Getting started - Eindhoven University of Technologyrjordans/5LIM0/01-compiler organization.pdf•...

Getting started

Roel Jordans

• Translate a program from a high level language into something the processor can execute efficiently

• So before we start we need to know how this processor executes a program

2

Goal

Processor architectureby example: AVR

• Separate instruction and Data memories

• All 32 General purpose registers connected to ALU

• IO Modules connected to Data Bus and accessible via Special Function Registers

AVR CPU core

4

Memory organization

• One general purpose register file• Two memories– Program memory– Data memory

• And some peripherals–Memory mapped• e.g. IO ports, timers, EEPROM, ...

5

General purpose registers• 32x8-bit registers• 16-bit registers formed using

pairs of 8 bit registers• X, Y, and Z 16-bit registers– Can be used for indirect

addressing

• Operation input/output schemes:– One 8-bit operand, 8-bit result– Two 8-bit operands, 8-bit result– Two 8-bit operands, 16-bit result– One 16-bit operand, 16-bit result

• Hardware stack pointer

6

Program memory

• On chip, in system programmable

• 32K bytes, in 16K 2 byte instructions

• Accessible via special instructions (LPM,SPM)

• Bootloader support from boot flash section

• Contains interrupt vector table– Program starts with reset (IRQ0)

7

Data memory• Registers mapped in

SRAM memory space– 32 GPR– 64 SFR

• SFR's accessed through in/out instructions (IO registers)

• 2KB of internal SRAM from address 0x060

• Holds the stack!– Required for call instruction– Stack grows down from the

top of memory

8

Important information

• The AVR datasheet– Only 448 pages long!– Also includes info on the available peripherals– Focus on the following sections

• 6 & 7 for the architecture and memory organization• 31 for the instruction-set overview• 30 for the special function register

• Provided on the oncourse website

9

Compilation stages

Compilation flow overview

Assembly Source Files

C/C++ Source and Header

Files


User-created files

compiler assembler

Object Files

Shared Object

File

Linkable Image File

Executable Image File

Link Map File

Linker and Locator

preprocessor

Linker Script

FileMakefile

Make Utility

Library Files

Archive Utility

11

compiler assembler Linker and Locatorpreprocessor

12

Preprocessor

13assembler Linker and Locatorpreprocessor compiler

• Adds some really useful extra features–#include–#define–#ifdef

• Also useful for other languages– Like the assembler

The compiler

• Compilation in three steps– AST, IR, Code generation

14compiler assembler Linker and Locatorpreprocessor

An example from C


int foo(int a, int b) { return a + b; }

Abstract Syntax Tree (AST)


TranslationUnitDecl ...|TypedefDecl ...`FunctionDecl ... <f.c:1:1, line:3:1> line:1:5 foo 'int(int,int)' |ParmVarDecl ... <col:9, col:13> col:13 used a 'int' |ParmVarDecl ... <col:16, col:20> col:20 used b 'int' `CompoundStmt ... <col:23, line:3:1> `ReturnStmt ... <line:2:2, col:11> `BinaryOperator ... 'int' '+' |ImplicitCastExpr ... 'int' <LValueToRValue> | `DeclRefExpr ... 'int' lvalue ParmVar ... 'a' 'int' `ImplicitCastExpr ... 'int' <LValueToRValue> `DeclRefExpr ... 'int' lvalue ParmVar ... 'b' 'int'

$ clang Xclang astdump fsyntaxonly f.c


AST Tools


• Direct translation back to C– Code formatting with clang-format

• AST analysis– clang-check

• Buffer checking C code

– clang-tidy• Cleans up common programming mistakes

• Transforming AST– Source-to-source transformations– clang-modernize

• Update code to use newer language features

Intermediate Representation (IR)


$ clang target=avr emitllvm S f.c

; ModuleID = 'f.c'target datalayout = "..."target triple = "avrunknownunknown"

; Function Attrs: nounwinddefine i16 @foo(i16 %a, i16 %b) #0 { %1 = alloca i16, align 2 %2 = alloca i16, align 2 store i16 %a, i16* %1, align 2 store i16 %b, i16* %2, align 2 %3 = load i16* %1, align 2 %4 = load i16* %2, align 2 %5 = add nsw i16 %3, %4 ret i16 %5}


IR optimizations


• Code simplification– Dead-code elimination (DCE)– Common sub-expression elimination (CSE)

• Code analysis– Alias analysis– Control-flow analysis

• Code transformation– Loop transformation– Auto vectorization

Enabling IR optimizations


$ opt S O1 f.ll o f.opt.ll

; ModuleID = 'f.ll'target datalayout = "..."target triple = "avrunknownunknown"

; Function Attrs: nounwinddefine i16 @foo(i16 %a, i16 %b) #0 { %add = add nsw i16 %a, %b ret i16 %add}



$ llc f.opt.ll

.text .file "f.opt.ll" .globl foo .align 2 .type foo,@functionfoo: ; @foo; BB#0: ; %entry add r24, r22 adc r25, r23 ret.Lfunc_end0: .size foo, .Lfunc_end0foo

CodeGen


The assembler

• From human readable to executable• Produces object files– Usually in ELF or COFF format


Executable and Linkable Format

• Contains the binary data for the processor– Code & Data

• Optionally contains extra information– Debug information– Symbol tables– Relocation information

compiler assembler Linker and Locatorpreprocessor23

Linux ‘Executable’ ELF files

• The Executable ELF files produced by the Linux linker are configured for execution in a private ‘virtual’ address space, whereby every program gets loaded at the identical virtual memory-address


Executable versus Linkable

Section-Header Table(optional)

ELF Header

Section 2 Data

Section 3 Data

…Section n Data

Segment 1 Data

Segment 2 Data

Segment 3 Data

…Segment n Data

Linkable File Executable File

Section-Header Table

Program-Header Table(optional)

Program-Header Table

ELF Header

Section 1 Data


ELF Header


Section 1 Data

Section 2 Data

…Section n Data

Linkable File

ELF Header


Section 1 Data

Section 2 Data

…Section n Data

Linkable File

Role of the LinkerELF Header

Program-Header Table

Segment 1 Data

Segment 2 Data

…Segment n Data

Executable File


The linker

• Controllable by a linker script– Default provided for simple architectures

• Required when the mapping of sections to memories becomes more complex–Multiple data memories–Moving code around at runtime– Fat binaries: supporting multiple architectures– Encrypted binaries


C runtime

• Initializing the processor• What happens before and after main– Setting up the stack– Calling main– Stopping execution


To summarize


C/C++ Source and Header

Files


User-created files

compiler assembler

Object Files

Shared Object

File

Linkable Image File

Executable Image File

Link Map File

Linker and Locator

preprocessor

Linker Script

FileMakefile

Make Utility

Library Files

Archive Utility

29

The compiler driver

The compiler driver

• Many different tools in the compilation process–Mostly useful when debugging the compiler

• Needs simple interface

• The frontend program (clang) can actually call the right programs for you!

30

PCP Assignment 1:AVR backend

The application

Overview

33

• Practice makes perfect• Assignments on the ES group servers

– Login information can be found on oncourse.tue.nl

– Password distributed in “My grades”

• Start with “Tool setup”• Download and follow the assignment

instructions• Answer the Assignment 1.A quiz

questions on oncourse

Bonus exercise 1

• Make (or Makefiles)• As mentioned earlier in this lecture

– Manages compilation steps– Very useful to know– (Ab)used for many other purposes

• Also available on oncourse– Take home exercise with questions

online– Try to automate the build process of

assignment 1

AcknowledgementsThese slides were partially based on the following:• www.cs.usfca.edu/~cruse/cs630f06/lesson15.ppt

• http://embsys.technikum-wien.at/staff/veigl/MDBA/slides/II-AVR-Basics%281%29.ppt

36

http://www.cs.usfca.edu/~cruse/cs630f06/lesson15.ppt

http://embsys.technikum-wien.at/staff/veigl/MDBA/slides/II-AVR-Basics%281%29.ppt

Getting started - Eindhoven University of Technologyrjordans/5LIM0/01-compiler organization.pdf•...

Documents

Transcript of Getting started - Eindhoven University of Technologyrjordans/5LIM0/01-compiler organization.pdf•...