0.020594561100006105,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,16384,0.0048377600312232965,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,16384,0.02043999910354614,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,32768,0.0051385599374771124,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,32768,0.020408000946044922,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,65536,0.005731199979782105,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,65536,0.020430080890655518,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,131072,0.006959360241889953,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,131072,0.020511360168457032,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,262144,0.01050976037979126,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,262144,0.02040992021560669,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,524288,0.012122880220413208,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,524288,0.020425920486450193,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,1048576,0.017761919498443603,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,1048576,0.023929600715637205,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,2097152,0.029984960556030272,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,2097152,0.0372979211807251,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,4194304,0.053559999465942386,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,4194304,0.06448224067687988,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,8388608,0.09692511558532715,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,8388608,0.10614080429077148,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,16777216,0.1712371253967285,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,16777216,0.18984800338745117,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,33554432,0.320384635925293,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,33554432,0.3531155014038086,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,67108864,0.6114688110351563,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,67108864,0.6749430084228516,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,134217728,1.1809113311767578,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,134217728,1.3086854553222655,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,268435456,2.311841583251953,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,268435456,2.5700628662109373,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,536870912,4.5562307739257815,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,536870912,5.048380126953125,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,128,0.004570559859275817,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,128,0.02170432090759277,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,256,0.004627839922904968,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,256,0.02203200101852417,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,512,0.004592640101909637,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,512,0.021942079067230225,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,1024,0.004667519927024842,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,1024,0.02193376064300537,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,2048,0.004666880071163178,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,2048,0.0218668794631958,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,4096,0.0049001601338386535,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,4096,0.022008318901062012,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,8192,0.005155519843101501,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,8192,0.02198080062866211,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,16384,0.005648959875106811,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,16384,0.021569600105285643,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,32768,0.006398720145225525,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,32768,0.021586880683898926,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,65536,0.007833279967308045,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,65536,0.021765758991241453,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,131072,0.01403872013092041,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,131072,0.02178015947341919,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,262144,0.014915839433670045,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,262144,0.02170367956161499,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,524288,0.01650048017501831,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,524288,0.022110719680786133,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,1048576,0.020200319290161133,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,1048576,0.026328001022338864,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,2097152,0.03373215913772583,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,2097152,0.04186880111694336,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,4194304,0.06104351997375488,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,4194304,0.0739254379272461,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,8388608,0.12132575988769531,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,8388608,0.12630175590515136,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,16777216,0.1956425666809082,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,16777216,0.1998281669616699,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,33554432,0.31695104598999024,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,33554432,0.3495404815673828,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,67108864,0.5654195022583008,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,67108864,0.6235196685791016,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,134217728,1.0529926300048829,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,134217728,1.1778441619873046,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,268435456,2.039561309814453,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,268435456,2.288059844970703,sglang_eager
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,536870912,3.998787231445313,sglang_graph
SGLang,0.5.9,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,536870912,4.518257751464843,sglang_eager
