name,prefill_H100,prefill_MI300x,abs_diff Attention,4.6992201900505775,16.732406578152904,12.033186388102326 Llama3RotaryEmbedding,1.654183418481141,4.335820968684168,2.6816375502030274 LogitsProcessor,2.6814815672764203,0.8598186434823126,-1.8216629237941078 "MergedColumnParallelLinear(weight=bfloat16[28672, 4096])",41.146126621848175,33.08889975623442,-8.057226865613757 "QKVParallelLinear(weight=bfloat16[6144, 4096])",10.96928445819218,9.34991231474652,-1.6193721434456592 RMSNorm(weight=bfloat16[4096]) <- LlamaDecoderLayer,3.354452265386841,3.3751253573296616,0.02067309194282041 "RowParallelLinear(weight=bfloat16[4096, 14336])",20.778851360320235,20.448087215830284,-0.33076414448995095 "RowParallelLinear(weight=bfloat16[4096, 4096])",8.061780077104702,7.733184817216182,-0.32859525988851956 Sampler,0.8995345280289777,0.8750792223722248,-0.024455305656752913 SiluAndMul,5.5758567828256185,2.9901399503933863,-2.585716832432232 others,0.05437592877606764,0.051256658571903924,-0.003119270204163714 vocab_embed_ops,0.12485280170906228,0.1602685169860263,0.03541571527696401