framework,version,device,op_name,kernel_source,allreduce_dtype,num_gpus,message_size,latency,backend
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,128,0.0069753599166870114,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,128,0.01766592025756836,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,256,0.006981760263442993,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,256,0.017192000150680543,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,512,0.006717439889907836,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,512,0.017256319522857666,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,1024,0.006652160286903382,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,1024,0.017361279726028443,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,2048,0.007013440132141113,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,2048,0.017327359914779662,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,4096,0.007004479765892029,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,4096,0.017246400117874143,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,8192,0.007197120189666747,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,8192,0.017181440591812133,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,16384,0.0073907202482223515,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,16384,0.017056000232696534,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,32768,0.007405120134353638,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,32768,0.01719071984291077,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,65536,0.007783039808273315,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,65536,0.017799680233001706,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,131072,0.007905279994010925,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,131072,0.017675520181655885,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,262144,0.009630720019340514,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,262144,0.020251200199127198,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,524288,0.013507519960403442,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,524288,0.0232041597366333,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,1048576,0.019911999702453616,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,1048576,0.03038464069366455,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,2097152,0.03266272068023682,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,2097152,0.042783679962158205,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,4194304,0.06118527889251709,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,4194304,0.07002304077148438,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,8388608,0.057719678878784174,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,8388608,0.05963776111602783,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,16777216,0.08562399864196776,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,16777216,0.09325568199157715,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,33554432,0.1527827167510986,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,33554432,0.17180479049682618,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,67108864,0.2817292785644531,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,67108864,0.3212799835205078,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,134217728,0.5322518539428711,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,134217728,0.6125936126708984,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,268435456,0.9848604583740235,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,268435456,1.1507596588134765,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,2,536870912,1.8922915649414063,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,2,536870912,2.2167068481445313,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,128,0.0072780799865722655,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,128,0.017875839471817017,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,256,0.006962239742279053,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,256,0.01810207962989807,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,512,0.007029119729995728,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,512,0.017957760095596315,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,1024,0.006949440240859986,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,1024,0.017974079847335817,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,2048,0.007209920287132262,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,2048,0.0178985595703125,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,4096,0.007215039730072021,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,4096,0.017861759662628172,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,8192,0.007400959730148315,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,8192,0.01815840005874634,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,16384,0.007890239953994752,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,16384,0.01803712010383606,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,32768,0.008189759850502014,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,32768,0.01796671986579895,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,65536,0.00825439989566803,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,65536,0.01861024022102356,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,131072,0.008848000168800354,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,131072,0.018462400436401367,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,262144,0.017939840555191038,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,262144,0.028973119258880613,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,524288,0.018946880102157594,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,524288,0.030354878902435302,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,1048576,0.02754528045654297,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,1048576,0.03924031972885132,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,2097152,0.04368512153625488,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,2097152,0.05493311882019043,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,4194304,0.07506815910339355,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,4194304,0.08436511993408204,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,8388608,0.08190336227416992,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,8388608,0.07752607822418213,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,16777216,0.12173600196838379,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,16777216,0.129039363861084,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,33554432,0.18550111770629882,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,33554432,0.21133760452270506,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,67108864,0.3529657745361328,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,67108864,0.3967558288574219,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,134217728,0.6805635070800781,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,134217728,0.766627197265625,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,268435456,1.2988365173339844,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,268435456,1.4673350524902342,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,4,536870912,2.539725799560547,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,4,536870912,2.8647753906249998,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,128,0.007440959811210632,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,128,0.019036480188369752,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,256,0.007039039731025695,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,256,0.019065279960632325,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,512,0.007131519913673401,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,512,0.019083839654922486,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,1024,0.007151039838790893,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,1024,0.0189081597328186,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,2048,0.007441920042037964,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,2048,0.018905279636383058,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,4096,0.007567359805107117,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,4096,0.01906048059463501,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,8192,0.007826560139656066,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,8192,0.01904800057411194,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,16384,0.008425279855728149,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,16384,0.01900864005088806,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,32768,0.00866111993789673,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,32768,0.018969919681549072,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,65536,0.009422720074653626,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,65536,0.01941663980484009,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,131072,0.025661439895629884,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,131072,0.03370527982711792,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,262144,0.026003201007843018,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,262144,0.036362879276275635,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,524288,0.027322239875793457,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,524288,0.03770112037658692,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,1048576,0.028983039855957033,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,1048576,0.040694079399108886,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,2097152,0.04651648044586182,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,2097152,0.05787903785705566,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,4194304,0.08041088104248047,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,4194304,0.09121408462524414,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,8388608,0.10819135665893556,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,8388608,0.10645376205444337,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,16777216,0.16937088012695312,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,16777216,0.17317855834960938,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,33554432,0.27615711212158206,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,33554432,0.2871232032775879,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,67108864,0.3972275161743164,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,67108864,0.42370113372802737,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,134217728,0.7140882873535157,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,134217728,0.7869961547851563,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,268435456,1.3415420532226563,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,268435456,1.503417663574219,sglang_eager
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_graph,float16,8,536870912,2.59987548828125,sglang_graph
SGLang,0.5.10,NVIDIA B200,all_reduce,SGLang_CustomAllReduce_eager,float16,8,536870912,2.9278143310546874,sglang_eager
