Skip to main content

Posts

Featured

My MLIR Track #1 - Installation

**Disclaimer** : This is a documentation of my learning process only. Following these steps does not guarantee identical results. In the beginning, LLVM was developed to investigate dynamic compilation techniques for both static and dynamic programming languages. However, now day LLVM severs as portable, high-level assembly language, called language-independent intermediate representation (IR) to optimize the computation [1]. MLIR (Multi-Level Intermediate Representation ) was developed in 2018 by Google and released as a sub-project of LLVM in 2019. It is developed to manage the complexity of heterogeneous (hybrid) architectures, enabling precise orchestration of computation loading and memory usage across diverse hardware targets[2]. In order to delicately manage heterogeneous architectures, working on the compiler becomes now an essential pain to truly orchestrate data movement and workload balance. ## Installing MLIR It needs some dependencies to install the MLIR. ```bash # Ub...

Latest Posts

Pipeline: Double-Buffering

NVIDIA RTX Blackwell architecture

Parallel Computing: Challenges and Memory Models of MPI and OpenMP

平行計算:MPI 與 OpenMP 的挑戰與記憶體模型

【讀書筆記】我想開始去跑步

【讀書筆記】研究所這條路,我還沒走完。

【讀書筆記】改變世界的十七個方程式

關於鋼鐵人的弧形反應器