Different optimizer interfaces are called due to different engines that each sub-graph belongs to. To thoroughly utilize the calculation ability of the CUBE module in AICORE, A novel data layout format for faster hardware fetch is applied and the transition between normal 4D to this special format is performed in this stage. Such an operation guarantees less data handling between RAMs and CUBEs. Certain combination of operators is fused into a single big operator to further reduce the computation costs. This fusion is carried out in this stage as well.
Different optimizerA interfaces are called due to different engines that each sub-graph belongs to. To thoroughly utilize the calculation ability of the CUBE module in AICORE, A novel data layout format for faster hardware fetch is applied and the transition between normal 4D to this special format is performed in this stage. Such an operation guarantees less data handling between RAMs and CUBEs. Certain combination of operators is fused into a single big operator to further reduce the computation costs. This fusion is carried out in this stage as well.
ywx998152
4 years ago