/deepcpu

try to reproduce https://www.usenix.org/conference/atc18/presentation/zhang-minjia

DeepCPU

The repo aims to implement the paper to boost the rnn (eg. rnn, dru, lstm) by PCP(private-cache partition) and WCS(weight-centric-streamlining). The sequential MKL and the MLAS(Microsoft blas of its onnxruntime) will be selected as the low-level implementation of the repo. The repo is borrowed from the microsoft/onnxruntime and has no warranty of its quality.

WARNING

The repo has no relation with Microsoft and is just my paratice. The repo has no warranty of its quality.