framework,version,device,op_name,kernel_source,allreduce_dtype,num_gpus,message_size,latency,backend
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,128,0.004415999948978424,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,128,0.019665600061416627,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,256,0.004436799883842468,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,256,0.018549439907073976,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,512,0.004524480104446412,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,512,0.018479039669036863,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,1024,0.00454367995262146,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,1024,0.01847391963005066,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,2048,0.004467839896678925,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,2048,0.01852671980857849,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,4096,0.00455487996339798,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,4096,0.018557440042495727,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,8192,0.004589439928531646,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,8192,0.01851104021072388,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,16384,0.004614079892635346,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,16384,0.01847040057182312,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,32768,0.004744639992713929,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,32768,0.018510400056838988,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,65536,0.004842239916324616,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,65536,0.018424960374832152,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,131072,0.005158720016479492,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,131072,0.01847488045692444,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,262144,0.006329280138015747,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,262144,0.018477760553359986,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,524288,0.008375040292739869,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,524288,0.018614720106124878,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,1048576,0.012396800518035888,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,1048576,0.018544000387191773,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,2097152,0.020371520519256593,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,2097152,0.026984319686889646,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,4194304,0.035391039848327636,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,4194304,0.04642911911010742,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,8388608,0.07635551929473877,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,8388608,0.08464927673339843,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,16777216,0.13668928146362305,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,16777216,0.15372575759887694,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,33554432,0.24737279891967773,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,33554432,0.2779385566711426,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,67108864,0.46416862487792965,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,67108864,0.528598403930664,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,134217728,0.8875734710693359,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,134217728,1.0146947479248047,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,268435456,1.6824143981933595,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,268435456,1.9475184631347655,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,536870912,3.2163519287109374,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,536870912,3.7408023071289067,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,128,0.004534080028533936,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,128,0.020130879878997802,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,256,0.004521279931068421,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,256,0.02038304090499878,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,512,0.004589119851589203,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,512,0.020345919132232666,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,1024,0.004618560075759888,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,1024,0.020233919620513914,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,2048,0.004581120014190674,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,2048,0.020176639556884764,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,4096,0.004731520116329193,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,4096,0.02035840034484863,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,8192,0.0047881600260734555,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,8192,0.020207359790802,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,16384,0.005005760192871094,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,16384,0.020361599922180177,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,32768,0.005315200090408325,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,32768,0.02034208059310913,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,65536,0.005824000239372253,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,65536,0.020221760272979734,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,131072,0.007087680101394653,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,131072,0.02029439926147461,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,262144,0.010532480478286744,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,262144,0.02046720027923584,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,524288,0.01198591947555542,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,524288,0.020388801097869873,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,1048576,0.017698559761047363,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,1048576,0.023808960914611814,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,2097152,0.029786880016326907,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,2097152,0.03702208042144776,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,4194304,0.05361087799072266,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,4194304,0.06468255996704102,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,8388608,0.09739263534545899,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,8388608,0.10502143859863282,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,16777216,0.1716441535949707,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,16777216,0.1893065643310547,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,33554432,0.32092159271240234,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,33554432,0.3521820831298828,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,67108864,0.6101500701904297,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,67108864,0.6748719787597656,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,134217728,1.1827935791015625,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,134217728,1.3086834716796876,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,268435456,2.3128703308105467,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,268435456,2.57037353515625,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,536870912,4.518193054199219,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,536870912,5.044471435546875,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,128,0.004773760139942169,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,128,0.020197439193725585,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,256,0.00468288004398346,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,256,0.020182080268859863,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,512,0.004770239889621735,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,512,0.02017888069152832,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,1024,0.004799680113792419,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,1024,0.020055360794067383,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,2048,0.004798080027103424,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,2048,0.01993056058883667,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,4096,0.005067520141601563,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,4096,0.02009216070175171,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,8192,0.005230079889297486,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,8192,0.020002880096435548,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,16384,0.005746880173683166,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,16384,0.020056641101837157,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,32768,0.006510720252990723,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,32768,0.02010528087615967,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,65536,0.00794975996017456,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,65536,0.020204479694366458,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,131072,0.014184000492095946,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,131072,0.02012351989746094,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,262144,0.01497215986251831,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,262144,0.02054527997970581,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,524288,0.016677119731903077,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,524288,0.02200416088104248,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,1048576,0.020045440196990966,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,1048576,0.026361279487609864,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,2097152,0.03353280067443848,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,2097152,0.04093887805938721,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,4194304,0.06059360027313232,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,4194304,0.07355391979217529,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,8388608,0.12099295616149904,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,8388608,0.12603103637695312,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,16777216,0.1962499237060547,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,16777216,0.20063711166381837,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,33554432,0.34821151733398437,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,33554432,0.34294464111328127,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,67108864,0.5633609771728516,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,67108864,0.6225040054321289,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,134217728,1.0530233764648438,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,134217728,1.1793231964111328,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,268435456,2.0342745971679688,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,268435456,2.291820220947266,sglang_eager
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,536870912,4.001587524414062,sglang_graph
SGLang,0.5.10,NVIDIA H200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,536870912,4.511891479492188,sglang_eager
