趋近智
Compiler Optimizations for Machine Learning
章节 1: Anatomy of Deep Learning Compilers
Compilation Pipeline Overview
Computational Graphs and DAGs
Static Single Assignment in ML
High-Level vs Low-Level IR
Hands-on Practical: Inspecting Relay IR
章节 2: Graph-Level Transformations
Operator Fusion Strategies
Algebraic Simplification
Layout Transformation
Dead Code and Common Subexpression Elimination
Hands-on Practical: Implementing a Fusion Pass
章节 3: Kernel Optimization and Loop Schedules
The Loop Optimization Space
Loop Tiling and Blocking
Vectorization and SIMD
Loop Reordering and Unrolling
Hands-on Practical: Optimizing Matrix Multiplication
章节 4: The MLIR Infrastructure
MLIR Architecture and Dialects
The Linalg Dialect
Affine Dialect and Analysis
Pattern Rewriting and Lowering
Hands-on Practical: Creating a Custom Dialect
章节 5: Hardware-Aware Code Generation
GPU Memory Hierarchy Mapping
Thread Binding and Warp Divergence
Tensor Core Intrinsics
Shared Memory Banking and Conflicts
Hands-on Practical: Writing Triton Kernels
章节 6: Auto-Tuning and Search Spaces
Defining the Search Space
Search Algorithms: Random to Genetic
Machine Learning Cost Models
Ansor and AutoTVM Architecture
Hands-on Practical: Auto-Tuning a ResNet Block
Compiler Optimizations for Machine Learning
先修课程 Strong Python, ML concepts
级别:
高级
Intermediate Representations
Design and manipulate high-level and low-level intermediate representations for tensor computations.
Graph Optimization
Implement graph-level transformations such as operator fusion, constant folding, and dead code elimination.
Loop Transformations
Apply advanced loop optimizations including tiling, unrolling, and reordering to maximize cache locality.
MLIR Infrastructure
Utilize the Multi-Level Intermediate Representation framework to build custom dialects and lowering pipelines.
Auto-tuning
Develop strategies for automated kernel search and cost modeling to find optimal execution schedules.
本课程没有先修课程。
目前没有推荐的后续课程。
登录以撰写评论
分享您的反馈以帮助其他学习者。
© 2025 ApX Machine Learning用心打造