name,decode_1_H100,decode_1_MI300x,abs_diff Attention,8.752717489220126,7.27695149430903,-1.4757659949110966 Llama3RotaryEmbedding,1.8000202471177884,2.104369711457971,0.3043494643401825 LogitsProcessor,4.964075664737471,4.514901920960399,-0.449173743777072 "MergedColumnParallelLinear(weight=bfloat16[28672, 4096])",37.824289906697395,35.46849998215996,-2.355789924537433 "QKVParallelLinear(weight=bfloat16[6144, 4096])",9.576314761464358,10.226939006282445,0.6506242448180863 RMSNorm(weight=bfloat16[4096]) <- LlamaDecoderLayer,2.9357176888315397,5.175429327375672,2.239711638544132 RMSNorm(weight=bfloat16[4096]) <- LlamaForCausalLM,0.04621376037593292,0.07626616238825962,0.030052402012326694 "RowParallelLinear(weight=bfloat16[4096, 14336])",19.938635218431514,19.464116135902668,-0.4745190825288468 "RowParallelLinear(weight=bfloat16[4096, 4096])",8.381148560652987,7.621727382262613,-0.7594211783903742 Sampler,1.731400248341767,3.2102290660877353,1.4788288177459683 SiluAndMul,4.0082858755763136,4.787734424960685,0.7794485493843712 others,0.041180578552811514,0.07283538585257543,0.031654807299763915