1 =============================================
2 My First Language Frontend with LLVM Tutorial
3 =============================================
19 **Requirements:** This tutorial assumes you know C++, but no previous
20 compiler experience is necessary.
22 Welcome to the "My First Language Frontend with LLVM" tutorial. Here we
23 run through the implementation of a simple language, showing
24 how fun and easy it can be. This tutorial will get you up and running
25 fast and show a concrete example of something that uses LLVM to generate
28 This tutorial introduces the simple "Kaleidoscope" language, building it
29 iteratively over the course of several chapters, showing how it is built
30 over time. This lets us cover a range of language design and LLVM-specific
31 ideas, showing and explaining the code for it all along the way,
32 and reduces the overwhelming amount of details up front. We strongly
33 encourage that you *work with this code* - make a copy and hack it up and
36 **Warning**: In order to focus on teaching compiler techniques and LLVM
38 this tutorial does *not* show best practices in software engineering
39 principles. For example, the code uses global variables
40 pervasively, doesn't use
41 `visitors <http://en.wikipedia.org/wiki/Visitor_pattern>`_, etc... but
42 instead keeps things simple and focuses on the topics at hand.
44 This tutorial is structured into chapters covering individual topics,
45 allowing you to skip ahead as you wish:
47 - `Chapter #1: Kaleidoscope language and Lexer <LangImpl01.html>`_ -
48 This shows where we are
49 going and the basic functionality that we want to build. A lexer
50 is also the first part of building a parser for a language, and we
51 use a simple C++ lexer which is easy to understand.
52 - `Chapter #2: Implementing a Parser and AST <LangImpl02.html>`_ -
53 With the lexer in place, we can talk about parsing techniques and
54 basic AST construction. This tutorial describes recursive descent
55 parsing and operator precedence parsing.
56 - `Chapter #3: Code generation to LLVM IR <LangImpl03.html>`_ - with
57 the AST ready, we show how easy it is to generate LLVM IR, and show
58 a simple way to incorporate LLVM into your project.
59 - `Chapter #4: Adding JIT and Optimizer Support <LangImpl04.html>`_ -
60 One great thing about LLVM is its support for JIT compilation, so
61 we'll dive right into it and show you the 3 lines it takes to add JIT
62 support. Later chapters show how to generate .o files.
63 - `Chapter #5: Extending the Language: Control Flow <LangImpl05.html>`_ - With
64 the basic language up and running, we show how to extend
65 it with control flow operations ('if' statement and a 'for' loop). This
66 gives us a chance to talk about SSA construction and control
68 - `Chapter #6: Extending the Language: User-defined Operators
69 <LangImpl06.html>`_ - This chapter extends the language to let
70 users define arbitrary unary and binary operators - with assignable
71 precedence! This allows us to build a significant piece of the
72 "language" as library routines.
73 - `Chapter #7: Extending the Language: Mutable Variables
74 <LangImpl07.html>`_ - This chapter talks about adding user-defined local
75 variables along with an assignment operator. This shows how easy it is
76 to construct SSA form in LLVM: LLVM does *not* require your front-end
77 to construct SSA form in order to use it!
78 - `Chapter #8: Compiling to Object Files <LangImpl08.html>`_ - This
79 chapter explains how to take LLVM IR and compile it down to object
80 files, like a static compiler does.
81 - `Chapter #9: Debug Information <LangImpl09.html>`_ - A real language
82 needs to support debuggers, so we
83 add debug information that allows setting breakpoints in Kaleidoscope
84 functions, print out argument variables, and call functions!
85 - `Chapter #10: Conclusion and other tidbits <LangImpl10.html>`_ - This
86 chapter wraps up the series by discussing ways to extend the language
87 and includes pointers to info on "special topics" like adding garbage
88 collection support, exceptions, debugging, support for "spaghetti
91 By the end of the tutorial, we'll have written a bit less than 1000 lines
92 of (non-comment, non-blank) lines of code. With this small amount of
93 code, we'll have built up a nice little compiler for a non-trivial
94 language including a hand-written lexer, parser, AST, as well as code
95 generation support - both static and JIT! The breadth of this is a great
96 testament to the strengths of LLVM and shows why it is such a popular
97 target for language designers and others who need high performance code