Fr

ee

LLVM is currently being used by many companies and has

a very active open source community. It provides a compiler
infrastructure, which can be used to write a compiler for a
language, generate machine code for a target architecture,
and also with a lot of other utility tools that can be easily
integrated into compiler projects.

Who this book is written for

P U B L I S H I N G

Implement advanced LLVM IR paradigms

Understand the LLVM IR Optimization
Pass Manager's infrastructure and write
an optimization pass
Imbibe basics of LLVM IR transformations
Understand the steps involved in converting
LLVM IR to Selection DAG
Implement a custom target using the
LLVM infrastructure

$ 24.99 US
15.99 UK

community experience distilled

pl

Building LLVM IR

C o m m u n i t y

Mayur Pandey

This book is intended for those who already know some

of the concepts concerning compilers and want to quickly
become familiar with the LLVM's infrastructure and the rich
set of libraries that it provides.

Get an introduction to LLVM modular design

and LLVM tools

Suyog Sarda

This book details how you can use LLVM compiler

infrastructure libraries effectively and will enable you to
quickly design your own custom compiler with LLVM. We
start with the basics, where you'll get to know all about
LLVM. We then cover how you can use LLVM library calls
to produce intermediate representations (IR) of simple
and complex high-level language paradigms. Moving on,
we show you how to implement optimizations at different
levels, write an optimization pass, generate code that is
independent of a target, and then map the generated code
to a backend. The book also walks you through IR-to-IR
transformations, advanced IR block transformations, and
target machines. By the end of this book, you'll be able to
easily utilize the LLVM libraries in your own projects.

What you will learn from this book

LLVM Essentials

LLVM Essentials

Sa
m

E x p e r i e n c e

LLVM Essentials
Get familiar with the LLVM infrastructure and start using LLVM
libraries to design a compiler

Prices do not include

local sales tax or VAT
where applicable

Visit www.PacktPub.com for books, eBooks,

code, downloads, and PacktLib.

D i s t i l l e d

Suyog Sarda

Mayur Pandey

In this package, you will find:

The authors biography

A preview chapter from the book, Chapter 1 'Playing with LLVM'
A synopsis of the books content
More information on LLVM Essentials

About the Authors

Suyog Sarda is a professional software engineer and an open source enthusiast.

He focuses on compiler development and compiler tools. He is an active contributor

to the LLVM open source community. Suyog was also involved in code performance
improvements for the ARM and X86 architectures. He has been a part of the compiler
team for the Tizen project. His interest in compiler development lies more in code
optimization and vectorization.
Previously, he has authored a book on LLVM, titled LLVM Cookbook by
Packt Publishing.
Apart from compilers, Suyog is also interested in Linux Kernel Development. He has
published a technical paper titled Secure Co-resident Virtualization in Multicore Systems
by VM Pinning and Page Coloring at the IEEE Proceedings of the 2012 International
Conference on Cloud Computing, Technologies, Applications, and Management
at the Birla Institute of Technology, Dubai. He has earned a bachelor's degree in
computer technology from the College of Engineering, Pune, India.

Mayur Pandey is a professional software engineer and open source enthusiast

focused on compiler development and tools. He is an active contributor to the
LLVM open source community. He has been a part of the compiler team for the
Tizen project and has hands-on experience of other proprietary compilers.
He has earned a bachelor's degree in Information Technology from Motilal
Nehru National Institute of Technology, Allahabad, India. Currently, he lives
in Bengaluru, India.

Preface

Preface
LLVM is one of the very hot topics in recent times. It is an open source project
with an ever-increasing number of contributors. Every programmer comes across
a compiler at some point or the other while programming. Simply speaking, a
compiler converts a high-level language to machine-executable code. However,
what goes on under the hood is a lot of complex algorithms at work. So, to get
started with compiler, LLVM will be the simplest infrastructure to study. Written
in object-oriented C++, modular in design, and with concepts that are very easy to
map to theory, LLVM proves to be attractive for experienced compiler programmers
and for novice students who are willing to learn.
As authors, we maintain that simple solutions frequently work better and are easier
to grasp than complex solutions. Throughout the book we will look at various topics
that will help you enhance your skills and drive you to learn more.
We also believe that this book will be helpful for people not directly involved in
compiler development as knowledge of compiler development will help them
write code optimally.

What this book covers

Chapter 1, Playing with LLVM, introduces you to the modular design of LLVM and
LLVM Intermediate Representation. In this chapter, we also look into some of the
tools that LLVM provides.
Chapter 2, Building LLVM IR, introduces you to some basic function calls provided by
the LLVM infrastructure to build LLVM IR. This chapter demonstrates building of
modules, functions, basic blocks, condition statements, and loops using LLVM APIs.
Chapter 3, Advanced LLVM IR, introduces you to some advanced IR paradigms. This
chapter explains advanced IR to the readers and shows how LLVM function calls
can be used to emit them in the IR.

Preface

Chapter 4, Basic IR Transformations, deals with basic transformation optimizations at

the IR level using the LLVM optimizer tool opt and the LLVM Pass infrastructure.
You will learn how to use the information of one pass in another and then look into
Instruction Simplification and Instruction Combining Passes.
Chapter 5, Advanced IR Block Transformations, deals with optimizations at block level
on IR. We will discuss various optimizations such as Loop Optimizations, Scalar
Evolution, Vectorization, and so on, followed by the summary of this chapter.
Chapter 6, IR to Selection DAG phase, takes you on a journey through the abstract
infrastructure of a target-independent code generator. We explore how LLVM IR
is converted to Selection DAG and various phases thereafter. It also introduces you
to instruction selection, scheduling, register allocation, and so on.
Chapter 7, Generating Code for Target Architecture, introduces the readers to the
tablegen concept. It shows how target architecture specifications such as register sets,
instruction sets, calling conventions, and so on can be represented using tablegen,
and how the output of tablegen can be used to emit code for a given architecture.
This chapter can be used by readers as a reference for bootstrapping a target
machine code generator.

Playing with LLVM

The LLVM Compiler infrastructure project, started in 2000 in University of Illinois,
was originally a research project to provide modern, SSA based compilation
technique for arbitrary static and dynamic programming languages. Now it has
grown to be an umbrella project with many sub projects within it, providing a set
of reusable libraries having well defined interfaces.
LLVM is implemented in C++ and the main crux of it is the LLVM core libraries it
provides. These libraries provide us with opt tool, the target independent optimizer,
and code generation support for various target architectures. There are other tools
which make use of core libraries, but our main focus in the book will be related to the
three mentioned above. These are built around LLVM Intermediate Representation
(LLVM IR), which can almost map all the high-level languages. So basically, to
use LLVM's optimizer and code generation technique for code written in a certain
programming language, all we need to do is write a frontend for a language that
takes the high level language and generates LLVM IR. There are already many
frontends available for languages such as C, C++, Go, Python, and so on. We will
cover the following topics in this chapter:

Modular design and collection of libraries

Getting familiar with LLVM IR

LLVM Tools and using them at command line

[1]

Playing With LLVM

Modular design and collection of libraries

The most important thing about LLVM is that it is designed as a collection of libraries.
Let's understand these by taking the example of LLVM optimizer opt. There are
many different optimization passes that the optimizer can run. Each of these passes
is written as a C++ class derived from the Pass class of LLVM. Each of the written
passes can be compiled into a .o file and subsequently they are archived into a .a
library. This library will contain all the passes for opt tool. All the passes in this
library are loosely coupled, that is they mention explicitly the dependencies on
other passes.
When the optimizer is ran, the LLVM PassManager uses the explicitly mentioned
dependency information and runs the passes in optimal way. The library based
design allows the implementer to choose the order in which passes will execute and
also choose which passes are to be executed based on the requirements. Only the
passes that are required are linked to the final application, not the entire optimizer.
The following figure demonstrates how each pass can be linked to a specific
object file within a specific library. In the following figure, the PassA references
LLVMPasses.a for PassA.o, whereas the custom pass refers to a different library
MyPasses.a for the MyPass.o object file.

The code generator also makes use of this modular design like the Optimizer, for
splitting the code generation into individual passes, namely, instruction selection,
register allocation, scheduling, code layout optimization, and assembly emission.

[2]

Chapter 1

In each of the following phases mentioned there are some common things for
almost every target, such as an algorithm for assigning physical registers available
to virtual registers even though the set of registers for different targets vary. So,
the compiler writer can modify each of the passes mentioned above and create
custom target-specific passes. The use of the tablegen tool helps in achieving this
using table description .td files for specific architectures. We will discuss how this
happens later in the book.
Another capability that arises out of this is the ability to easily pinpoint a bug to a
particular pass in the optimizer. A tool name Bugpoint makes use of this capability
to automatically reduce the test case and pinpoint the pass that is causing the bug.

Getting familiar with LLVM IR

LLVM Intermediate Representation (IR) is the heart of the LLVM project. In general
every compiler produces an intermediate representation on which it runs most of
its optimizations. For a compiler targeting multiple-source languages and different
architectures the important decision while selecting an IR is that it should neither be
of very high-level, as in very closely attached to the source language, nor it should
be very low-level, as in close to the target machine instructions. LLVM IR aims to be
a universal IR of a kind, by being at a low enough level that high-level ideas may be
cleanly mapped to it. Ideally the LLVM IR should have been target-independent, but
it is not so because of the inherent target dependence in some of the programming
languages itself. For example, when using standard C headers in a Linux system,
the header files itself are target dependent, which may specify a particular type
to an entity so that it matches the system calls of the particular target architecture.
Most of the LLVM tools revolve around this Intermediate Representation. The
frontends of different languages generate this IR from the high-level source
language. The optimizer tool of LLVM runs on this generated IR to optimize the
code for better performance and the code generator makes use of this IR for target
specific code generation. This IR has three equivalent forms:

An in-memory compiler IR

An on-disk bitcode representation

A Human readable form (LLVM Assembly)

[3]

Playing With LLVM

Now let's take an example to see how this LLVM IR looks like. We will take a small
C code and convert it into LLVM IR using clang and try to understand the details
of LLVM IR by mapping it back to the source language.
$ cat add.c
int globvar = 12;
int add(int a) {
return globvar + a;
}

Use the clang frontend with the following options to convert it to LLVM IR:
$ clang -emit-llvm -c -S add.c
$ cat add.ll
; ModuleID = 'add.c'
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
target triple = "x86_64-unknown-linux-gnu"
@globvar = global i32 12, align 4
; Function Attrs: nounwind uwtable
define i32 @add(i32 %a) #0 {
%1 = alloca i32, align 4
store i32 %a, i32* %1, align 4
%2 = load i32, i32* @globvar, align 4
%3 = load i32, i32* %1, align 4
%4 = add nsw i32 %2, %3
ret i32 %4
}
attributes #0 = { nounwind uwtable "less-precise-fpmad"="false" "noframe-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fpmath"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8"
"target-cpu"="x86-64" "unsafe-fp-math"="false" "use-soft-float"="false" }
!llvm.ident = !{!0}

Now let's look at the IR generated and see what it is all about. You can see the very
first line giving the ModuleID, that it defines the LLVM module for add.c file. An
LLVM module is a toplevel data structure that has the entire contents of the input
LLVM file. It consists of functions, global variables, external function prototypes,
and symbol table entries.
[4]

Chapter 1

The following lines show the target data layout and target triple from which we can
know that the target is x86_64 processor with Linux running on it. The datalayout
string tells us what is the endianess of machine ('e' meaning little endian), and the
name mangling (m : e denotes elf type). Each specification is separated by ''and
each following spec gives information about the type and size of that type. For
example, i64:64 says 64 bit integer is of 64 bits.
Then we have a global variable globvar. In LLVM IR all globals start with '@' and
all local variables start with '%'. There are two main reasons why the variables are
prefixed with these symbols. The first one being, the compiler won't have to bother
about a name clash with reserved words, the other being that the compiler can come
up quickly with a temporary name without having to worry about a conflict with
symbol table conflicts. This second property is useful for representing the IR in
static single assignment (SSA) from where each variable is assigned only a single
time and every use of a variable is preceded by its definition. So, while converting
a normal program to SSA form, we create a new temporary name for every
redefinition of a variable and limit the range of earlier definition till this redefinition.
LLVM views global variables as pointers, so an explicit dereference of the global
variable using load instruction is required. Similarly, to store a value, an explicit
store instruction is required.
Local variables have two categories:

Register allocated local variables: These are the temporaries and allocated
virtual registers. The virtual registers are allocated physical registers during
the code generation phase which we will see in a later chapter of the book.
They are created by using a new symbol for the variable like:
%1 = some value

Stack allocated local variables: These are created by allocating variables

on the stack frame of a currently executing function, using the alloca
instruction. The alloca instruction gives a pointer to the allocated type
and explicit use of load and store instructions is required to access and
store the value.
%2 = alloca i32

Now let's see how the add function is represented in LLVM IR. define i32
@add(i32 %a) is very similar to how functions are declared in C. It specifies the
function returns integer type i32 and takes an integer argument. Also, the function
name is preceded by '@', meaning it has global visibility.

[5]

Playing With LLVM

Within the function is actual processing for functionality. Some important things to
note here are that LLVM uses a three-address instruction, that is a data processing
instruction, which has two source operands and places the result in a separate
destination operand (%4 = add i32 %2, %3). Also the code is in SSA form, that is
each value in the IR has a single assignment which defines the value. This is useful
for a number of optimizations.
The attributes string that follows in the generated IR specifies the function attributes
which are very similar to C++ attributes. These attributes are for the function that
has been defined. For each function defined there is a set of attributes defined in
the LLVM IR.
The code that follows the attributes is for the ident directive that identifies the
module and compiler version.

LLVM tools and using them in the

command line
Until now, we have understood what LLVM IR (human readable form) is and how it
can be used to represent a high-level language. Now, we will take a look at some of
the tools that LLVM provides so that we can play around with this IR converting to
other formats and back again to the original form. Let's take a look at these tools one
by one along with examples.

llvm-as: This is the LLVM assembler that takes LLVM IR in assembly form
(human readable) and converts it to bitcode format. Use the preceding add.
ll as an example to convert it into bitcode. To know more about the LLVM
Bitcode file format refer to http://llvm.org/docs/BitCodeFormat.html
$ llvm-as add.ll o add.bc

To view the content of this bitcode file, a tool such as hexdump can be used.
$ hexdump c add.bc

llvm-dis: This is the LLVM disassembler. It takes a bitcode file as input and
outputs the llvm assembly.
$ llvm-dis add.bc o add.ll

If you check add.ll and compare it with the previous version, it will be the
same as the previous one.

[6]

Chapter 1

llvm-link: llvm-link links two or more llvm bitcode files and outputs one
llvm bitcode file. To view a demo write a main.c file that calls the function
in the add.c file.
$ cat main.c
#include<stdio.h>
extern int add(int);
int main() {
int a = add(2);
printf("%d\n",a);
return 0;
}

Convert the C source code to LLVM bitcode format using the following
command.
$ clang -emit-llvm -c main.c

Now link main.bc and add.bc to generate output.bc.

$ llvm-link main.bc add.bc -o output.bc

lli: lli directly executes programs in LLVM bitcode format using a just-in-time
compiler or interpreter, if one is available for the current architecture. lli is not
like a virtual machine and cannot execute IR of different architecture and can
only interpret for host architecture. Use the bitcode format file generated by
llvm-link as input to lli. It will display the output on the standard output.
$ lli output.bc
14

llc: llc is the static compiler. It compiles LLVM inputs (assembly form/
bitcode form) into assembly language for a specified architecture. In the
following example it takes the output.bc file generated by llvm-link and
generates the assembly file output.s.
$ llc output.bc o output.s

Let's look at the content of the output.s assembly, specifically the two
functions of the generated code, which is very similar to what a native
assembler would have generated.
Function main:
.type main,@function
main:
.cfi_startproc
# BB#0:

# @main

[7]

Playing With LLVM

pushq %rbp
.Ltmp0:
.cfi_def_cfa_offset 16
.Ltmp1:
.cfi_offset %rbp, -16
movq %rsp, %rbp
.Ltmp2:
.cfi_def_cfa_register %rbp
subq $16, %rsp
movl $0, -4(%rbp)
movl $2, %edi
callq add
movl %eax, %ecx
movl %ecx, -8(%rbp)
movl $.L.str, %edi
xorl %eax, %eax
movl %ecx, %esi
callq printf
xorl %eax, %eax
addq $16, %rsp
popq %rbp
retq
.Lfunc_end0:

Function: add
add:
.cfi_startproc
# BB#0:
pushq %rbp
.Ltmp3:
.cfi_def_cfa_offset 16
.Ltmp4:
.cfi_offset %rbp, -16
movq %rsp, %rbp
.Ltmp5:
.cfi_def_cfa_register %rbp
movl %edi, -4(%rbp)
addl globvar(%rip), %edi
movl %edi, %eax
popq %rbp
retq
.Lfunc_end1:

# @add

[8]

Chapter 1

opt: This is modular LLVM analyzer and optimizer. It takes the input file and
runs the optimization or analysis specified on the command line. Whether it
runs the analyzer or optimizer depends on the command-line option.
opt [options] [input file name]

When the analyze option is provided it performs various analysis on

the input. There is a set of analysis options already provided that can be
specified through command line or else one can write down their own
analysis pass and provide the library to that analysis pass. Some of the useful
analysis passes that can be specified using the following command line
arguments are:

basicaa: basic alias analysis

da: dependence analysis

instcount: count the various instruction types.

loops: information about loops

scalar evolution: analysis of scalar evolution

When the analyze option is not passed, the opt tool does the actual
optimization work and tries to optimize the code depending upon the
command-line options passed. Similarly to the preceding case, you can use
some of the optimization passes already present or write your own pass for
optimization. Some of the useful optimization passes that can be specified
using the following command-line arguments are:

constprop: simple constant propagation.

dce: dead code elimination pass

globalopt: pass for global variable optimization

inline: pass for function inlining

instcombine: for combining redundant instructions

licm: loop invariant code motion

tailcallelim: Tail Call elimination

Before going ahead we must note that all the tools mentioned in this
chapter are meant for compiler writers. An end user can directly use
clang for compilation of C code without converting the C code into
intermediate representation

[9]

Playing With LLVM

Downloading the example code

You can download the example code files from your account at
http://www.packtpub.com for all the Packt Publishing books
you have purchased. If you purchased this book elsewhere, you can
visit http://www.packtpub.com/support and register to have
the files e-mailed directly to you.

Summary
In this chapter, we looked into the modular design of LLVM: How it is used in the
opt tool of LLVM, and how it is applicable across LLVM core libraries. Then we took
a look into LLVM intermediate representation, and how various entities (variables,
functions etc.) of a language are mapped to LLVM IR. In the last section, we discussed
about some of the important LLVM tools, and how they can be used to transform the
LLVM IR from one form to another.
In the next chapter, we will see how we can write a frontend for a language that can
output LLVM IR using the LLVM machinery.

[ 10 ]

Get more information LLVM Essentials

Where to buy this book

You can buy LLVM Essentials from the Packt Publishing website.
Alternatively, you can buy the book from Amazon, BN.com, Computer Manuals and most internet
book retailers.
Click here for ordering and shipping details.

www.PacktPub.com

Stay Connected: