Book Image

Learn LLVM 17 - Second Edition

By : Kai Nacke, Amy Kwan
Book Image

Learn LLVM 17 - Second Edition

By: Kai Nacke, Amy Kwan

Overview of this book

LLVM was built to bridge the gap between the theoretical knowledge found in compiler textbooks and the practical demands of compiler development. With a modular codebase and advanced tools, LLVM empowers developers to build compilers with ease. This book serves as a practical introduction to LLVM, guiding you progressively through complex scenarios and ensuring that you navigate the challenges of building and working with compilers like a pro. The book starts by showing you how to configure, build, and install LLVM libraries, tools, and external projects. You’ll then be introduced to LLVM's design, unraveling its applications in each compiler stage: frontend, optimizer, and backend. Using a real programming language subset, you'll build a frontend, generate LLVM IR, optimize it through the pipeline, and generate machine code. Advanced chapters extend your expertise, covering topics such as extending LLVM with a new pass, using LLVM tools for debugging, and enhancing the quality of your code. You'll also focus on just-in-time compilation issues and the current state of JIT-compilation support with LLVM. Finally, you’ll develop a new backend for LLVM, gaining insights into target description and how instruction selection works. By the end of this book, you'll have hands-on experience with the LLVM compiler development framework through real-world examples and source code snippets.
Table of Contents (20 chapters)
1
Part 1: The Basics of Compiler Construction with LLVM
4
Part 2: From Source to Machine Code Generation
10
Part 3: Taking LLVM to the Next Level
14
Part 4: Roll Your Own Backend

Global instruction selection

Instruction selection via the selection DAG produces fast code, but it takes time to do so. The speed of the compiler is often critical for developers, who want to quickly try out the changes they’ve made. Usually, the compiler should be very fast at optimization level 0, but it can take more time with increased optimization levels. However, constructing the selection DAG costs so much time that this approach does not scale as required. The first solution was to create another instruction selection algorithm called FastISel, which is fast but does not generate good code. It also does not share code with the selection DAG implementation, which is an obvious problem. Because of this, not all targets support FastISel.

The selection DAG approach does not scale because it is a large, monolithic algorithm. If we can avoid creating a new data structure such as the selection DAG, then we should be able to perform the instruction selection using small...