FusionStitching, Deep Fusion and Code Generation for Tensorflow Computations on GPUs | Don't Respond #19
Description
https://hjchen2.github.io/2019/11/27/DeepFusion/
FusionStitching系统概述 屏幕快照 2019-11-25 13.56.40 输入HloModule，经过以下三个阶段，最终输出LLVM IR。 Computation Fusion Schedule Planning Code Generation 论文主要针对XLA Fusion算法进行了改进，提出了实现Block合并策略的Schedule和Shared Memory Planni