I am working at Alibaba’s T-Head as a Senior Staff Engineer, focusing on AI Compiler, MLIR and Polyhedral Compilation.

I earned my PhD degree from Zhejiang University under supervised by Professor Shi Zheng. And then I worked as a Senior software R&D at Semitronix responsible for the innovation and development of EDA tools.

Before joining T-Head, I worked at Huawei as a Technical Expert of AI compiler based on Polyhedral Compilation Technology. I am one of the main architect of MindAKG(GitHub, Gitee) which is an open-sourced AI compiler used in Huawei AI framework MindSpore for Huawei Ascend 910 and Nvidia GPU. During the period, I carried out some research and technical cooperation of AI Compilation with Professor Zhaojie, Professor Cedirc Bastoul and Professor Jiang Li.

Research Interests

• Polyhedral Compilation, Parallel Programming and Optimizations for Heterogeneous Systems
• AI Compiler, Multi-Level IR Compiler Framework (MLIR) and TVM
• Computational Lithography, Design For Manufacturing (DFM)

News

Our team is looking for self-motivated technical experts and outstanding college graduates, who are interested in Parallel Programming, AI Compiler, Polyhedral Compilation and MLIR. Please no hesitate to contact me. (dylangeng@163.com).

• 2023.11 : Our paper “Modeling the Interplay between Loop Tiling and Fusion in Optimizing Compilers using Affine Relations” was accepted by ACM TOCS. We are so lucky to be authors in one of the last rounds of TOCS publications, as the journal is no longer accepting submissions and will cease publication after 2023.
• 2023.05 : I was awarded CCF Senior Member.
• 2022.08 : Our paper “Parallelizing Neural Network Models Effectively on GPU by Implementing Reductions Atomically” is accepted by PACT 2022.
• 2022.01 : Our paper “Apollo: Automatic Partition-based Operator Fusion through Layer by Layer Optimization” is accepted by MLSys 2022.
• 2021.03: Our paper “AKG: Automatic Kernel Generation for Neural Processing Units using Polyhedral Transformations” is accepted by PLDI 2021.
• 2020.06: MindAKG is open-sourced. (GitHub, Gitee).