framework,version,device,op_name,kernel_source,allreduce_dtype,num_gpus,message_size,latency,backend
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,2,128,0.004586560130119323,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,2,128,0.03605983972549438,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,2,256,0.004455040097236633,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,2,256,0.03568896055221558,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,2,512,0.004449599981307983,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,2,512,0.03590656042098999,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,2,1024,0.004436799883842468,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,2,1024,0.03599647998809814,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,2,2048,0.004473600089550019,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,2,2048,0.036025280952453616,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,2,4096,0.0044841599464416505,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,2,4096,0.035910079479217524,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,2,8192,0.004599680006504059,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,2,8192,0.035866239070892335,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,2,16384,0.0046003198623657225,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,2,16384,0.03593183994293213,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,2,32768,0.00472351998090744,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,2,32768,0.03615231990814209,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,2,65536,0.004912959933280945,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,2,65536,0.03606623888015747,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,2,131072,0.005356159806251526,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,2,131072,0.03615583896636963,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,2,262144,0.006514880061149597,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,2,262144,0.03627903938293457,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,2,524288,0.008559679985046387,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,2,524288,0.0362608003616333,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,2,1048576,0.01263167977333069,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,2,1048576,0.03657279968261719,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,2,2097152,0.020436160564422608,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,2,2097152,0.036561279296875,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,2,4194304,0.05277279853820801,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,2,4194304,0.057555198669433594,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,2,8388608,0.08101247787475586,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,2,8388608,0.09373920440673829,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,2,16777216,0.14163904190063475,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,2,16777216,0.165316162109375,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,2,33554432,0.2503027153015137,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,2,33554432,0.2954438400268555,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,2,67108864,0.4682150268554688,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,2,67108864,0.5583692932128906,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,2,134217728,0.8866553497314452,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,2,134217728,1.0663807678222657,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,2,268435456,1.6949430847167968,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,2,268435456,2.0492005920410157,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,2,536870912,3.2216381835937495,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,2,536870912,3.9365518188476565,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,4,128,0.004725120067596436,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,4,128,0.034895999431610106,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,4,256,0.004749439954757691,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,4,256,0.03455712080001831,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,4,512,0.004715520143508911,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,4,512,0.03460319995880127,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,4,1024,0.0047257599234580994,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,4,1024,0.03485951900482177,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,4,2048,0.004802879989147186,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,4,2048,0.03487519979476929,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,4,4096,0.0049801599979400635,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,4,4096,0.03481312036514282,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,4,8192,0.004967679977416992,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,4,8192,0.034693760871887205,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,4,16384,0.005112640261650085,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,4,16384,0.03477663993835449,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,4,32768,0.0054323202371597285,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,4,32768,0.035016000270843506,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,4,65536,0.006011199951171875,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,4,65536,0.03494688034057617,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,4,131072,0.007355520129203796,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,4,131072,0.03593600034713745,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,4,262144,0.011499840021133422,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,4,262144,0.035253760814666746,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,4,524288,0.01304800033569336,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,4,524288,0.035164799690246586,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,4,1048576,0.018801599740982056,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,4,1048576,0.03548480033874511,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,4,2097152,0.030601279735565184,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,4,2097152,0.03877887964248657,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,4,4194304,0.06611936092376709,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,4,4194304,0.07056159973144531,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,4,8388608,0.1034182357788086,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,4,8388608,0.11589823722839357,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,4,16777216,0.174847354888916,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,4,16777216,0.20011936187744142,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,4,33554432,0.3248886489868164,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,4,33554432,0.3711097717285156,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,4,67108864,0.6169311904907226,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,4,67108864,0.7076576232910157,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,4,134217728,1.184810562133789,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,4,134217728,1.3617289733886717,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,4,268435456,2.319433898925781,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,4,268435456,2.674194030761719,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,4,536870912,4.553898010253906,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,4,536870912,5.252532958984375,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,8,128,0.005782080292701721,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,8,128,0.035449280738830566,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,8,256,0.004769279956817627,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,8,256,0.035225279331207275,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,8,512,0.004806399941444397,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,8,512,0.03504960060119629,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,8,1024,0.004797439873218536,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,8,1024,0.035578560829162595,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,8,2048,0.0049526399374008176,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,8,2048,0.03661952018737793,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,8,4096,0.005120000243186951,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,8,4096,0.0366972804069519,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,8,8192,0.005596479773521424,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,8,8192,0.06962528228759765,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,8,16384,0.005850880146026612,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,8,16384,0.06986112117767333,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,8,32768,0.006728640198707581,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,8,32768,0.07052671909332275,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,8,65536,0.008138880133628845,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,8,65536,0.06987552165985107,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,8,131072,0.015130239725112914,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,8,131072,0.07009600162506104,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,8,262144,0.016038399934768677,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,8,262144,0.07022367954254151,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,8,524288,0.01768736004829407,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,8,524288,0.07051904201507568,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,8,1048576,0.021312639713287354,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,8,1048576,0.07091392040252685,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,8,2097152,0.03480479955673218,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,8,2097152,0.0709007978439331,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,8,4194304,0.08067968368530273,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,8,4194304,0.09488032341003419,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,8,8388608,0.12597311973571776,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,8,8388608,0.15229344367980957,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,8,16777216,0.21353727340698242,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,8,16777216,0.24125919342041016,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,8,33554432,0.3637705612182617,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,8,33554432,0.40985153198242186,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,8,67108864,0.6958422088623047,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,8,67108864,0.7850646209716797,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,8,134217728,1.3543235778808593,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,8,134217728,1.5271778869628907,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,8,268435456,2.639378662109375,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,8,268435456,2.9960147094726564,vllm_eager
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_graph,float16,8,536870912,5.2082934570312505,vllm_graph
vLLM,0.12.0,NVIDIA H100 80GB HBM3,all_reduce,vLLM_custom_eager,float16,8,536870912,5.9235424804687495,vllm_eager
