llvm-learning

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

LLVM Learning Skill

LLVM 学习技能

This skill provides curated learning paths and resources for mastering LLVM, Clang, and compiler development.

本技能为掌握LLVM、Clang及编译器开发提供了精心整理的学习路径与资源。

Learning Paths

学习路径

Beginner Path

入门路径

Start with Kaleidoscope: Official LLVM tutorial building a simple language
Understand LLVM IR: Learn the intermediate representation
Write Simple Passes: Transform IR with custom passes
Explore Clang: Understand C/C++ frontend

从Kaleidoscope开始：官方LLVM教程，构建一门简单语言
理解LLVM IR：学习中间表示
编写简单Pass：用自定义Pass转换IR
探索Clang：理解C/C++前端

Intermediate Path

进阶路径

Deep Dive into Optimization: Study built-in optimization passes
Backend Development: Target code generation
Analysis Frameworks: Pointer analysis, dataflow
Clang Tooling: LibTooling, AST matchers

深入优化技术：研究内置优化Pass
后端开发：目标代码生成
分析框架：指针分析、数据流分析
Clang工具链：LibTooling、AST匹配器

Advanced Path

高级路径

MLIR: Multi-level IR for domain-specific compilation
JIT Compilation: ORC JIT framework
Security Features: Sanitizers, CFI implementation
Contribute to LLVM: Patches, reviews, community

MLIR：面向领域特定编译的多层IR
JIT编译：ORC JIT框架
安全特性：Sanitizers、CFI实现
贡献LLVM：补丁、代码评审、社区参与

Official Resources

官方资源

Documentation

文档

LLVM Docs: https://llvm.org/docs
Clang Docs: https://clang.llvm.org/docs
LLVM Language Reference: Complete IR specification
Programmer's Manual: Core APIs and usage

LLVM Docs：https://llvm.org/docs
Clang Docs：https://clang.llvm.org/docs
LLVM Language Reference：完整的IR规范
Programmer's Manual：核心API及使用指南

Tutorials

教程

Kaleidoscope: https://llvm.org/docs/tutorial/MyFirstLanguageFrontend/index.html
Writing an LLVM Pass: https://llvm.org/docs/WritingAnLLVMPass.html
Building LLVM: https://llvm.org/docs/GettingStarted.html

Kaleidoscope：https://llvm.org/docs/tutorial/MyFirstLanguageFrontend/index.html
Writing an LLVM Pass：https://llvm.org/docs/WritingAnLLVMPass.html
Building LLVM：https://llvm.org/docs/GettingStarted.html

Community Resources

社区资源

Books and Guides

书籍与指南

Learn LLVM 12 by Kai Nacke: Comprehensive practical guide
LLVM Techniques, Tips, and Best Practices: Advanced patterns
Getting Started with LLVM Core Libraries: Foundation concepts
LLVM Essentials: Quick start guide

Learn LLVM 12 by Kai Nacke：综合性实用指南
LLVM Techniques, Tips, and Best Practices：高级模式
Getting Started with LLVM Core Libraries：基础概念
LLVM Essentials：快速入门指南

Video Resources

视频资源

LLVM Developer Meetings: Recorded talks on YouTube
LLVM Weekly: Newsletter with latest developments
EuroLLVM/US LLVM: Conference recordings

LLVM Developer Meetings：YouTube上的录制演讲
LLVM Weekly：最新进展通讯
EuroLLVM/US LLVM：会议录制内容

GitHub Learning Projects

GitHub学习项目

undefined

undefined

Highly recommended starter projects

banach-space/llvm-tutor # Comprehensive pass examples hunterzju/llvm-tutorial # Step-by-step tutorials jauhien/iron-kaleidoscope # Kaleidoscope in Rust bigconvience/llvm-ir-in-action # IR examples

undefined

banach-space/llvm-tutor # Comprehensive pass examples hunterzju/llvm-tutorial # Step-by-step tutorials jauhien/iron-kaleidoscope # Kaleidoscope in Rust bigconvience/llvm-ir-in-action # IR examples

undefined

LLVM IR Essentials

LLVM IR 核心知识

Basic Structure

基本结构

llvm

; Module level
@global_var = global i32 42

; Function definition
define i32 @add(i32 %a, i32 %b) {
entry:
    %sum = add i32 %a, %b
    ret i32 %sum
}

; External declaration
declare i32 @printf(i8*, ...)

llvm

; Module level
@global_var = global i32 42

; Function definition
define i32 @add(i32 %a, i32 %b) {
entry:
    %sum = add i32 %a, %b
    ret i32 %sum
}

; External declaration
declare i32 @printf(i8*, ...)

Common Instructions

常见指令

llvm

; Arithmetic
%result = add i32 %a, %b
%result = sub i32 %a, %b
%result = mul i32 %a, %b

; Memory
%ptr = alloca i32
store i32 %value, i32* %ptr
%loaded = load i32, i32* %ptr

; Control flow
br i1 %cond, label %then, label %else
br label %next

; Function calls
%retval = call i32 @function(i32 %arg)

llvm

; Arithmetic
%result = add i32 %a, %b
%result = sub i32 %a, %b
%result = mul i32 %a, %b

; Memory
%ptr = alloca i32
store i32 %value, i32* %ptr
%loaded = load i32, i32* %ptr

; Control flow
br i1 %cond, label %then, label %else
br label %next

; Function calls
%retval = call i32 @function(i32 %arg)

Type System

类型系统

llvm

; Integer types: i1, i8, i16, i32, i64, i128
; Floating point: half, float, double
; Pointers: i32*, [10 x i32]*, { i32, i64 }*
; Arrays: [10 x i32]
; Structs: { i32, i64, i8* }
; Vectors: <4 x i32>

llvm

; Integer types: i1, i8, i16, i32, i64, i128
; Floating point: half, float, double
; Pointers: i32*, [10 x i32]*, { i32, i64 }*
; Arrays: [10 x i32]
; Structs: { i32, i64, i8* }
; Vectors: <4 x i32>

Writing LLVM Passes

编写LLVM Pass

New Pass Manager (Recommended)

新Pass管理器（推荐）

cpp

#include "llvm/Passes/PassPlugin.h"
#include "llvm/Passes/PassBuilder.h"

struct MyPass : public llvm::PassInfoMixin<MyPass> {
    llvm::PreservedAnalyses run(llvm::Function &F,
                                 llvm::FunctionAnalysisManager &FAM) {
        for (auto &BB : F) {
            for (auto &I : BB) {
                llvm::errs() << I << "\n";
            }
        }
        return llvm::PreservedAnalyses::all();
    }
};

// Plugin registration
extern "C" LLVM_ATTRIBUTE_WEAK ::llvm::PassPluginLibraryInfo
llvmGetPassPluginInfo() {
    return {LLVM_PLUGIN_API_VERSION, "MyPass", LLVM_VERSION_STRING,
            [](llvm::PassBuilder &PB) {
                PB.registerPipelineParsingCallback(
                    [](llvm::StringRef Name, llvm::FunctionPassManager &FPM,
                       llvm::ArrayRef<llvm::PassBuilder::PipelineElement>) {
                        if (Name == "my-pass") {
                            FPM.addPass(MyPass());
                            return true;
                        }
                        return false;
                    });
            }};
}

cpp

#include "llvm/Passes/PassPlugin.h"
#include "llvm/Passes/PassBuilder.h"

struct MyPass : public llvm::PassInfoMixin<MyPass> {
    llvm::PreservedAnalyses run(llvm::Function &F,
                                 llvm::FunctionAnalysisManager &FAM) {
        for (auto &BB : F) {
            for (auto &I : BB) {
                llvm::errs() << I << "\n";
            }
        }
        return llvm::PreservedAnalyses::all();
    }
};

// Plugin registration
extern "C" LLVM_ATTRIBUTE_WEAK ::llvm::PassPluginLibraryInfo
llvmGetPassPluginInfo() {
    return {LLVM_PLUGIN_API_VERSION, "MyPass", LLVM_VERSION_STRING,
            [](llvm::PassBuilder &PB) {
                PB.registerPipelineParsingCallback(
                    [](llvm::StringRef Name, llvm::FunctionPassManager &FPM,
                       llvm::ArrayRef<llvm::PassBuilder::PipelineElement>) {
                        if (Name == "my-pass") {
                            FPM.addPass(MyPass());
                            return true;
                        }
                        return false;
                    });
            }};
}

Running Custom Passes

运行自定义Pass

bash

undefined

bash

undefined

Build as shared library

clang++ -shared -fPIC -o MyPass.so MyPass.cpp

llvm-config --cxxflags --ldflags

clang++ -shared -fPIC -o MyPass.so MyPass.cpp

llvm-config --cxxflags --ldflags

Run with opt

opt -load-pass-plugin=./MyPass.so -passes="my-pass" input.ll -o output.ll

undefined

opt -load-pass-plugin=./MyPass.so -passes="my-pass" input.ll -o output.ll

undefined

Clang Development

Clang开发

AST Exploration

AST探索

bash

undefined

bash

undefined

Dump AST

clang -Xclang -ast-dump -fsyntax-only source.cpp

AST as JSON

clang -Xclang -ast-dump=json -fsyntax-only source.cpp

undefined

clang -Xclang -ast-dump=json -fsyntax-only source.cpp

undefined

LibTooling Basics

LibTooling基础

cpp

#include "clang/Tooling/Tooling.h"
#include "clang/ASTMatchers/ASTMatchers.h"
#include "clang/ASTMatchers/ASTMatchFinder.h"

using namespace clang::ast_matchers;

// Match all function declarations
auto matcher = functionDecl().bind("func");

class Callback : public MatchFinder::MatchCallback {
    void run(const MatchFinder::MatchResult &Result) override {
        if (auto *FD = Result.Nodes.getNodeAs<FunctionDecl>("func")) {
            llvm::outs() << "Found: " << FD->getName() << "\n";
        }
    }
};

cpp

#include "clang/Tooling/Tooling.h"
#include "clang/ASTMatchers/ASTMatchers.h"
#include "clang/ASTMatchers/ASTMatchFinder.h"

using namespace clang::ast_matchers;

// Match all function declarations
auto matcher = functionDecl().bind("func");

class Callback : public MatchFinder::MatchCallback {
    void run(const MatchFinder::MatchResult &Result) override {
        if (auto *FD = Result.Nodes.getNodeAs<FunctionDecl>("func")) {
            llvm::outs() << "Found: " << FD->getName() << "\n";
        }
    }
};

AST Matchers Reference

AST匹配器参考

cpp

// Common matchers
functionDecl()               // Match function declarations
varDecl()                    // Match variable declarations
binaryOperator()             // Match binary operators
callExpr()                   // Match function calls
ifStmt()                     // Match if statements

// Narrowing matchers
hasName("foo")               // Match by name
hasType(asString("int"))     // Match by type
isPrivate()                  // Match private members

// Traversal matchers
hasDescendant(...)           // Match in subtree
hasAncestor(...)             // Match in parent tree
hasBody(...)                 // Match function body

cpp

// Common matchers
functionDecl()               // Match function declarations
varDecl()                    // Match variable declarations
binaryOperator()             // Match binary operators
callExpr()                   // Match function calls
ifStmt()                     // Match if statements

// Narrowing matchers
hasName("foo")               // Match by name
hasType(asString("int"))     // Match by type
isPrivate()                  // Match private members

// Traversal matchers
hasDescendant(...)           // Match in subtree
hasAncestor(...)             // Match in parent tree
hasBody(...)                 // Match function body

Development Tools

开发工具

Essential Commands

核心命令

bash

undefined

bash

undefined

Compile to LLVM IR

clang -S -emit-llvm source.c -o source.ll

Optimize IR

opt -O2 source.ll -o optimized.ll

Disassemble bitcode

llvm-dis source.bc -o source.ll

Assemble to bitcode

llvm-as source.ll -o source.bc

Generate native code

llc source.ll -o source.s

View CFG

opt -passes=view-cfg source.ll

undefined

opt -passes=view-cfg source.ll

undefined

Debugging LLVM IR

调试LLVM IR

bash

undefined

bash

undefined

With debug info

clang -g -S -emit-llvm source.c

Debug IR transformation

opt -debug-pass=Structure -O2 source.ll

Verify IR validity

opt -passes=verify source.ll

undefined

opt -passes=verify source.ll

undefined

Common Analysis APIs

常用分析API

cpp

// Dominator Tree
#include "llvm/Analysis/Dominators.h"
DominatorTree &DT = FAM.getResult<DominatorTreeAnalysis>(F);
if (DT.dominates(BB1, BB2)) { /* BB1 dominates BB2 */ }

// Loop Analysis
#include "llvm/Analysis/LoopInfo.h"
LoopInfo &LI = FAM.getResult<LoopAnalysis>(F);
for (Loop *L : LI) { /* process loops */ }

// Alias Analysis
#include "llvm/Analysis/AliasAnalysis.h"
AAResults &AA = FAM.getResult<AAManager>(F);
AliasResult R = AA.alias(Ptr1, Ptr2);

// Call Graph
#include "llvm/Analysis/CallGraph.h"
CallGraph &CG = MAM.getResult<CallGraphAnalysis>(M);

cpp

// Dominator Tree
#include "llvm/Analysis/Dominators.h"
DominatorTree &DT = FAM.getResult<DominatorTreeAnalysis>(F);
if (DT.dominates(BB1, BB2)) { /* BB1 dominates BB2 */ }

// Loop Analysis
#include "llvm/Analysis/LoopInfo.h"
LoopInfo &LI = FAM.getResult<LoopAnalysis>(F);
for (Loop *L : LI) { /* process loops */ }

// Alias Analysis
#include "llvm/Analysis/AliasAnalysis.h"
AAResults &AA = FAM.getResult<AAManager>(F);
AliasResult R = AA.alias(Ptr1, Ptr2);

// Call Graph
#include "llvm/Analysis/CallGraph.h"
CallGraph &CG = MAM.getResult<CallGraphAnalysis>(M);

Practice Projects

实践项目

Beginner

入门级

Instruction counter pass
Function call logger
Dead code finder

指令计数器Pass
函数调用日志器
死代码查找工具

Intermediate

进阶级

Simple constant propagation
Branch probability estimator
Memory access pattern analyzer

简单常量传播
分支概率估算器
内存访问模式分析器

Advanced

高级

Custom optimization pass
Security vulnerability detector
Domain-specific language frontend

自定义优化Pass
安全漏洞检测器
领域特定语言前端

Community

社区

Getting Help

获取帮助

LLVM Discourse: https://discourse.llvm.org
LLVM Discord: Real-time chat
LLVM Mailing Lists: llvm-dev, cfe-dev

LLVM Discourse：https://discourse.llvm.org
LLVM Discord：实时聊天
LLVM邮件列表：llvm-dev, cfe-dev

Contributing

贡献代码

Phabricator: Code reviews (being migrated to GitHub)
GitHub: https://github.com/llvm/llvm-project
Bug Tracker: https://github.com/llvm/llvm-project/issues

Phabricator：代码评审（正在迁移至GitHub）
GitHub：https://github.com/llvm/llvm-project
Bug追踪器：https://github.com/llvm/llvm-project/issues

Resources Index

资源索引

For a comprehensive list of tutorials, books, and example projects, see the LLVM Tutorial and Clang Tutorial sections in the main README.md.

如需获取教程、书籍及示例项目的完整列表，请查看主README.md中的LLVM教程与Clang教程章节。

Getting Detailed Information

获取详细信息

When you need detailed and up-to-date resource links, tutorials, or learning materials, fetch the latest data from:

https://raw.githubusercontent.com/gmh5225/awesome-llvm-security/refs/heads/main/README.md

This README contains comprehensive curated lists of:

LLVM tutorials and learning resources (LLVM Tutorial section)
Clang tutorials and AST guides (Clang Tutorial section)
C++ modern programming guides (CPP Tutorial section)
Compiler and security books

当你需要详细且最新的资源链接、教程或学习资料时，请从以下地址获取最新数据：

https://raw.githubusercontent.com/gmh5225/awesome-llvm-security/refs/heads/main/README.md

该README包含精心整理的以下内容：

LLVM教程与学习资源（LLVM Tutorial章节）
Clang教程与AST指南（Clang Tutorial章节）
C++现代编程指南（CPP Tutorial章节）
编译器与安全相关书籍