name,decode_1_H100,decode_1_MI300x,abs_diff Attention,7.66453436872326,8.0284914956388,0.3639571269155404 Llama3RotaryEmbedding,1.7494507609966288,2.1471268792258464,0.39767611822921767 LogitsProcessor,5.112673045317336,4.599264651697881,-0.5134083936194545 "MergedColumnParallelLinear(weight=bfloat16[28672, 4096])",37.83081076377858,35.60876904150703,-2.2220417222715483 "QKVParallelLinear(weight=bfloat16[6144, 4096])",9.760101441762473,9.615342268135976,-0.14475917362649682 RMSNorm(weight=bfloat16[4096]) <- LlamaDecoderLayer,3.0871978635252053,5.3236893643251175,2.236491500799912 RMSNorm(weight=bfloat16[4096]) <- LlamaForCausalLM,0.04648338109080121,0.07917001526701674,0.03268663417621553 "RowParallelLinear(weight=bfloat16[4096, 14336])",20.27652623002319,19.178927528919335,-1.0975987011038555 "RowParallelLinear(weight=bfloat16[4096, 4096])",8.522408056297714,7.513383564505963,-1.0090244917917506 Sampler,1.8217729154778657,3.2192338402399026,1.3974609247620369 SiluAndMul,4.084844495629638,4.614384286676602,0.5295397910469637 others,0.043196677377310214,0.0722170638605179,0.02902038648320769