The Tiger Compiler Project

Nul n'est censé ignorer la loi.
Everything exposed in this document is expected to be known.


Node: Top, Next: , Up: (dir)

The Tiger Compiler Project

This document (split and beautiful or in one simple piece) details the various tasks the "Compilation" students must complete.

It was last edited on September 17, 2003, using:

     $ tc --version
     tc (LRDE Tiger Compiler 0.62)
     Revision 0.1004 Tue, 09 Sep 2003 16:44:26 +0200
     
     This package was written by and with the assistance of
     
     * Akim Demaille                akim@freefriends.org
       - Maintenance.
     
     * Alexandre Duret-Lutz         duret_g@epita.fr
     
     * Cedric Bail                  bail_c@epita.fr
       - Initial escaping static link computation framework.
     
     * Alexis Brouard               brouar_a@epita.fr
       - Portability of tc-check to NetBSD.
     
     * Benoît Perrot                benoit@lrde.epita.fr
       - Extensive documentation.
       - Redesign of the Task system.
       - Design and implementation of target handling.
       - Deep clean up of every single module.
     
     * Daniel Gazard                gazard_d@epita.fr
       - Initial framework from LIR to MIPS.
     
     * Francis Maes
       - Generation of static C++ Tree As Types.
     
     * Pierre-Yves Strub            strub_p@epita.fr
       - Redesign of the AST.
       - Design of Symbol.
     
     * Quôc Peyrot                  chojin@lrde.epita.fr
       - Initial Task framework.
     
     * Raphaël Poss                 r.poss@online.fr
       - Conversion of AST to using pointers instead of references.
       - Breakup between interfaces and implementations (.hh only -> .hxx, .cc)
       - Miscellaneous former TODO items.
     
     * Robert Anisko                anisko_r@epita.fr
     
     * Sébastien Broussaud          brouss_s@epita.fr
       - Escapes torture tests.
     
     * Stéphane Molina              molina_s@epita.fr
       - Configuration files in tc-check.
     
     * Thierry Géraud               theo@epita.fr
       - Initial idea for visitors.
       - Initial idea for tasks.
       - Initial implementation of AST.
       - Initial implementation of Tree.
     
     * Valentin David               david_v@epita.fr
       - Some additional tests.
     
     * Yann Popo                    popo_y@epita.fr
       - Implementation of the Timer class.
     
     * Yann Régis-Gianas            yann@lrde.epita.fr
       - Reimplementation of graphs
     
     Copyright (C) 2003 LRDE.
     
     Example 1: tc --version
     
     $ havm --version
     HAVM 0.20
     Written by Robert Anisko.
     
     Copyright (C) 2003 Laboratoire de Recherche et Développement de l'EPITA.
     This is free software; see the source for copying conditions.  There is NO
     warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
     Example 2: havm --version
     
     $ mipsy --version
     mipsy (Mipsy) 0.5
     Written by Benoit Perrot.
     
     Copyright (C) 2003 Benoit Perrot.
     mipsy comes with ABSOLUTELY NO WARRANTY.
     This is free software, and you are welcome to redistribute and modify it
     under certain conditions; see source for details.
     
     Example 3: mipsy --version
     

Table of Contents


Node: Introduction, Next: , Previous: Top, Up: Top

Introduction

This document presents the Tiger Project as part of the EPITA scholarship. It aims at the implementation of a Tiger compiler (see Modern Compiler Implementation) in C++.


Node: How to Read this Document, Next: , Up: Introduction

How to Read this Document

If you are a newcomer, you might be afraid by its sheer size. Don't worry, but in any case, do not give up: as stated in the very beginning of this document:

Nul n'est censé ignorer la loi.
That is to say everything exposed in this document is considered to be known. If it is written but you didn't know, you are wrong. If it is not written and was not clearly reported in the news, I am wrong.

Basically this document contains three kinds of informations:

Initial and Permanent
What you must read and know since the very beginning of the project. This includes most the following chapters: Introduction, Tarballs, and Evaluation.
Incremental
You should read these parts as and when needed. This includes mostly Compiler Stages.
Auxiliary
This information is provided to help you: just go there when you feel the need, Tools, and Project Layout.


Node: Why the Tiger Project, Next: , Previous: How to Read this Document, Up: Introduction

Why the Tiger Project

This project is quite different from most other EPITA projects, and has aims at several different goals, in different areas:

Several iterations
This project is about the only one with which you will live for 9 months, with the constant needs to fix errors found in earlier stages.
Complete Project
While the evaluation of most student projects is based on the code, this project restores the deserved emphasis on documentation and testing. Because of the duration of the project, you will value the importance of a good (developer's) documentation (why did we write this 4 months ago?), and of a good test suite (why does T2 fails now that we implemented T4? When did we break it?).

This also means that you have to design a test suite, and maintain it through out the project. The test suite is an integral part of the project.

Team Management
The Tiger Compiler is a long project, running from January to September (and optionally further). Each four person team is likely to experience nasty "human problems". This is explicitly a part of the project: the team management is a task you have to address. That may well include exclusion of lazy members.
C++
C++ is by no means an adequate language to study compilers (C would be even worse). Languages such as Haskell, Ocaml, Stratego are much better suited (actually the latter is even designed to this end). But, as already said, the primary goal is not to learn how to write a compiler: for an EPITA student, learning C++, Design Patterns, and Object Oriented Design is much more important.
English
English is the language for this project, starting with this very document, written by a French person, for French students. Nevertheless, it is impossible to be a good computer scientist with absolutely no fluency in English. Any attempt to break the importance of English is wrong. For instance, do not translate this document nor any other. Ask support to the Yakas, or to the English team.

By the past, some oral and written examinations were made in English. It may well be back some day.

Compiler
The project aims at the implementation of a compiler, but this is a minor issue. The field of compilers is a wonder place where most of computer science is concentrated, that's why this topic is extremely convenient as long term project. But it is not the major goal, the full list of all these items is.


Node: History, Previous: Why the Tiger Project, Up: Introduction

History

The Tiger Compiler Project evolves every year, so as to improve its infrastructure, to demonstrate more instructional material and so forth. This section tries to keep a list of these changes, together with the most constructive criticisms from students (or ourselves).

If you have information, including criticisms, that should be mentioned here, please send it to me.

The years correspond to the class, e.g., Tiger 2005 refers to EPITA class 2004, i.e., the project ran from January 2003 to September 2003.


Node: Fair Criticism, Next: , Up: History

Fair Criticism

Before diving into the history of the Tiger Compiler Project in EPITA, a whole project in itself for ourselves, with experimental tries, failures etc. it might be good to review some constrains that can explain why things are the way they are. Understanding these constraints will make it easier to criticize actual flaws, instead of focusing on issues that are mandated by other factors.

Tiger is an instructional project, the purpose of which is detailed above, see Why the Tiger Project. Because the input is a stream of students with virtually no knowledge whatsoever in C++, and our target is a stream of students with good fluency in the many constructs and an understanding of more complex matters, we have to gradually transform them via intermediate forms with increasing skills. In particular this means that by the end of the project, evolved techniques can and should be used, while at the beginning only introductory facts should be needed. As a consequence, we cannot have a nice and high-tech AST for instance.

Because the insight of compilers is not the primary goal, when a choice is to be made between (i) more interesting work on compiler internals with little C++ novelty, and (ii) providing most of this work and focusing on something else, then we are most likely to select the first option. This means that the Tiger Project is doomed to be a low-tech featureless compiler, with no call graph, no default optimization, no debugging support (outputting comments in the assembly showing the original code), no bells, no whistles etc. This also implies that sometimes good and interested students will feel we "stole" the pleasure to write nice pieces of code from them; understand that we actually provided code to the other students. However, you are free to rewrite everything if you wish.


Node: Tiger 2002, Next: , Previous: Fair Criticism, Up: History

Tiger 2002

Wrapping a tarball is impossible
During the first edition of the Tiger Compiler project, students had to write their own Makefiles -- after all, knowing Make is considered mandatory for an Epitean. This had the most dramatic effects, with a wide range of creative and imaginative ways to have your project fail; for instance:

As a result I grew tired of fixing the tarballs, and in order to have a robust, efficient (albeit some piece of pain in the neck sometimes) distributions 1 we moved to using Automake, and hence Autoconf.

There are reasons not to be happy with it, agreed. But there are many more reasons to be sad without it. So Autoconf and Automake are here to stay.

Note, however, that you are free to use another system if you wish. Just obey to the standard package interface (see Delivery).

The SemantVisitor is a nightmare to maintain
The SemantVisitor, which performs both the type checking and the translation to intermediate code, was near to impossible to deliver in pieces to the students: because type checking and translation were so much intertwined, it was not possible to deliver as a first step the type checking machinery template, and then the translation pieces. Students had to fight with non applicable patches. This was fixed in Tiger 2003 by splitting the SemantVisitor into TypeVisitor and TranslationVisitor. The negative impact, of course, is a performance loss.
Akim is tired during the student defenses
Seeing every single group for each compiler stage is a nightmare. Sometimes I was not enough aware.


Node: Tiger 2003, Next: , Previous: Tiger 2002, Up: History

Tiger 2003

The ast is rigid
Because the members of the ast objects were references, it was impossible to implement any change on it: simplifications, optimization etc. This is fixed in Tiger 2004 where all the members are now pointers, but the interface to these classes still uses references.
Akim is even more tired during the student defenses
Just as the previous year, see Tiger 2002, but with more groups and more stages. But now there are enough competent students to create a group of assistants, the Yakas, to help the students, and to share the load of defenses.
Upgrading is not easy
Only tarballs were delivered, making upgrades delicate, error prone, and time consuming. The systematic use of patches between tarballs since the 2004 edition solves this issue.
Upgraded tarballs don't compile
Students would like at least to be able to compile a tarball with its holes. To this end, much of the removed code is now inside functions, leaving just what it needed to satisfy the prototype. Unfortunately this is not very easy to do, and conflicts with the next complaint:
Filling holes is not interesting
In order to scale down the amount of code students have to write, in order to have them focus on instructional material, more parts are delivered almost complete except for a few interesting places. Unfortunately, some students decided to answer the question completely mechanically (copy, paste, tweak until it compiles), instead of focusing of completing their own education. There is not much I can do about this. Some parts will therefore grow; typically some files will be left empty instead of having most of the skeleton ready (prototypes and so forth). This means more work, but more interesting I guess. But it conflicts with the previous item...


Node: Tiger 2004, Next: , Previous: Tiger 2003, Up: History

Tiger 2004

The driver is not maintainable
The compiler driver was a nightmare to maintain, extend etc. when delivering additional modules etc. This was fixed in 2005 by the introduction of the Task model.
No sane documentation
This was addressed by the use of Doxygen in 2005.
No UML documentation
The solution is yet to be found.
Too many visitors
It seems that some students think there were too many visitors to implement. I do not subscribe to this view (after all, why not complain that "there are too many programs to implement", or, in a more C++ vocabulary "there are too many classes to implement"), nevertheless in Tiger 2005 this was addressed by making the EscapeVisitor "optional" (actually it became a rush).
Too many memory leaks
The only memory properly reclaimed is that of the ast. No better answer for the rest of the compiler. This is the most severe flaw in this project, and definitely the worst thing to remember of: what we showed is not what student should learn to do. Note too, that even though using a garbage collector is tempting and well suited for our tasks, its pedagogical content is less interesting: students should be taught how to properly manage the memory.
Upgraded tarballs don't compile
Filling holes is not interesting
See Tiger 2003.


Node: Tiger 2005, Previous: Tiger 2004, Up: History

Tiger 2005

Too many memory leaks
See Tiger 2004. The 2006 edition will pay strict attention to memory allocation. To demonstrate more memory management styles, the intermediate representation will be converted to smart pointers implementing reference counting.
Too long to compile
Too much code was in *.hh files. Since then the policy wrt file contents was defined (see File Conventions), and in Tiger 2006 was adjusted to obey these conventions. Unfortunately, although the improvement was significant, it was not measured precisely.

The interfaces between modules have also been cleaned to avoid excessive inter dependencies. Also, when possible, opaque types are used to avoid additional includes. For instance, ast/at-tasks.hh today includes:

          namespace ast
          {
            // Forward decl.
            class Exp;
          
            namespace tasks
            {
              /// Global root node of abstract syntax tree.
              extern ast::Exp* the_program;
          // ...
            }
          }
          

where it used to include all the ast headers to define exactly the type ast::Exp.

Upgraded tarballs don't compile
Filling holes is not interesting
See Tiger 2003.


Node: Tarballs, Next: , Previous: Introduction, Up: Top

Tarballs

There are a few mandatory requirements over the tarballs.


Node: Given Tarballs, Next: , Up: Tarballs

Given Tarballs

The naming scheme for provided tarballs is different from the scheme you must follow (see Delivery). Our naming scheme looks like 2004-tc-2.0.tar.bz2. If we update the tarballs, they will be named 2004-tc-2.x.tar.bz2. But your tarball must be named login-tc-2.tar.bz2, even if you send a second version of your project.

We also (try to) provide patches from one tarball to another. For instance 2005-tc-1.0-2.0.diff.bz2 is the difference from 2005-tc-1.0.tar.bz2 to 2005-tc-2.0.tar.bz2. You are encouraged to read this file as understanding a patch is expected from any Unix programmer. Just run bzless 2005-tc-1.0-2.0.diff.bz2.

To apply the patch:

  1. go into the top level of your current tarball
  2. remove any file which name might cause confusion afterwards (find . -name '*.orig' -o -name '*.rej' | xargs rm)
  3. run bzcat 2005-tc-1.0-2.0.diff.bz2 | patch -p1
  4. look for all the failures (find . -name '*.rej) and fix them by hand once you understood why the patch did not apply

You might need to repeat the process to jump from a version x to x + 2 via version x + 1.


Node: Coding Style, Next: , Previous: Given Tarballs, Up: Tarballs

Coding Style


Node: No Draft Allowed, Next: , Up: Coding Style

No Draft Allowed

The code you deliver must be clean. In particular, when some code is provided, and you have to fill in the blanks denoted by FIXME: Some code has been deleted.. Sometimes you will have to write the code from scratch.

In any case, dead code and dead comments must be removed. You are free to leave comments spotting places where you fixed a FIXME:, but never leave a fixed FIXME: in your code. Nor any irrelevant comment.

The official compiler for this project, is GNU C++ Compiler, 3.2 or higher (see GCC).


Node: Use of C++ Features, Next: , Previous: No Draft Allowed, Up: Coding Style

Use of C++ Features

Hunt code duplication
Code duplication is your enemy: the code is less exercised (if there are two routines instead of one, then the code is run half of the time only), and whenever an update is required, you are likely to forget to update all the other places. You should strive to prevent code duplication to sneak into your code. Every C++ feature is good to prevent code duplication: inheritance, templates etc.
Prefer dynamic_cast of references
Of the following two snippets, the first is preferred:
          const IntExp &ie = dynamic_cast <const IntExp &> (exp);
          int val = ie.value_get ();
          
          const IntExp *iep = dynamic_cast <const IntExp &> (exp);
          assert (iep);
          int val = iep->value_get ();
          

While upon type mismatch the second aborts, the first throws a std::bad_cast: they are equally safe.


Node: Use of STL, Next: , Previous: Use of C++ Features, Up: Coding Style

Use of STL

ESn refers to item n in Effective STL (see Bibliography).

Specify comparison types for associative containers of pointers (ES20)
For instance, instead of declaring
          typedef set::set<const Temp *> temp_set_t;
          

declare

     /** Object function to compare two Temp*. */
     struct temp_compare :
       public binary_function<const Temp *, const Temp*, bool>
     {
       bool
       operator() (const Temp *s1, const Temp *s2) const
       {
         return *s1 < *s2;
       }
     };
     
     typedef set::set<const Temp *, temp_compare> temp_set_t;
     

Scott Meyers mentions several good reasons, but leaves implicit a very important one: if you don't, since the outputs will be based on the order of the pointers in memory, and since (i) this order may change if your allocation pattern changes and (ii) this order depends of the environment you run, then you cannot compare outputs (including traces). Needless to say that, at least during development, this is a serious misfeature.

Prefer algorithm call to hand-written loops (ES43)
Using for_each, find, find_if, transform etc. is preferred over explicit loops. This is for (i) efficiency, (ii) correctness, and (iii) maintainability. Knowing these algorithms is mandatory for who claims to be a C++ programmer.
Prefer member functions to algorithms with the same names (ES44)
For instance, prefer my_set.find (my_item) to find (my_item, my_set.begin (), my_set.end ()). This is for efficiency: the former has a logarithmic complexity, versus... linear for the latter! You may find the Item 44 of Effective STL on the Internet.


Node: File Conventions, Next: , Previous: Use of STL, Up: Coding Style

File Conventions

There are some strict conventions to obey wrt the files and their contents.

Declarations in *.hh
The *.hh should contain only declarations, i.e., prototypes, extern for variables etc. Inlined short methods are accepted when there are few of them, otherwise, create an *.hxx file. The documentation should be here too.

There is no good reason for huge objects to be defined here.

Inlined definitions in *.hxx
If there are definitions that should be loaded in different places (definitions of templates, inline functions etc.), then declare and document them in the *.hh file, and implement them in the *.hxx file.
Definitions of functions and variables in *.cc
Big objects should be defined in the *.cc file corresponding to the declaration/documentation file *.hh.
lib*.hh and lib*.cc are pure
There should be only pure functions in the interface of a module. That means that the functions in these files should not depend upon globals, nor have side effects of global objects. Of course no global variable can be defined here either.
*-tasks.hh and *-tasks.cc are impure
Tasks, as designed currently, are the place for side effects. That's where globals such as the current ast, the current assembly program, etc., are defined and modified.


Node: Matters of Style, Previous: File Conventions, Up: Coding Style

Matters of Style

The following items are more a matter of style than the others. Nevertheless, you are asked to follow this style.

Prefer Doxygen Documentation to plain comments
We use Doxygen (see Doxygen) to maintain the developer documentation of the Tiger Compiler.
Write Documentation in Doxygen
Documentation is a genuine part of programming, just as testing. The quality of this documentation can change the grade.
Use \directive
Prefer backslash (\) to the commercial at (@) to specify directives.
Prefer C Comments for Long Comments
Prefer C comments (/** ... */) to C++ comments (/// ...). This is to ensure consistency with the style we use.
Prefer C++ Comments for One Line Comments
Because it is lighter, instead of
               /** \brief Name of this program. */
               extern const char *program_name;
               

prefer

               /// Name of this program.
               extern const char *program_name;
               

For instance, instead of

     /* Construct an InterferenceGraph. */
     InterferenceGraph (const std::string &name,
                        const assem::instrs_t& instrs, bool trace = false);
     

or

     /** @brief Construct an InterferenceGraph.
      ** @param name    its name, hopefully based on the function name
      ** @param instrs  the code snippet to study
      ** @param trace   trace flag
      **/
     InterferenceGraph (const std::string &name,
                        const assem::instrs_t& instrs, bool trace = false);
     

or

     /// \brief Construct an InterferenceGraph.
     /// \param name    its name, hopefully based on the function name
     /// \param instrs  the code snippet to study
     /// \param trace   trace flag
     InterferenceGraph (const std::string &name,
                        const assem::instrs_t& instrs, bool trace = false);
     

write

     /** \brief Construct an InterferenceGraph.
         \param name    its name, hopefully based on the function name
         \param instrs  the code snippet to study
         \param trace   trace flag
       */
     InterferenceGraph (const std::string &name,
                        const assem::instrs_t& instrs, bool trace = false);
     

Of course, Doxygen documentation is not appropriate everywhere.

Use foo_get, not get_foo
Accessors have standardized names: foo_get and foo_set.


Node: Delivery, Next: , Previous: Coding Style, Up: Tarballs

Delivery

Each group must provide a tarball, made via make distcheck. All the information about the delivery per se is given on the Yaka's Delivery Page.

If bardec_f is the head of your group, the tarball must be bardec_f-tc-n.tar.bz2 where n is the number of the "release" (see Package Name and Version). The following commands must work properly:

     $ bunzip2 -cd bardec_f-tc-n.tar.bz2 | tar xvf -
     $ cd bardec_f-tc-n
     $ export CC=gcc-3.2
     $ export CXX=g++-3.2
     $ ./configure
     $ make
     $ cd src
     $ ./tc /tmp/test.tig
     $ cd ..
     $ make distcheck
     

For more information on the tools, see The GNU Build System, GCC.

Your tarball must be done via make distcheck (see Making a Tarball). Any tarball which is not built thanks to make distcheck (this is easy to see: they include files we don't want, and don't contain some files we need...) will be penalized with at least ### tarball_not_clean.


Node: Project Layout, Next: , Previous: Delivery, Up: Tarballs

Project Layout

This section describes the mandatory layout of the tarball.


Node: The Top Level, Next: , Up: Project Layout

The Top Level

AUTHORS
In the top level of the distribution, there must be a file AUTHORS which contents is as follows:
          Fabrice Bardèche     <bardec_f@epita.fr>
          Jean-Paul Sartre     <sartre_j@epita.fr>
          Jean-Paul Deux       <deux_j@epita.fr>
          Jean-Paul Belmondo   <belmon_j@epita.fr>
          

The group leader is the first in the list. Do not include emails other than those of EPITA. I repeat: give the 6_1@epita.fr address. Note that the file AUTHORS is automatically distributed, but pay attention to the spelling.

ChangeLog
Optional. The list of the changes made in the compiler, with the dates and names of the people who worked on it. See the Emacs key binding C-x 4 a.
README
Various free information.
argp/
The command line parser we use.
src/
All the sources are in this directory.
tests/
Your own test suite. You should make it part of the project, and ship it like the rest of the package. Actually, it is abnormal not to have a test suite here.


Node: src, Next: , Previous: The Top Level, Up: Project Layout

The src Directory

common.hh (src/) File
Used throughout the project.

tc (src/) File
Your compiler.

tc.cc (src/) File
Main entry. Called, the driver.


Node: src/misc, Next: , Previous: src, Up: Project Layout

The src/misc Directory

Convenient C++ routines.

contract.hh (src/misc/) File
A useful improvement over cassert.

escape.hh (src/misc/) File
This file implements a means to output string while escaping non printable characters. An example:
            cout << "escape (\"\111\") = " << escape ("\"\111\"") << endl;
          

Understanding how escape works is required starting from T2.

set.hh (src/misc/) File
A wrapper around std::set that introduce convenient operators (operator+ and so forth).

timer.hh (src/misc/) File
timer.cc (src/misc/) File
A class that makes it possible to have timings of the compilation process, as when using --time-report with gcc, or --report=time with bison. It is used in the Task machinery, but can be used to provide better timings (e.g., separating the scanner from the parser).


Node: src/task, Next: , Previous: src/misc, Up: Project Layout

The src/task Directory

No namespace for the time being, but it should be task. Delivered for T1. A generic scheme to handle the components of our compiler, and their dependencies.


Node: src/symbol, Next: , Previous: src/task, Up: Project Layout

The src/symbol Directory

Namespace symbol, delivered for T1 or T2.

symbol.hh (src/symbol/) File
The handling of the symbols.

table.hh (src/symbol/) File
The handling of generic symbol tables, i.e., it is independent of functions, types and variables.


Node: src/ast, Next: , Previous: src/symbol, Up: Project Layout

The src/ast Directory

Namespace ast, delivered for T2. Implementation of the abstract syntax tree. The file ast/README gives an overview of the involved class hierarchy.

location.hh (src/ast/) File
position.hh (src/ast/) File
These files are now simply forwarding the definitions of yy::Position and yy::Location as provided by Bison.

visitor.hh (src/ast/) File
Abstract base class of the compiler's visitor hierarchy. Actually, it defines a class template GenVisitor, which expects an argument which can be either non_const_kind or const_kind. This allows to define to parallel hierarchies: ConstVisitor and Visitor, similar to iterator and const_iterator.

The understanding of the template programming used is not required at this stage as it is quite delicate, and goes far beyond your (average) current understanding of templates.

default-visitor.hh (src/ast/) File
Implementation of the DefaultVisitor class, which walks the abstract syntax tree, doing nothing. It is mainly used as a basis for deriving other visitors. Actually, just as above, there is a template, so that we have two different default visitors: DefaultVisitor<const_kind> and DefaultVisitor<non_const_kind>.

print-visitor.hh (src/ast/) File
Implementation of the PrintVisitor class, which performs pretty-printing in the tiger compiler.


Node: src/parse, Next: , Previous: src/ast, Up: Project Layout

The src/parse Directory

Namespace parse. Delivered during T1.

scantiger.ll (src/parse/) File
The scanner.

parsetiger.yy (src/parse/) File
The parser.

position.hh (src/ast/) File
Keeping track of a point (cursor) in a file.

location.hh (src/ast/) File
Keeping track of a range (two cursors) in a (or two) file.

libparse.hh (src/ast/) File
which prototypes what tc.cc needs to know about the module parse.


Node: src/type, Next: , Previous: src/parse, Up: Project Layout

The src/type Directory

Namespace type. Type checking.

libtype.hh (src/type/) File
The interface of the Type module. It exports a single procedure, type_check.

types.hh (src/type/) File
The definition of all the types. You are free to use whatever layout you wish (several files); we have a single types.hh file.

type-entry.hh (src/type/) File
Definitions of type::TypeEntry, type::VarEntry, and type::FunEntry, used in type::TypeEnv to associate data to types, variables, and functions (obviously).

type-env.hh (src/type/) File
The types environment, comprising three symbol tables: types, functions, and variables, used by the type::TypeVisitor.


Node: src/temp, Next: , Previous: src/type, Up: Project Layout

The src/temp Directory

Namespace temp, delivered for T5.

temp.hh (src/temp/) File
So called temporaries are pseudo-registers: we may allocate as many temporaries as we want. Eventually the register allocator will map those temporaries to either an actual register, or it will allocate a slot in the activation block (aka frame) of the current function.

label.hh (src/temp/) File
We need labels for jumps, for functions, strings etc.


Node: src/tree, Next: , Previous: src/temp, Up: Project Layout

The src/tree Directory

Namespace tree, delivered for T5. The implementation of the intermediate representation. The file tree/README should give enough explanations to understand how it works.

Reading the corresponding explanations in Appel's book is mandatory.

It is worth noting that contrary to A. Appel, just as we did for ast, we use n-ary structures. For instance, where Appel uses a binary seq, we have an n-ary seq which allows us to put as many statements as we want.

To avoid gratuitous name clashes, what Appel denotes exp is denoted sxp (Statement Expression), implemented in translate::Sxp.

Please, pay extra attention to the fact that there are temp::Temp used to create unique temporaries (similar to symbol::Symbol), and tree::Temp which is the intermediate representation instruction denoting a temporary (hence a tree::Temp needs a temp::Temp). Similarly, on the one hand, there is temp::Label which is used to create unique labels, and on the other hand there are tree::Label which is the IR statement to define to a label, and tree::Name used to refer to a label (typically, a tree::Jump needs a tree::Name which in turn needs a temp::Label).


Node: src/frame, Next: , Previous: src/tree, Up: Project Layout

The src/frame Directory

Namespace frame, delivered for T5.

access.hh (src/frame/) File
access.cc (src/frame/) File
An Access is a location of a variable: on the stack, or in a temporary.

frame.hh (src/frame/) File
frame.cc (src/frame/) File
A Frame knows only what are the "variables" it contains.


Node: src/translate, Next: , Previous: src/frame, Up: Project Layout

The src/translate Directory

Namespace translate. Translation to intermediate code translation. It includes:

libtranslate.hh (src/translate/) File
The interface.

libtranslate.cc (src/translate/) File
The compiled module.

fragment.hh (src/translate/) File
It implements translate::Fragment, an abstract class, translate::DataFrag to store the literal strings, and translate::ProcFrag to store the routines.

access.hh (src/translate/) File
access.cc (src/translate/) File
Static link aware versions of level::Access.

level.hh (src/translate/) File
level.cc (src/translate/) File
translate::Level are wrappers frame::Frame that support the static links, so that we can find an access to the variables of the "parent function".

exp.hh (src/translate/) File
Implementation of translate::Ex (expressions), Nx (instructions), Cx (conditions), and Ix (if) shells. They wrap tree::Node to delay their translation until the actual use is known.

level-entry.hh (src/translate/) File
All the information that the environment must keep about variables and functions.

level-env.hh (src/translate/) File
The levels environment, containing LevelVarEntry's and LevelFunEntry's. We don't need to store information related to types here.

translation.hh (src/translate/) File
functions used by the translate::TranslateVisitor to translate the AST into HIR. For instance, it contains Exp *simpleVar (const Access &access, const Level &level), Exp *callExp (const temp::Label &label, std::list<Exp *> args) etc. which are routines that produce some Tree::Exp. They handle all the unCx etc. magic.

translate-visitor.hh (src/translate/) File
Implements the class TranslateVisitor which performs the IR generation thanks to translation.hh. It must not be polluted with translation details: it is only coordinating the AST traversal with the invocation of translation routines. For instance, here is the translation of a ast::SimpleVar:
          virtual void visit (const SimpleVar& e)
          {
            const Access &access = _env.var_access_get (e.name_get ());
            _exp = simpleVar (access, *_level);
          }
          


Node: src/canon, Next: , Previous: src/translate, Up: Project Layout

The src/canon Directory

Namespace tree.


Node: src/assem, Next: , Previous: src/canon, Up: Project Layout

The src/assem Directory

Namespace assem, delivered for T7.

This directory contains the implementation of the Assem language: yet another intermediate representation that aims at encoding an assembly language, plus a few need features so that register allocation can be performed afterwards. Given in full.

instr.hh (src/assem/) File
move.hh (src/assem/) File
oper.hh (src/assem/) File
label.hh (src/assem/) File
Implementation of the basic types of assembly instructions.

fragment.hh (src/assem/) File
fragment.cc (src/assem/) File
Implementation of assem::Fragment, assem::ProcFrag, and assem::DataFrag. They are comparable to translate::Fragment: aggregate some informations that must remain together, such as a frame::Frame and the instructions (a list of assem::Instr).

visitor.hh (src/assem/) File
The root of assembler visitors.

layout.hh (src/assem/) File
A pretty printing visitor for assem::Fragment.

libassem.hh (src/assem/) File
libassem.cc (src/assem/) File
The interface of the module, and its implementation.


Node: src/target, Next: , Previous: src/assem, Up: Project Layout

The src/target Directory

Namespace target, delivered for T7. Some data on the back end. Given in full.

cpu.hh (src/target/) File
Description of a CPU: everything about its registers, and its word size.

target.hh (src/target/) File
Description of a target (language): its CPU, its assembly (codegen::Assembly), and it translator (codegen::Codegen).

mips-cpu.hh (src/target/) File
mips-target.hh (src/target/) File
The description of the MIPS (actually, SPIM/Mipsy) target.

ia32-cpu.hh (src/target/) File
ia32-target.hh (src/target/) File
Description of the i386. This is not part of the project, it is left only as an incomplete source of inspiration.

target-tasks.cc (src/target/) File
target-tasks.hh (src/target/) File
The command line interface to specify the target architecture.


Node: src/codegen, Next: , Previous: src/target, Up: Project Layout

The src/codegen Directory

Namespace codegen, delivered for T7.

mips (src/codegen/) File
ia32 (src/codegen/) File
The instruction selection per se split into a generic part, and a target specific (MIPS and IA32) part. See src/codegen/mips, and src/codegen/ia32.

assembly.hh (src/codegen/) File
The abstract class codegen::Assembly which is the interface for elementary assembly instructions generation.

codegen.hh (src/codegen/) File
The abstract class codegen::Codegen which is the interface for all our back ends.

libcodegen.hh (src/codegen/) File
libcodegen.cc (src/codegen/) File
Converting translate::Fragments into assem::Fragments.

codegen-tasks.hh (src/codegen/) File
codegen-tasks.cc (src/codegen/) File
Command line interface.

tiger-runtime.c (src/codegen/) File
This is the Tiger runtime, written in C, based on Andrew Appel's runtime.c. The actual runtime.s file for MIPS was written by hand, but the ia32 was a compiled version of this file. It should be noted that:
Strings
Strings are implemented as 4 bytes to encode the length, and then a 0-terminated a` la C string. The length part is due to conformance to the Tiger Reference Manual, which specifies that 0 is a regular character that can be part of the strings, but it is nevertheless terminated by 0 to be compliant with SPIM/Mipsy's print syscall. This might change in the future.
Special Strings
There are some special strings: 0 and 1 character long strings are all implemented via a singleton. That is to say there is only one allocated string "", a single "1" etc. These singletons are allocated by main. It is essential to preserve this invariant/convention in the whole runtime.
strcmp vs. stringEqual
I don't know how Appel wants to support "bar" < "foo" since he doesn't provide strcmp. We do. But note that anyway his implementation of "foo" != "fooo" is more efficient than ours, since he can decide just be looking at the lengths. That could be improved in the future...
main
The runtime has some initializations to make, such as strings singletons, and then calls the compiled program. This is why the runtime provides main, and calls t_main, which is the "main" that your compiler should provide.


Node: src/codegen/mips, Next: , Previous: src/codegen, Up: Project Layout

The src/codegen/mips Directory

Namespace codegen::mips, delivered for T7. Code generation for MIPS R2000.

runtime.s (src/codegen/mips/) File
runtime.cc (src/codegen/mips/) File
The Tiger runtime in MIPS assembly language: print etc. The C++ file runtime.cc is built from runtime.s: do not edit the former. See src/codegen, tiger-runtime.

spim-assembly.hh (src/codegen/mips/) File
spim-assembly.cc (src/codegen/mips/) File
Our assembly language (syntax, opcodes and layout); it abstracts the generation of MIPS 2000 instructions. codegen::mips::SpimAssembly derives from codegen::Assembly.

codegen.hh (src/codegen/mips/) File
codegen.cc (src/codegen/mips/) File
Our real and only back end: a translator from LIR to ASSEM using the MIPS 2000 instruction set defined by codegen::mips::SpimAssembly. It is implemented as a maximal munch. codegen::mips::Codegen derives from codegen::Codegen.

spim-layout.hh (src/codegen/mips/) File
spim-layout.cc (src/codegen/mips/) File
How MIPS (and SPIM/Mipsy) fragments are to be displayed. In other words, that's where the (global) syntax of the target assembly file is selected.


Node: src/codegen/ia32, Next: , Previous: src/codegen/mips, Up: Project Layout

The src/codegen/ia32 Directory

Namespace codegen::ia32, delivered for T7. Code generation for IA32. This is not part of the student project, but it is left to satisfy their curiosity. In addition its presence is a sane invitation to respect the constraints of a multi-back-end compiler.

runtime.s (src/codegen/ia32/) File
runtime.cc (src/codegen/ia32/) File
The Tiger runtime in IA32 assembly language: print etc. The C++ file runtime.cc is built from runtime.s: do not edit the former. See src/codegen, tiger-runtime.

gas-assembly.hh (src/codegen/ia32/) File
gas-assembly.cc (src/codegen/ia32/) File
Our assembly language (syntax, opcodes and layout); it abstracts the generation of IA32 instructions using Gas' syntax. codegen::ia32::GasAssembly derives from codegen::Assembly.

codegen.hh (src/codegen/ia32/) File
codegen.cc (src/codegen/ia32/) File
The IA32 back-end: a translator from LIR to ASSEM using the IA32 instruction set defined by codegen::ia32::GasAssembly. It is implemented as a maximal munch. codegen::ia32::Codegen derives from codegen::Codegen.

gas-layout.hh (src/codegen/ia32/) File
gas-layout.cc (src/codegen/ia32/) File
How IA32 fragments are to be displayed. In other words, that's where the (global) syntax of the target assembly file is selected.


Node: src/graph, Next: , Previous: src/codegen/ia32, Up: Project Layout

The src/graph Directory

Namespace graph, a generic implementation of graphs. Delivered for T7.

graph.hh (src/graph/) File
graph.hxx (src/graph/) File
Oriented and undirected graphs.

handler.hh (src/graph/) File
handler.hxx (src/graph/) File
Abstractions/indirections for graph nodes and edges.

iterator.hh (src/graph/) File
iterator.hxx (src/graph/) File
Iterating over nodes and edges of graphs.

test-graph.cc (src/graph/) File
Exercising this nodule.


Node: src/liveness, Next: , Previous: src/graph, Up: Project Layout

The src/liveness Directory

Namespace liveness, delivered for T8.

flowgraph.hh (src/liveness/) File
FlowGraph implementation.

test-flowgraph.cc (src/liveness/) File
FlowGraph test.

liveness.hh (src/liveness/) File
liveness.cc (src/liveness/) File
Computing the live-in and live-out information from the FlowGraph.

interference-graph.hh (src/liveness/) File
interference-graph.cc (src/liveness/) File
Computing the InterferenceGraph from the live-in/live-out information.


Node: src/regalloc, Previous: src/liveness, Up: Project Layout

The src/regalloc Directory

Namespace regalloc, register allocation, delivered for T9.

color.hh (src/regalloc/) File
Coloring an interference graph.

regallocator.hh (src/regalloc/) File
Repeating the coloration until it succeeds (no spills).

libregalloc.hh (src/regalloc/) File
libregalloc.cc (src/regalloc/) File
Removing useless moves once the register allocation performed, and allocating the register for fragments.

test-regalloc.cc (src/regalloc/) File
Exercising this.

regalloc-tasks.hh (src/regalloc/) File
regalloc-tasks.cc (src/regalloc/) File
Command line interface.


Node: Given Test Cases, Previous: Project Layout, Up: Tarballs

Given Test Cases

We provide a few test cases: you must write your own tests. Writing tests is part of the project. Do not just copy test cases from other groups, as you will not understand why they were written.

The initial test suite is available for download at tests.tgz. It contains the following directories:

good
These programs are correct.
scan
These programs have lexical errors.
parse
These programs have syntax errors.
type
These programs contain type mismatches.


Node: Evaluation, Next: , Previous: Tarballs, Up: Top

Evaluation

Some stages are evaluated only by a program, and others are evaluated both by humans, and a program.


Node: Automated Evaluation, Next: , Up: Evaluation

Automated Evaluation

Each stage of the compiler will be evaluated by an automatic corrector. As soon as the tarball are delivered, the logs are available on http://www.lrde.epita.fr/~akim/compil, in the directory corresponding to your class and stage. For instance, 2004 students ought to read http://www.lrde.epita.fr/~akim/compil/2004/4/bardec_f-tc-4.log.

We stress that automated evaluation enforces the requirements: you must stick to what is being asked. For instance, for T3 it is explicitly asked to display something like:

     var /* escaping */ i : int := 2
     

so if you display any of the following outputs

     var i : int /* escaping */ := 2
     var i /* escaping */ : int := 2
     var /* Escapes */ i : int := 2
     

be sure to fail all the tests, even if the computation is correct.

If you find some unexpected errors (your project does compile with the reference compiler, some files are missing, your output is slightly incorrect etc.) immediately send a new tarball to yaka@epita.fr with [Tiger] as prefix of the subject. This corresponds to ### patch.

Do not wait for the final marks to be computed, this is extremely irritating, and doomed to failure. You must understand that (i) you increase our workload, and (ii) anyway this is the wrong approach, the Tiger Compiler is a big project which must be continuously improved.

If, anyway, you send a tarball to fix your problems long after the initial date, you will be flagged as ### super_late, which impact on the mark is quite bad...


Node: During the Examination, Next: , Previous: Automated Evaluation, Up: Evaluation

During the Examination

When you are defending your projects, here are a few rules to follow:

Don't talk
Don't talk unless you are asked to: when a person is asked a question, s/he is the only one to answer. You must not talk to each other either: often, when one cannot answer a question, the question is asked to another member. It is then obvious why the members of the group shall not talk.
Don't touch the screen
Don't touch my display! You have nice fingers, but I don't need their prints on my screen.
Tell the truth
If there is something the examiner must know (someone did not work on the project at all, some files are coming from another group etc.), say it immediately, for, if we discover that by ourselves, you will be severely sanctioned.
Learn
It is explicitly stated that you can not have worked on a stage provided this was an agreement with the group. But it is also explicitly stated that you must have learned what was to be learned from that compiler stage, which includes C++ techniques, Bison and Flex mastering, object oriented concepts, design patterns and so forth.
Complain now!
If you don't agree with the notation, say it immediately. Private messages about "this is unfair: I worked much more than bardec_f but his grade is better than mine" are thrown away.

Conversely, there is something I wish to make clear: I, Akim, and the other examiners, will probably be harsh (maybe even very harsh), but this does not mean I disrespect you, or judge you badly.

You are here to defend your project and knowledge, I'm here to stress them, to make sure they are right. Learning to be strong under pressure is part of the exercise. Don't burst into tears, react! Don't be shy, that's not the proper time: you are selling me something, and I will never buy something from someone who cries when I'm criticizing his product.

You should also understand that human examination is the moment where we try to evaluate who, or what group, needs help. We are here to diagnose your project and provide solutions to your problems. If you know there is a problem in your project, but you failed to fix it, tell it to the examiner! Work with her/him to fix your project.


Node: Human Evaluation, Next: , Previous: During the Examination, Up: Evaluation

Human Evaluation

The point of this evaluation is to measure:

the quality of the code
How clean it is, amount of code duplication, bad hacks, standards violations (e.g., stderr is forbidden in proper C++ code) and so forth. It also aims at detecting cheaters, who will be severely punished (mark = -42).
the knowledge each member acquired
While we do not require that each member worked on a stage, we do require that each member (i) knows how the stage works and (ii) has perfectly understood the (C++, Bison etc.) techniques needed to implement the stage.

In results in an evaluation file (login-tc-stage.eval, e.g., bardec_f-tc-2.eval), structured as follows.

The format is free, except some lines starting with specific markers, when they are at the first column. The first marker is *, used to denote the login of the group head:

     * bardec_f
     ----------------------------------------
     2002-03-07: T2 par akim
     tc-check: Summary:
     tc-check: == Testing ./tc --parse-trace -l --ast-display ==
     Successes: 177/177
     
     Suppléments:
     - --stdin pour parser stdin.
     - Le parser fait attention à ses leaks, même sur les Symbol.
     - --gcc-ast
     - ChangeLog
     - Ne meurt pas quand on lui donne un répertoire en entrée.
     
     Le code est joli, et en 80 col !
     Plutôt que des listes vides, passent 0.
     

As you can see, the person who evaluated is allowed to put whatever comment comes to her mind. Specific features, specific failures, explanations to help the students to fix their project should be included. Each time an entry is made, there should be the date, and who wrote it (2002-03-07: T2 par akim).

Then, the mark of the project, aka, the note of the group:

     ### Note T2 = 20
     

and optionally:

### late
The code was delivered reasonably late (the same day).
### super_late
More than one day.
### patch
Within 24 hours from the first delivery date, the group submitted another (hopefully) fixed tarball.
### super_patch
Before the second delivery date, the group submitted another tarball.
### not_compile
It doesn't compile, and the examiner had to change the code.
### tarball_not_clean
Clear enough: bad name, bad content etc.
### cheater
The group has cheated. The examiner is then asked to write as many details as possible, and to flag the group as cheaters. The mark is likely to be negative, and exclusion from EPITA is possible.
### reevaluation
The group had a mark < 8, and is now being reexamined.
### bonus
The group has implemented additional features, such as XML output etc. and the project works properly. We do not care about extra features on top of a broken project.
Important note to the examiners: Note field.

The examiner should not take (too much) the automated tests into account to decide the mark. This is because the mark is computed later, taking this into account, so don't do it twice. Similarly, a very nice project, which is super late, shall be flagged as super late, but should have a good ## Note entry.

Important note to the examiners: broken tarballs.

If you fixed the tarball (### not_compile, or ### tarball_not_clean or whatever modification, you must run make distcheck again, and replace the tarball they delivered with the new one. Do not keep the old tarball, do not install it in a special place: just replace the first tarball with it, but say so in the eval file.

The rationale is simple: only tarballs pass the tests, and every tarball must be able to pass the tests. If you don't do that, then someone else will have to do it again.

The next bits is the evaluation of each member of the group:

     ----------------------------------------
     ## bardec_f:100:120
     Parser, locations, scanner.  A pratiquement e'crit tout T2.
     
     ##sartre_j:100:40
     AST, Visitor
     Pourtant ne comprend rien au Visitor.  Se fout de ma gueule.
     
     ##deux_j:80:80
     Pas de problème C++.
     FIXME: Sa note de T1 est à revoir (resoutenance).
     
     ##belmon_j:100:100
     Actions, les delete dans le parser.
     A implémenté la sortie en AST compatible avec GCC.
     

The formalism is ## login:bonus1:bonus2. If the stage is T2, then bonus1 refers to a bonus on T1, and bonus2 on T2. For T4, it covers T3 and T4, and so forth.

The values can be "negative" (80, 60, 40, 0), or positive if a member deserves it (100, 120, or even more if justifiable).

There should be information about who did what.


Node: Marks Computation, Next: , Previous: Human Evaluation, Up: Evaluation

Marks Computation

Because the Tiger Compiler is a project with stages, the computation of the marks depends on the stages too. To spell it out explicitly:

A stage is penalized by bad results on tests performed for previous stages.

It means, for instance, that a T3 compiler will be exercised on T1, T2, and T3. If there are still errors on T1 and T2 tests, they will pessimize the result of T3 tests. The older the errors are, the more expensive they are.

As an example, here are the formulas to compute the global success rate of T3 and T5:

     global-rate-T3 := rate-T3 * (+ 2 * rate-T1
                                  + 1 * rate-T2) / 3
     global-rate-T5 := rate-T5 * (+ 4 * rate-T1
                                  + 3 * rate-T2
                                  + 2 * rate-T3
                                  + 1 * rate-T4) / 10
     

Because a project which fail half of the time is not a project that deserves half of 20, the global-rate is elevated to 1.7 before computing the mark:

     mark-T3 := roundup (power (global-rate-T3, 1.7) * 20 - malus-T3, 1)
     

where roundup (x, 1) is x rounded up to one decimal (roundup (15, 1) = 15, roundup (15.01, 1) = 15.1).

When the project is also evaluated by a human, power is not used. Rather, the success rate modifies the mark given by the examiner:

     mark-T2 := roundup (eval-T2 * global-rate-T2 - malus-T2, 1)
     


Node: Reevaluation, Previous: Marks Computation, Up: Evaluation

Reevaluation

It happens that some groups have big problems with their project. Here is how to solve these issues.

But first of all, you must know that cheating, stealing another group's project, is not a solution, because:


Because the Tiger Compiler is a long project, you continuously have to improve it, but marks starting from 8 to higher are definitive. If your mark is less than 8, can be re-examined: you must provide a better tarball, and in the case of human evaluated stages, another audition will be made.

This new audition will be flagged as ### reevaluation. It cannot be better than 12/20, even if your project is excellent: a good project out of date cannot be judged better than a reasonable good project on time.

Pay attention that when providing an updated tarball for reexamination of stage n, it must follow the naming scheme of stage n, even if the contents goes further. For instance, it is perfectly valid to ask for a T2 reevaluation with a compiler that implements T4, but the tarball must be bardec_f-tc-2.tar.bz2.

You have to ask for a reevaluation, we will not look after you.


Node: Compiler Stages, Next: , Previous: Evaluation, Up: Top

Compiler Stages

The compiler will be written in several steps, described below.


Node: T0, Next: , Up: Compiler Stages

T0, Naive Scanner and Parser

This section has been updated for EPITA-2005.

T0 is a weak form of T1: the scanner and the parser are written, but there is a set of simplifications:

the hierarchy
We don't need several directories, you can program in the top level of the package.
the build
The GNU Build System is not used: there is no need for Autoconf, Automake etc.
no driver
The driver is reduced to its simplest form:
          int
          main (int argc, const char *argc[])
          {
            assert (argc == 1);
            yyin = fopen (argv[1]);
            assert (yyin);
            return !!yyparse ();
          }
          

i.e., there is no support for options at all.

SCAN, PARSE
There is no support for options, but that does not mean that you cannot use the tracing/debugging features from Flex/Bison. It is required that you bind tracing to the environment variables SCAN and PARSE. I.e., running
          PARSE=1 ./tc foo.tig
          

will set yydebug to 1, which causes the traces of the parsing to be displayed.

YYPRINT
There is no requirement to implement YYPRINT support.
yylval
All the values are supported: strings, integers and even symbols. Nevertheless, symbols are returned as plain strings for the time being: the class symbol::Symbol is to be implemented in T1.


Node: T0 Goals, Next: , Up: T0

T0 Goals

Things to learn during this stage that you should remember:


Node: T0 Code to Write, Previous: T0 Goals, Up: T0

T0 Code to Write

You must write

scantiger.ll
The scanner.
parsetiger.yy
The parser, and maybe main if you wish.
tc.cc
Optionally, you may write your driver, i.e., main, in this file. Putting it into parsetiger.yy is OK in T0.
Makefile
This file is mandatory. Running make must build the binary tc.

The requirements on the tarball are the same as usual, see Tarballs.


Node: T1, Next: , Previous: T0, Up: Compiler Stages

T1, Scanner and Parser

2005-T1 delivery is Friday, February 14th 2003 at noon.
This section is updated for EPITA-2005.

Scanner and parser are properly running, but the abstract syntax tree is not built yet. Differences with T0 include:

GNU Build System
Autoconf, Automake are used.
Options, Tasks
The compiler supports basic options via in the Task module.
locations
The locations are computed are properly reported in the error messages


Node: T1 Goals, Next: , Up: T1

T1 Goals

Things to learn during this stage that you should remember:


Node: T1 Samples, Next: , Previous: T1 Goals, Up: T1

T1 Samples

The only information the compiler will give is about lexical and syntax errors.

If there are no errors, the compiler shuts up, and exits successfully:

     /* an array type and an array variable */
     let
       type  arrtype = array of int
       var arr1 : arrtype := arrtype [10] of 0
     in
       arr1[2]
     end
     File 4: test01.tig
     
     $ tc test01.tig
     Example 5: tc test01.tig
     

If there are lexical errors, the exit status is 2, and a an error message is output on the standard error output. Note that its format is standard: file, (precise) location, and then the message.

     1
      /* This comments starts at /* 2.2 */
     File 6: unterminated-comment.tig
     
     $ tc unterminated-comment.tig
     error-->unterminated-comment.tig:2.1-3.0: unexpected end of file in a comment
     =>2
     Example 7: tc unterminated-comment.tig
     

If there are syntax errors, the exit status is set to 3:

     let var a : nil := ()
     in
       1
     end
     File 8: type-nil.tig
     
     $ tc type-nil.tig
     error-->type-nil.tig:1.12-14: syntax error, unexpected "nil", expecting "identifier"
     error-->Parsing Failed
     =>3
     Example 9: tc type-nil.tig
     

If there are errors which are non lexical, nor syntactic (Windows will not pass by me):

     $ tc C:/TIGER/SAMPLE.TIG
     error-->tc: cannot open `C:/TIGER/SAMPLE.TIG': No such file or directory
     =>1
     Example 10: tc C:/TIGER/SAMPLE.TIG
     

The option --parse-trace, which relies on Bison's %debug directive, and the use of YYPRINT, must work properly:

     $ cat foo.tig
     a + "a"
     $ ./tc --parse-trace foo.tig
     Starting parse
     Entering state 0
     Reading a token: Next token is 258 (ID a)
     Shifting token 258 (ID), Entering state 2
     Reading a token: Next token is 270 (PLUS)
     Reducing via rule 74 (line 318), ID  -> varid
     state stack now 0
     Entering state 16
     Reducing via rule 34 (line 196), varid  -> lvalue
     state stack now 0
     Entering state 13
     Next token is 270 (PLUS)
     Reducing via rule 33 (line 192), lvalue  -> exp
     state stack now 0
     Entering state 12
     Next token is 270 (PLUS)
     Shifting token 270 (PLUS), Entering state 30
     Reading a token: Next token is 257 (STRING a)
     Shifting token 257 (STRING), Entering state 1
     Reducing via rule 15 (line 151), STRING  -> exp
     state stack now 0 12 30
     Entering state 65
     Reading a token: Now at end of input.
     Reducing via rule 27 (line 182), exp PLUS exp  -> exp
     state stack now 0
     Entering state 12
     Now at end of input.
     Reducing via rule 1 (line 106), exp  -> program
     state stack now 0
     Entering state 159
     Now at end of input.
     Shifting token 0 ($), Entering state 160
     Now at end of input.
     $ echo $?
     =>0
     

Note that (i), it cannot see that the variable is not declared nor that there is a type checking error, since type checking... is not implemented, and (ii), the output might be slightly different, depending upon the version of Bison you use. But what matters is that one can see the items: ID a, STRING a.


Node: T1 Given Code, Next: , Previous: T1 Samples, Up: T1

T1 Given Code

Some code is provided: 2005-tc-1.0.tar.bz2. See The Top Level, src, src/parse, src/misc.


Node: T1 Code to Write, Next: , Previous: T1 Given Code, Up: T1

T1 Code to Write

Be sure to read Flex and Bison documentations and tutorials, see Flex & Bison.

src/parse/scantiger.ll
The scanner must be completed. It must be able to read strings, identifiers etc. Strings and identifiers will be stored as C++ std::string. See the following code for the basics.
          ...
          \"          yylval->str = new std::string (); BEGIN STATE_STRING;
          
          <STATE_STRING>{ /* Handling of the strings.  Initial " is eaten. */
               \" {
                 BEGIN INITIAL;
                 return STRING;
               }
          ...
               \\x[0-9a-fA-F]{2}  {
                 yylval->str->append (1, strtol (yytext + 2, 0, 16));
               }
          ...
          }
          

The locations are tracked.

src/parse/parsetiger.yy
The grammar must be complete, but do not implement actions. You have to implement support for --parse-trace (see T1 Samples). Pay special attention to the display of strings and identifiers.

Bison will certainly complain because of a type clash for some actions. For instance, if you have given a type to STRING, but none to exp, then it will choke on:

          exp: STRING;
          

because it actually means

          exp: STRING { $$ = $1; };
          

which is not type coherent. So write this instead:

          exp: STRING {};
          

src/ast/position.hh
The class ast::Position is completed.
src/ast/location.hh
The class ast::Location must be completed.
src/symbol/symbol.hxx
The class symbol::Symbol keeps a single copy of identifiers, so that (i) we save space, and (ii) symbol comparison is fast. The file src/symbol/symbol.hh describes the interface of the class symbol::Symbol, but the implementation is to be written in src/symbol/symbol.hxx.


Node: T1 FAQ, Previous: T1 Code to Write, Up: T1

T1 FAQ

Options
The first version of argp we used (1.1) was buggy: on ./tc foo.tig -A, it considered -A was a file, not an option. This bug is now fixed. The patch is from Niels Möller:
          diff -u -r1.16 -r1.17
          --- argp-parse.c        18 Feb 2001 22:40:03 -0000      1.16
          +++ argp-parse.c        4 Feb 2003 19:52:30 -0000       1.17
          @ -1021,6 +1021,8 @
                    *arg_ebadkey = 1;
                    if (parser->first_nonopt != parser->last_nonopt)
                      {
          +             exchange(parser);
          +
                        /* Start processing the arguments we skipped previously. */
                        parser->state.next = parser->first_nonopt;
          
          

Go into argp/, run patch -p0, paste the patch, type <CTRL>-d.


Node: T2, Next: , Previous: T1, Up: Compiler Stages

T2, Building the Abstract Syntax Tree

This section was last updated for EPITA-2005 on 2003-02-25.

2005-T2 delivery is Friday, March 14th 2003 at noon.
At the end of this stage, the compiler can build abstract syntax trees of Tiger programs and pretty-print them. The parser is equipped with error recovery.


Node: T2 Goals, Next: , Up: T2

T2 Goals

Things to learn during this stage that you should remember:


Node: T2 Samples, Next: , Previous: T2 Goals, Up: T2

T2 Samples

Here are a few examples of expected features.

The parser builds abstract syntax trees that can be output by a pretty-printing module:

     /* define a recursive function */
     let
       /* calculate n! */
       function fact (n : int) : int =
         if  n = 0
           then 1
           else n * fact (n - 1)
     in
       fact (10)
     end
     File 11: simple-fact.tig
     
     $ tc -A simple-fact.tig
     /* == Abstract Syntax Tree. == */
     let
        function fact (n : int) : int =
           if (n = 0)
              then 1
              else (n * fact ((n - 1)))
     in
        fact (10)
     end
     Example 12: tc -A simple-fact.tig
     

The output from your pretty-printer must be valid Tiger code, and be equivalent to the input.

By valid, we mean that any Tiger compiler must be able to parse with success your output. Pay attention to the banners such as == Abstract...: you should use comments: /* == Abstract... */. Pay attention to special characters too.

     print ("\"\x45\x50ITA\n\"")
     File 13: string-escapes.tig
     
     $ tc -A string-escapes.tig
     /* == Abstract Syntax Tree. == */
     print ("\"EPITA\n\"")
     Example 14: tc -A string-escapes.tig
     

By equivalent, we mean that except for syntactic sugar, the output and the input are equal. Syntactic sugar refers to &, |, unary -, and in some cases if then:

     1 = 1 & 2 = 2
     File 15: 1s-and-2s.tig
     
     $ tc -A 1s-and-2s.tig
     /* == Abstract Syntax Tree. == */
     if (1 = 1)
        then (2 = 2)
        else 0
     Example 16: tc -A 1s-and-2s.tig
     
     $ tc -A 1s-and-2s.tig >output.tig
     Example 17: tc -A 1s-and-2s.tig >output.tig
     
     $ tc -A output.tig
     /* == Abstract Syntax Tree. == */
     if (1 = 1)
        then (2 = 2)
        else 0
     Example 18: tc -A output.tig
     

For loops must be properly displayed, i.e., although we use a ast::VarDec for the index of the loop, you must not display var:

     /* valid let and for */
     let
       var a := 0
     in
       for i := 0 to 100 do (a := a+1; ())
     end
     File 19: for-loop.tig
     
     $ tc -A for-loop.tig
     /* == Abstract Syntax Tree. == */
     let
        var a := 0
     in
        for i := 0 to 100 do
           (
              a := (a + 1);
              ()
           )
     end
     Example 20: tc -A for-loop.tig
     

Notice too that parentheses must not stack for free. In fact, you must even remove them.

     % cat parens.tig
     (((0)))
     % ./tc -A parens.tig
     /* == Abstract Syntax Tree. == */
     0
     

As a result, anything output by tc -A is equal to what tc -A | tc -A - displays!

Another part of T2 is the improvement of your parser: it must be robust to some forms of errors. Observe that on the following input:

     (
       1;
       (2, 3);
       (4, 5);
       6
     )
     File 21: multiple-parse-errors.tig
     

several parse errors are reported, not merely the first one:

     $ tc multiple-parse-errors.tig
     error-->multiple-parse-errors.tig:3.4: syntax error, unexpected ",", expecting ";"
     error-->multiple-parse-errors.tig:4.4: syntax error, unexpected ",", expecting ";"
     =>3
     Example 22: tc multiple-parse-errors.tig
     

Of course, the exit status still reveals the parse error. Be sure that your error recovery does not break the rest of the compiler...

     $ tc -A multiple-parse-errors.tig
     error-->multiple-parse-errors.tig:3.4: syntax error, unexpected ",", expecting ";"
     error-->multiple-parse-errors.tig:4.4: syntax error, unexpected ",", expecting ";"
     /* == Abstract Syntax Tree. == */
     (
        1;
        ();
        ();
        6
     )
     =>3
     Example 23: tc -A multiple-parse-errors.tig
     


Node: T2 Given Code, Next: , Previous: T2 Samples, Up: T2

T2 Given Code

Some code is provided: 2004-tc-2.0.tar.bz2. See src/misc, src/symbol, and src/ast.


Node: T2 Code to Write, Next: , Previous: T2 Given Code, Up: T2

T2 Code to Write

What is to be done:

src/symbol.*
Implement this class if you hadn't during T1. Understanding is required.
src/parse/scantiger.ll
The scanner must now use symbol::Symbol instead of std::string for identifiers. Of course, the parser must be adjusted too.

The scanner must be updated to keep track of locations of tokens in Tiger programs. To adjust your scanner, you are strongly encouraged to use YY_USER_ACTION, and also the yylex prologue:

          ...
          %%
          %{
            // Everything here is run each time yylex is invoked.
          %}
          "if"    return IF;
          ...
          %%
          ...
          

Have a look at the scanner and parser chapters of this draft.

src/parse/parsetiger.yy
Implement error recovery. There should be at least three uses of the token error. Read the Bison documentation about it.

The grammar must be changed to process declarations by chunks. In Tiger, the following program is invalid:

          let function foo () = ()
              function foo () = ()
              var      foo    := 0
          in
            ()
          end
          

while the following code is valid:

          let function foo () = ()
              var      foo    := 0
              function foo () = ()
          in
            ()
          end
          

this is because declarations are cut in "chunks" of declarations of the same kind. In the first example, there are two chunks: one chunk of two function declarations and one variable declaration. In the second example: there are three chunks: one chunk containing only one function declaration, one with one variable declaration, and the third one, with a single function declaration again.

The rule is "a chunk cannot define twice the same name".

In order to implement this easily, you must adjust your grammar so that declarations are parsed by chunks. Pay special attention to the implementation of ast::FunctionDecs, ast::VarDecs, and ast::TypeDecs (which are all implemented thanks to ast::AnyDecs): they are these chunks of declaration. Therefore, an ast::LetExp uses a list of chunks.

src/ast
The abstract syntax tree module must be completed. Basically, this means there should remain no FIXME: anywhere in the code we gave. Several files are missing (fieldvar.hh nilexp.hh intexp.hh stringexp.hh callexp.hh assignexp.hh whileexp.hh breakexp.hh arrayexp.hh). See src/ast/README for additional information on the missing classes.
src/ast/default-visitor.hh
The DefaultVisitor class must be completed, and must be able to walk whole abstract syntax trees. Do not forget that your DefaultVisitor must be a sound basis for your further work on the Tiger compiler.
src/ast/print-visitor.hh
The PrintVisitor class must be written entirely.


Node: T2 FAQ, Previous: T2 Code to Write, Up: T2

T2 FAQ

bison
Be sure to read its dedicated section: Flex & Bison.
escapes (escapes_get, etc.)
These are place holders for T3. You don't have to complete them. Actually, the easiest is probably just to comment these functions out. Or better yet: #if 0/#endif.
kinds (kind_get, etc.)
Some of the ast components have features such as _kind, kind_get and so forth. These are to be used only in T5, you don't have to complete them now.


Node: T3, Next: , Previous: T2, Up: Compiler Stages

T3, Computing the Escaping Variables

This section was updated for Tiger 2004. The project will be taken on Friday, March 15th, at noon.

At the end of this stage, the compiler must be able to compute and display the escaping variables. These features are triggered by the options --escapes-compute/-e and --escapes-display/-E.

Be sure to read the chapter "Escapes" in the lecture notes.


Node: T3 Goals, Next: , Up: T3

T3 Goals

Things to learn during this stage that you should remember:


Node: T3 Samples, Next: , Previous: T3 Goals, Up: T3

T3 Samples

This example demonstrates the computation and display of escaping variables/formals. Notice that by default, all variable must be considered as escaping, since it is safe to put a non escaping variable onto the stack, while the converse is unsafe.

     let
        var escaping := "I rule the world!\n"
        var not_escaping := "Peace on Earth for humans of good will.\n"
        function print_slogan (not_escaping: string) =
           (print (not_escaping); print (escaping))
     in
        print_slogan (not_escaping)
     end
     File 24: variable-escapes.tig
     
     $ tc -EeE variable-escapes.tig
     /* == Escapes. == */
     let
        var /* escaping */ escaping := "I rule the world!\n"
        var /* escaping */ not_escaping := "Peace on Earth for humans of good will.\n"
        function print_slogan (/* escaping */ not_escaping : string) =
           (
              print (not_escaping);
              print (escaping)
           )
     in
        print_slogan (not_escaping)
     end
     /* == Escapes. == */
     let
        var /* escaping */ escaping := "I rule the world!\n"
        var not_escaping := "Peace on Earth for humans of good will.\n"
        function print_slogan (not_escaping : string) =
           (
              print (not_escaping);
              print (escaping)
           )
     in
        print_slogan (not_escaping)
     end
     Example 25: tc -EeE variable-escapes.tig
     

You are strongly encouraged to run your compiler on merge.tig and to study its output. There is a number of silly mistakes that people usually do on T3: they are all easy to defeat when you do have a reasonable test suite, and once you understood that torturing your project is a good thing to do.


Node: T3 Code To Write, Next: , Previous: T3 Samples, Up: T3

T3 Code To Write

ast::PrintVisitor
Be sure to display the /* escaping */ flag where needed, and only where needed. If you don't pay attention, you might display meaningless flags due to implementation details.
escapes::EscapesVisitor
Write the class escapes::EscapesVisitor in src/escapes/escapes-visitor.hh.

You are suggested to implement three additional classes:

Definition
An abstract class which is used to instantiate the template class symbol::Table into Table <Definition>.

escape_set (void) virtual void
Sets the escape to true.

int _depth Variable
Depth at which this object has been created.

depth_get () const int
Returns the depth associated to this Definition object.

VariableDefinition
Inherits from Definition. It has one additional attribute, a VarDec &. The method escape_set is implemented, and when invoked, set the escapes flags of the corresponding VarDec.
FormalDefinition
Inherits from Definition. To be designed by yourself. Do not forget that the ast class used to register formals is used elsewhere, and it would be a pity that your implementation makes no difference... Be sure to write a test that verifies that your implementation is not abused. I have one such test...

Equip ast
All the sites where variables and formals (i.e., the arguments of the functions being defined, not being used) are introduced must be equipped with the escape_get and escape_set methods. Most probably the code was already given, and is using const_casts; try to use mutable instead.

Modify the code so that each definition of an escaping variable/formal is preceded by the comment /* escaping */ if the flag display_escapes_p is true. See the item "Driver" for an example.


Node: T3 FAQ, Previous: T3 Code To Write, Up: T3

T3 FAQ

Dwarf errors
It appears that at EPITA, the linker is unable to read the output of G++ 3.2 when given -ggdb. So don't pass it.


Node: T4, Next: , Previous: T3, Up: Compiler Stages

T4, Type Checking

This section was last updated for EPITA-2005 on 2003-04-08.

2005-T4 delivery is Friday, April 25th 2003 at noon.
At the end of this stage, the compiler type checks Tiger programs. Clear error messages are required.


Node: T4 Goals, Next: , Up: T4

T4 Goals

Things to learn during this stage that you should remember:


Node: T4 Samples, Next: , Previous: T4 Goals, Up: T4

T4 Samples

Type checking is optional, invoked by --types-check or -T:

     1 + "2"
     File 26: int-plus-string.tig
     
     $ tc int-plus-string.tig
     Example 27: tc int-plus-string.tig
     
     $ tc int-plus-string.tig --types-check
     error-->int-plus-string.tig:1.0-6: type mismatch
     error-->  right operand type: string
     error-->  expected type: int
     =>4
     Example 28: tc int-plus-string.tig --types-check
     

When there are several type errors, it is admitted that some remain hidden by others.

     unknown_function (unknown_variable)
     File 29: unknowns.tig
     
     $ tc unknowns.tig --types-check
     error-->unknowns.tig:1.0-34: unknown function: unknown_function
     =>4
     Example 30: tc unknowns.tig --types-check
     

Be sure to check the type of all the constructs.

     if 1 then 2
     File 31: bad-if.tig
     
     $ tc bad-if.tig --types-check
     error-->bad-if.tig:1.0-10: type mismatch
     error-->  then clause type: int
     error-->  else clause type: void
     =>4
     Example 32: tc bad-if.tig --types-check
     

Be aware that type and function declarations are recursive by chunks. For instance:

     let type one = { hd : int, tail : two }
         type two = { hd : int, tail : one }
         function one (hd : int, tail : two) : one
            = one { hd = hd, tail = tail }
         function two (hd : int, tail : one) : two
            = two { hd = hd, tail = tail }
         var one := one (11, two (22, nil))
     in
       print_int (one.tail.hd); print ("\n")
     end
     File 33: mutuals.tig
     
     $ tc mutuals.tig --types-check
     Example 34: tc mutuals.tig --types-check
     

In case you are interested, the result is:

     $ tc -H mutuals.tig >mutuals.hir
     Example 35: tc -H mutuals.tig >mutuals.hir
     
     $ havm mutuals.hir
     22
     Example 36: havm mutuals.hir
     


Node: T4 Given Code, Next: , Previous: T4 Samples, Up: T4

T4 Given Code

Some code is provided: 2005-tc-4.3.tar.bz2. The transition from the previous versions can be done thanks to the following diffs: 2005-tc-2.1-4.0.diff, 2005-tc-4.0-4.1.diff, 2005-tc-4.1-4.2.diff, 2005-tc-4.2-4.3.diff. See src/misc.


Node: T4 Code to Write, Next: , Previous: T4 Given Code, Up: T4

T4 Code to Write

What is to be done.

symbol::Table< class Entry_T >
Write the class template symbol::Table in src/symbol/table.hh which is a table of symbols dedicated to storing some data which type is Entry_T *. In short, it maps a symbol::Symbol to an Entry_T * (that should ring a bell...). You are encouraged to implement something simple, based on stacks (see std::stack or std::list) and maps (see std::map).

symbol::Table is a class template as it is used by virtually all the AST visitors (e.g., escapes::EscapesVisitor, type::TypeVisitor, translate::TranslateVisitor etc.)

symbol::Table must provide this interface:

scope_begin () void
Open a new scope.

scope_end () void
Close the last scope, forgetting everything since the latest scope_begin ().

put (Symbol key, Entry_T & value) void
Associate value to key in the current scope.

get (Symbol key) const Entry_T *
If key was associated to some Entry_T in the open scopes, return the most recent insertion. Otherwise return the empty pointer.

print (std::ostream & ostr) const void
Send the content of this table on ostr in a readable manner, the top of the stack being displayed last.

src/type/types.hh
The Singletons type::String, type::Int, and type::Void are to be implemented. Using templates would be particularly appreciated to factor the code between the four singleton classes.

type::Named is almost entirely given.

type::Array is even simpler than the four Singletons.

type::Record is somewhat incomplete.

Pay extra attention to the implementation of type::operator== (const Type& a, const Type& b), type::Type::assignable_to and type::Type::comparable_to.

src/type/type-entry.hh
This file is really a empty nutshell, so we give it complete so that you concentrate on things that matter. Nonetheless you will be asked questions on this file, so study it.
src/type/type-env.hh
The constructor of type::TypeEnv must be completed: it must fill the environment with the definition of builtin types and functions. See the Tiger Reference Manual.

The handling of types is left as an example, you still have to implement the variables and functions support.

type::TypeVisitor
Of course this is the most tricky part. I hope there are enough comments in there so that you understand what is to be done. Please, post your questions and help me improve it.
Tasks
Each module Foo exports its tasks via the file foo/foo-tasks.hh. You must clean up your code to use the latest sources for Tasks, and make sure that your configure.ac no longer includes foo/libfoo.hh in src/modules.hh.


Node: T4 Options, Next: , Previous: T4 Code to Write, Up: T4

T4 Options

These are features that you might want to implement in addition to the core features.

type::Error
One problem is that type error recovery can generate false errors. For instance our compiler usually considers that the type for incorrect constructs is Int, which can create cascades of errors:
          "666" = if 000 then 333 else "666"
          File 37: is_devil.tig
          
          $ tc is_devil.tig --types-check
          error-->is_devil.tig:1.8-33: type mismatch
          error-->  then clause type: int
          error-->  else clause type: string
          error-->is_devil.tig:1.0-33: type mismatch
          error-->  left operand type: string
          error-->  right operand type: int
          =>4
          Example 38: tc is_devil.tig --types-check
          

One means to avoid this issue consists in introducing a new type, type::Error, that the type checker would never complain about. This can be a nice complement to ast::Error.


Node: T4 FAQ, Previous: T4 Options, Up: T4

T4 FAQ

Stupid Types
One can legitimately wonder whether the following program is correct:
          let type weirdo = array of weirdo
          in
            print ("I'm a creep.\n")
          end
          

the answer is "yes", as nothing prevents this in the Tiger specifications. Note that this type is not usable though.

kinds (kind_get, etc.)
Some of the ast components have features such as _kind, kind_get and so forth. These are to be used only in T5, you don't have to complete them now.
The TypeVisitor is not a ConstVisitor
One of the tasks of the type checking is to pass additional information to the translation. For instance, since < is overloaded (for integers and strings), the translation needs to know the types of the arguments. In a traditional compiler, type checking and translation would be performed simultaneously, but our Tiger Compiler, in order to simplify its architecture, has two different passes for each. Hence, the TypeVisitor will have to leave notes on the AST for the TranslateVisitor, therefore it cannot be a const visitor once T5 implemented. It can perfectly be const during T4.


Node: T5, Next: , Previous: T4, Up: Compiler Stages

T5, Translating to the High Level Intermediate Representation

This section was last updated for EPITA-2005 on 2003-06-10.

2005-T56 delivery is Friday, June 20th, at noon.
At the end of this stage the compiler translates the AST into the high level intermediate representation, HIR for short. And, of course, all the errors of previous stages have been fixed.


Node: T5 Goals, Next: , Up: T5

T5 Goals

Things to learn during this stage that you should remember:


Node: T5 Samples, Next: , Previous: T5 Goals, Up: T5

T5 Samples

T5 can be started (and should be started if you don't want to finish it in a hurry) by first making sure your compiler can handle code that uses no variables. Then, you can complete your compiler to support more and more Tiger features.


Node: T5 Primitive Samples, Next: , Up: T5 Samples

T5 Primitive Samples

This example is probably the simplest Tiger program.

     0
     File 39: 0.tig
     
     $ tc --hir-display 0.tig
     /* == High Level Intermediate representation. == */
     # Routine: Main
     label l`Main'
     # Prologue
     # Body
     sxp
         const 0
     # Epilogue
     label end
     Example 40: tc --hir-display 0.tig
     

You should then probably try to make more difficult programs with literals only. Arithmetics is one of the easiest tasks.

     1 + 2 * 3
     File 41: arith.tig
     
     $ tc -H arith.tig
     /* == High Level Intermediate representation. == */
     # Routine: Main
     label l`Main'
     # Prologue
     # Body
     sxp
         binop (+)
             const 1
             binop (*)
                 const 2
                 const 3
     # Epilogue
     label end
     Example 42: tc -H arith.tig
     

You should use havm to exercise your output.

     $ tc -H arith.tig >arith.hir
     Example 43: tc -H arith.tig >arith.hir
     
     $ havm arith.hir
     Example 44: havm arith.hir
     

Unfortunately, without actually printing something, you won't see the final result, which means you need to implement calls. Fortunately, you can ask havm for a verbose execution:

     $ havm --trace arith.hir
     error-->plaining
     error-->unparsing
     error-->checking
     error-->checkingLow
     error-->evaling
     error--> call ( name Main ) []
     error-->8.8-8.15:  const 1
     error-->10.12-10.19:  const 2
     error-->11.12-11.19:  const 3
     error-->9.8-11.19:  binop (*) 2 3
     error-->7.4-11.19:  binop (+) 1 6
     error-->6.0-11.19:  sxp 7
     error--> end call ( name Main ) [] = 0
     Example 45: havm --trace arith.hir
     

If you look carefully, you will find an sxp 7 in there...

Then you are encouraged to implement control structures.

     if 101 then 102 else 103
     File 46: if-101.tig
     
     $ tc -H if-101.tig
     /* == High Level Intermediate representation. == */
     # Routine: Main
     label l`Main'
     # Prologue
     # Body
     seq
         cjump ne
             const 101
             const 0
             name l0
             name l1
         label l0
         sxp
             const 102
         jump
             name l2
         label l1
         sxp
             const 103
         label l2
     seq end
     # Epilogue
     label end
     Example 47: tc -H if-101.tig
     

And even more difficult control structure uses:

     while 101
       do (if 102 then break)
     File 48: while-101.tig
     
     $ tc -H while-101.tig
     /* == High Level Intermediate representation. == */
     # Routine: Main
     label l`Main'
     # Prologue
     # Body
     seq
         label l1
         cjump ne
             const 101
             const 0
             name l2
             name l0
         label l2
         seq
             cjump ne
                 const 102
                 const 0
                 name l3
                 name l4
             label l3
             jump
                 name l0
             jump
                 name l5
             label l4
             sxp
                 const 0
             label l5
         seq end
         jump
             name l1
         label l0
     seq end
     # Epilogue
     label end
     Example 49: tc -H while-101.tig
     


Node: T5 Optimizing Cascading If, Next: , Previous: T5 Primitive Samples, Up: T5 Samples

T5 Optimizing Cascading If

Our compiler optimizes the number of jumps needed to compute nested if, using translate::Ix where a plain use of translate::Cx, Nx, and Ex is possible, but less efficient.

Consider the following sample:

     if 11 | 22 then print ("OK\n")
     File 50: boolean.tig
     

a naive implementation will probably produce too many successive cjump instructions:

     $ tc --hir-naive -H boolean.tig
     /* == High Level Intermediate representation. == */
     
     label l3
             "OK\n"
     # Routine: Main
     label l`Main'
     # Prologue
     # Body
     seq
         cjump ne
             eseq
             seq
                 cjump ne
                     const 11
                     const 0
                     name l0
                     name l1
                 label l0
                 move
                     temp t0
                     const 1
                 jump
                     name l2
                 label l1
                 move
                     temp t0
                     const 22
                 jump
                     name l2
                 label l2
             seq end
                 temp t0
             const 0
             name l4
             name l5
         label l4
         sxp
             call
                 name l`print'
                 name l3
             call end
         jump
             name l6
         label l5
         sxp
             const 0
         jump
             name l6
         label l6
     seq end
     # Epilogue
     label end
     Example 51: tc --hir-naive -H boolean.tig
     
     $ tc --hir-naive -H boolean.tig >boolean-1.hir
     Example 52: tc --hir-naive -H boolean.tig >boolean-1.hir
     
     $ havm --profile boolean-1.hir
     error-->/* Profiling.  */
     error-->fetches from temporary : 1
     error-->fetches from memory    : 0
     error-->binary operations      : 0
     error-->function calls         : 1
     error-->stores to temporary    : 1
     error-->stores to memory       : 0
     error-->jumps                  : 2
     error-->conditional jumps      : 2
     error-->/* Execution time.  */
     error-->number of cycles : 16
     OK
     Example 53: havm --profile boolean-1.hir
     

If you carefully analyze the cause of this pessimization, it is related to the computation of an intermediary expression (the value of 11 | 22) which is later decoded as a condition. A proper implementation will produce:

     $ tc -H boolean.tig
     /* == High Level Intermediate representation. == */
     
     label l0
             "OK\n"
     # Routine: Main
     label l`Main'
     # Prologue
     # Body
     seq
         seq
             cjump ne
                 const 11
                 const 0
                 name l4
                 name l5
             label l4
             cjump ne
                 const 1
                 const 0
                 name l1
                 name l2
             label l5
             cjump ne
                 const 22
                 const 0
                 name l1
                 name l2
         seq end
         label l1
         sxp
             call
                 name l`print'
                 name l0
             call end
         jump
             name l3
         label l2
         sxp
             const 0
         label l3
     seq end
     # Epilogue
     label end
     Example 54: tc -H boolean.tig
     
     $ tc -H boolean.tig >boolean-2.hir
     Example 55: tc -H boolean.tig >boolean-2.hir
     
     $ havm --profile boolean-2.hir
     error-->/* Profiling.  */
     error-->fetches from temporary : 0
     error-->fetches from memory    : 0
     error-->binary operations      : 0
     error-->function calls         : 1
     error-->stores to temporary    : 0
     error-->stores to memory       : 0
     error-->jumps                  : 1
     error-->conditional jumps      : 2
     error-->/* Execution time.  */
     error-->number of cycles : 13
     OK
     Example 56: havm --profile boolean-2.hir
     


Node: T5 Builtin Calls Samples, Next: , Previous: T5 Optimizing Cascading If, Up: T5 Samples

T5 Builtin Calls Samples

But the game becomes more interesting when you implement function calls (which is easier than compiling functions). print_int is probably the first builtin to implement:

     (print_int (101); print ("\n"))
     File 57: print-101.tig
     
     $ tc -H print-101.tig >print-101.hir
     Example 58: tc -H print-101.tig >print-101.hir
     
     $ havm print-101.hir
     101
     Example 59: havm print-101.hir
     

Complex values, arrays and records, also need calls to the runtime system:

     let type list = { h: int, t: list }
         var list := list { h = 1, t = list { h = 2, t = nil } }
     in
       print_int (list.t.h); print ("\n")
     end
     File 60: print-list.tig
     
     $ tc -H print-list.tig
     /* == High Level Intermediate representation. == */
     
     label l0
             "\n"
     # Routine: Main
     label l`Main'
     # Prologue
     move
         temp t2
         temp fp
     move
         temp fp
         temp sp
     move
         temp sp
         binop (-)
             temp sp
             const 4
     # Body
     seq
         move
             mem
                 temp $fp
             eseq
             seq
                 move
                     temp t1
                     call
                         name l`malloc'
                         const 8
                     call end
                 move
                     mem
                         binop (+)
                             temp t1
                             const 0
                     const 1
                 move
                     mem
                         binop (+)
                             temp t1
                             const 4
                     eseq
                     seq
                         move
                             temp t0
                             call
                                 name l`malloc'
                                 const 8
                             call end
                         move
                             mem
                                 binop (+)
                                     temp t0
                                     const 0
                             const 2
                         move
                             mem
                                 binop (+)
                                     temp t0
                                     const 4
                             const 0
                     seq end
                         temp t0
             seq end
                 temp t1
         seq
             sxp
                 call
                     name l`print_int'
                     mem
                         binop (+)
                             mem
                                 binop (+)
                                     mem
                                         temp $fp
                                     const 4
                             const 0
                 call end
             sxp
                 call
                     name l`print'
                     name l0
                 call end
         seq end
     seq end
     # Epilogue
     move
         temp sp
         temp fp
     move
         temp fp
         temp t2
     label end
     Example 61: tc -H print-list.tig
     
     $ tc -H print-list.tig >print-list.hir
     Example 62: tc -H print-list.tig >print-list.hir
     
     $ havm print-list.hir
     2
     Example 63: havm print-list.hir
     


Node: T5 Samples with Variables, Previous: T5 Builtin Calls Samples, Up: T5 Samples

T5 Samples with Variables

Here is an example which demonstrates the usefulness of information about escapes: when escaping variables are not computed, they are all stored on the stack:

     let var a := 1
         var b := 2
         var c := 3
     in
       a := 2;
       c := a + b + c;
       print_int (c);
       print ("\n")
     end
     File 64: vars.tig
     
     $ tc -H vars.tig
     /* == High Level Intermediate representation. == */
     
     label l0
             "\n"
     # Routine: Main
     label l`Main'
     # Prologue
     move
         temp t0
         temp fp
     move
         temp fp
         temp sp
     move
         temp sp
         binop (-)
             temp sp
             const 12
     # Body
     seq
         move
             mem
                 temp $fp
             const 1
         move
             mem
                 binop (+)
                     temp $fp
                     const -4
             const 2
         move
             mem
                 binop (+)
                     temp $fp
                     const -8
             const 3
         seq
             move
                 mem
                     temp $fp
                 const 2
             move
                 mem
                     binop (+)
                         temp $fp
                         const -8
                 binop (+)
                     binop (+)
                         mem
                             temp $fp
                         mem
                             binop (+)
                                 temp $fp
                                 const -4
                     mem
                         binop (+)
                             temp $fp
                             const -8
             sxp
                 call
                     name l`print_int'
                     mem
                         binop (+)
                             temp $fp
                             const -8
                 call end
             sxp
                 call
                     name l`print'
                     name l0
                 call end
         seq end
     seq end
     # Epilogue
     move
         temp sp
         temp fp
     move
         temp fp
         temp t0
     label end
     Example 65: tc -H vars.tig
     

But once escaping variable computation implemented, we know none escape in this example, hence they can be stored in temporaries:

     $ tc -eH vars.tig
     /* == High Level Intermediate representation. == */
     
     label l0
             "\n"
     # Routine: Main
     label l`Main'
     # Prologue
     # Body
     seq
         move
             temp t0
             const 1
         move
             temp t1
             const 2
         move
             temp t2
             const 3
         seq
             move
                 temp t0
                 const 2
             move
                 temp t2
                 binop (+)
                     binop (+)
                         temp t0
                         temp t1
                     temp t2
             sxp
                 call
                     name l`print_int'
                     temp t2
                 call end
             sxp
                 call
                     name l`print'
                     name l0
                 call end
         seq end
     seq end
     # Epilogue
     label end
     Example 66: tc -eH vars.tig
     
     $ tc -eH vars.tig >vars.hir
     Example 67: tc -eH vars.tig >vars.hir
     
     $ havm vars.hir
     7
     Example 68: havm vars.hir
     

Then, you should implement the declaration of functions:

     let function fact (i: int) : int =
         if i = 0 then 1
                  else i * fact (i - 1)
     in
       print_int (fact (15));
       print ("\n")
     end
     File 69: fact15.tig
     
     $ tc -H fact15.tig
     /* == High Level Intermediate representation. == */
     # Routine: fact
     label l0
     # Prologue
     move
         temp t1
         temp fp
     move
         temp fp
         temp sp
     move
         temp sp
         binop (-)
             temp sp
             const 8
     move
         mem
             temp $fp
         temp i0
     move
         mem
             binop (+)
                 temp $fp
                 const -4
         temp i1
     # Body
     move
         temp $v0
         eseq
         seq
             cjump eq
                 mem
                     binop (+)
                         temp $fp
                         const -4
                 const 0
                 name l1
                 name l2
             label l1
             move
                 temp t0
                 const 1
             jump
                 name l3
             label l2
             move
                 temp t0
                 binop (*)
                     mem
                         binop (+)
                             temp $fp
                             const -4
                     call
                         name l0
                         mem
                             temp $fp
                         binop (-)
                             mem
                                 binop (+)
                                     temp $fp
                                     const -4
                             const 1
                     call end
             label l3
         seq end
             temp t0
     # Epilogue
     move
         temp sp
         temp fp
     move
         temp fp
         temp t1
     label end
     
     
     label l4
             "\n"
     # Routine: Main
     label l`Main'
     # Prologue
     # Body
     seq
         sxp
             call
                 name l`print_int'
                 call
                     name l0
                     temp $fp
                     const 15
                 call end
             call end
         sxp
             call
                 name l`print'
                 name l4
             call end
     seq end
     # Epilogue
     label end
     Example 70: tc -H fact15.tig
     
     $ tc -H fact15.tig >fact15.hir
     Example 71: tc -H fact15.tig >fact15.hir
     
     $ havm fact15.hir
     2004310016
     Example 72: havm fact15.hir
     

And finally, you should support escaping variables (see example 24):

     $ tc -eH variable-escapes.tig
     /* == High Level Intermediate representation. == */
     
     label l0
             "I rule the world!\n"
     
     label l1
             "Peace on Earth for humans of good will.\n"
     # Routine: print_slogan
     label l2
     # Prologue
     move
         temp t2
         temp fp
     move
         temp fp
         temp sp
     move
         temp sp
         binop (-)
             temp sp
             const 4
     move
         mem
             temp $fp
         temp i0
     move
         temp t1
         temp i1
     # Body
     seq
         sxp
             call
                 name l`print'
                 temp t1
             call end
         sxp
             call
                 name l`print'
                 mem
                     mem
                         temp $fp
             call end
     seq end
     # Epilogue
     move
         temp sp
         temp fp
     move
         temp fp
         temp t2
     label end
     
     # Routine: Main
     label l`Main'
     # Prologue
     move
         temp t3
         temp fp
     move
         temp fp
         temp sp
     move
         temp sp
         binop (-)
             temp sp
             const 4
     # Body
     seq
         move
             mem
                 temp $fp
             name l0
         move
             temp t0
             name l1
         sxp
             call
                 name l2
                 temp $fp
                 temp t0
             call end
     seq end
     # Epilogue
     move
         temp sp
         temp fp
     move
         temp fp
         temp t3
     label end
     Example 73: tc -eH variable-escapes.tig
     


Node: T5 Given Code, Next: , Previous: T5 Samples, Up: T5

T5 Given Code

Some code is provided, see T6 Given Code. See src/temp, src/tree, src/frame, src/translate.


Node: T5 Code to Write, Next: , Previous: T5 Given Code, Up: T5

T5 Code to Write

You are encouraged to try first very simple examples: nil, 1 + 2, "foo" < "bar" etc. Then consider supporting variables, and finally handle the case of the functions.

Driver
The driver must performs the translation when given --hir-compute, but displays the result iff the option -H was given. Obviously, an input that has not been type-checked cannot be translated, so --hir-compute implies --types-check.
TypeVisitor
The TranslateVisitor often needs additional type information to proceed, especially expression versus instruction. Hence, you'll have to update the TypeVisitor to leave notes on the AST using kind_set and so forth.
src/translate/fragment.hh
There remains to implement translate::ProcFrag::print which outputs the routine themselves plus the glue code (allocating the frame etc.).
src/translate/level-env.hh
Code is missing. In particular, bear in mind that the initial environment is not empty...
src/translate/translation.hh
There are many holes to fill.
src/translate/translate-visitor.hh
There are holes to fill.


Node: T5 Options, Previous: T5 Code to Write, Up: T5

T5 Options

This section documents possible extensions you could implement in T5.


Node: T5 Bounds Checking, Next: , Up: T5 Options

T5 Bounds Checking

Implementing bounds checking is quite simple: it consists in having the program die when the program accesses an invalid subscript in an array. For instance, the following code is "succeeds" with a non-bounds-checking compiler.

     let type int_array = array of int
         var  size  := 2
         var  arr1  := int_array [size] of 0
         var  arr2  := int_array [size] of 0
         var  two   := 2
         var  m_one := -1
     in
       arr1[two]   := 3;
       arr2[m_one] := -1;
     
       print_int (arr1[1]);
       print ("\n");
       print_int (arr2[0]);
       print ("\n")
     end
     File 74: bounds-violation.tig
     
     $ tc -H bounds-violation.tig >bounds-violation.hir
     Example 75: tc -H bounds-violation.tig >bounds-violation.hir
     
     $ havm bounds-violation.hir
     -1
     3
     Example 76: havm bounds-violation.hir
     

When run with --bound-checking2, your compiler produces code that diagnoses such cases, and exits with status 120. Something like:

     error-->bounds-violation.tig:8.2-17: index out of arr1 bounds (0 .. 1): 2
     =>120
     


Node: T5 Optimizing Static Links, Previous: T5 Bounds Checking, Up: T5 Options

T5 Optimizing Static Links

Warning: this optimization is difficult to do it perfectly, and therefore, expect a big bonus.

In a first and conservative extension, the compiler considers that all the functions (but the builtins!) need a static link. This is correct, but inefficient: for instance, the traditional fact function will spend almost as much time handling the static link, than its real argument.

Some functions need a static link, but don't need to save it on the stack. For instance, in the following:

     let var foo := 1
         function foo () : int = foo
     in
       foo ()
     end
     

the function foo does need a static link to access the variable foo, but does not need to store its static link on the stack.

It is suggested to address these problems in the following order:

  1. Implement the detection of functions that do not need a static link (see exercise 6.5 in "Modern Implementation of Compilers"), but still consider any static link escapes.
  2. Adjust the output of --escapes-display to display /* escaping sl */ before the first formal argument of the functions (declarations) that need the static link:
              $ tc -E fact.tig
              /* == Escapes. == */
              let
                 function fact (/* escaping sl *//* escaping */ n : int) : int =
                    if  (n = 0)
                       then 1
                       else  (n * fact ( (n - 1)))
              in
                 fact (10)
              end
              
              $ tc -eE fact.tig
              /* == Escapes. == */
              let
                 function fact (n : int) : int =
                    if  (n = 0)
                       then 1
                       else  (n * fact ( (n - 1)))
              in
                 fact (10)
              end
              
  3. Adjust your call and progFrag prologues.
  4. Improve your computation so that non escaping static links are detected:
              $ tc -eE escaping-sl.tig
              /* == Escapes. == */
              let
                 var      toto := 1
                 function outer (/* escaping sl */) : int =
                   let function inner (/* sl */) : int = toto
                   in inner () end
              in
                outer ()
              end
              

    Watch out, it is not trivial to find the minimum. What do you think about the static link of the function sister below?

              let
                 var      toto := 1
                 function outer () : int =
                   let function inner () : int = toto
                   in inner () end
                 function sister () : int = outer ()
              in
                sister ()
              end
              


Node: T6, Next: , Previous: T5, Up: Compiler Stages

T6, Translating to the Low Level Intermediate Representation

This section was last updated for EPITA-2005 on 2003-05-15.

2005-T56 delivery is Friday, June 20th, at noon.
There will be no additional code: there is no "holes" to fill, you have to write the whole thing. Consequently, you may start T6 as soon as you want.

At the end of this stage, the compiler produces low level intermediate representation: LIR. LIR is a subset of the HIR: some patterns are forbidden. This is why it is also named canonicalization.


Node: T6 Goals, Next: , Up: T6

T6 Goals

Things to learn during this stage that you should remember:


Node: T6 Samples, Next: , Previous: T6 Goals, Up: T6

T6 Samples

There are several stages in T6.


Node: T6 Canonicalization Samples, Next: , Up: T6 Samples

T6 Canonicalization Samples

The first task in T6 is getting rid of all the eseq. To do this, you have to move the statement part of an eseq at the end of the current sequence point, and keeping the expression part in place.

Compare for instance the HIR to the LIR in the following case:

     let function print_ints (a: int, b: int) =
         (print_int (a); print (", "); print_int (b); print ("\n"))
         var a := 0
     in
       print_ints (1,  (a := a + 1; a))
     end
     File 77: preincr-1.tig
     

One possible HIR translation is:

     $ tc -eH preincr-1.tig
     /* == High Level Intermediate representation. == */
     
     label l1
             ", "
     
     label l2
             "\n"
     # Routine: print_ints
     label l0
     # Prologue
     move
         temp t2
         temp fp
     move
         temp fp
         temp sp
     move
         temp sp
         binop (-)
             temp sp
             const 4
     move
         mem
             temp $fp
         temp i0
     move
         temp t0
         temp i1
     move
         temp t1
         temp i2
     # Body
     seq
         sxp
             call
                 name l`print_int'
                 temp t0
             call end
         sxp
             call
                 name l`print'
                 name l1
             call end
         sxp
             call
                 name l`print_int'
                 temp t1
             call end
         sxp
             call
                 name l`print'
                 name l2
             call end
     seq end
     # Epilogue
     move
         temp sp
         temp fp
     move
         temp fp
         temp t2
     label end
     
     # Routine: Main
     label l`Main'
     # Prologue
     # Body
     seq
         move
             temp t3
             const 0
         sxp
             call
                 name l0
                 temp $fp
                 const 1
                 eseq
                     move
                         temp t3
                         binop (+)
                             temp t3
                             const 1
                     temp t3
             call end
     seq end
     # Epilogue
     label end
     Example 78: tc -eH preincr-1.tig
     

A possible canonicalization is then:

     $ tc -eL preincr-1.tig
     /* == Low Level Intermediate representation. == */
     
     label l1
             ", "
     
     label l2
             "\n"
     # Routine: print_ints
     label l0
     # Prologue
     move
         temp t2
         temp fp
     move
         temp fp
         temp sp
     move
         temp sp
         binop (-)
             temp sp
             const 4
     move
         mem
             temp $fp
         temp i0
     move
         temp t0
         temp i1
     move
         temp t1
         temp i2
     # Body
     seq
         label l3
         sxp
             call
                 name l`print_int'
                 temp t0
             call end
         sxp
             call
                 name l`print'
                 name l1
             call end
         sxp
             call
                 name l`print_int'
                 temp t1
             call end
         sxp
             call
                 name l`print'
                 name l2
             call end
         label l4
     seq end
     # Epilogue
     move
         temp sp
         temp fp
     move
         temp fp
         temp t2
     label end
     
     # Routine: Main
     label l`Main'
     # Prologue
     # Body
     seq
         label l5
         move
             temp t3
             const 0
         move
             temp t5
             temp $fp
         move
             temp t3
             binop (+)
                 temp t3
                 const 1
         sxp
             call
                 name l0
                 temp t5
                 const 1
                 temp t3
             call end
         label l6
     seq end
     # Epilogue
     label end
     Example 79: tc -eL preincr-1.tig
     

But please note the example above is simple because 1 commutes with (a := a + 1; a): the order does not matter. But if you change the 1 into a, then you cannot exchange a and (a := a + 1; a), so the translation is different. Compare the previous LIR with the following, and pay attention to

     let function print_ints (a: int, b: int) =
         (print_int (a); print (", "); print_int (b); print ("\n"))
         var a := 0
     in
       print_ints (a,  (a := a + 1; a))
     end
     File 80: preincr-2.tig
     
     $ tc -eL preincr-2.tig
     /* == Low Level Intermediate representation. == */
     
     label l1
             ", "
     
     label l2
             "\n"
     # Routine: print_ints
     label l0
     # Prologue
     move
         temp t2
         temp fp
     move
         temp fp
         temp sp
     move
         temp sp
         binop (-)
             temp sp
             const 4
     move
         mem
             temp $fp
         temp i0
     move
         temp t0
         temp i1
     move
         temp t1
         temp i2
     # Body
     seq
         label l3
         sxp
             call
                 name l`print_int'
                 temp t0
             call end
         sxp
             call
                 name l`print'
                 name l1
             call end
         sxp
             call
                 name l`print_int'
                 temp t1
             call end
         sxp
             call
                 name l`print'
                 name l2
             call end
         label l4
     seq end
     # Epilogue
     move
         temp sp
         temp fp
     move
         temp fp
         temp t2
     label end
     
     # Routine: Main
     label l`Main'
     # Prologue
     # Body
     seq
         label l5
         move
             temp t3
             const 0
         move
             temp t5
             temp $fp
         move
             temp t6
             temp t3
         move
             temp t3
             binop (+)
                 temp t3
                 const 1
         sxp
             call
                 name l0
                 temp t5
                 temp t6
                 temp t3
             call end
         label l6
     seq end
     # Epilogue
     label end
     Example 81: tc -eL preincr-2.tig
     

As you can see, the output is the same for the HIR and the LIR:

     $ tc -eH preincr-2.tig >preincr-2.hir
     Example 82: tc -eH preincr-2.tig >preincr-2.hir
     
     $ havm preincr-2.hir
     0, 1
     Example 83: havm preincr-2.hir
     
     $ tc -eL preincr-2.tig >preincr-2.lir
     Example 84: tc -eL preincr-2.tig >preincr-2.lir
     
     $ havm preincr-2.lir
     0, 1
     Example 85: havm preincr-2.lir
     

Be very careful when dealing with mem. For instance, rewriting something like:

     call (foo, eseq (move (temp t, const 51), temp t))
     

into

     move temp t1, temp t
     move temp t, const 51
     call (foo, temp t)
     

is dead wrong: temp t is a subexpression: it is being defined here. You should produce:

     move temp t, const 51
     call (foo, temp t)
     

Another danger is the handling of move (mem, ). For instance:

     move (mem foo, x)
     

must be rewritten into:

     move (temp t, foo)
     move (mem (temp t), x)
     

not as:

     move (temp t, mem (foo))
     move (temp t, x)
     

In other words, the first subexpression of move (mem (foo), ) is foo, not mem (foo). The following example is a good crash test against this problem:

     let type int_array = array of int
         var tab := int_array [2] of 51
     in
       tab[0] := 100;
       tab[1] := 200;
       print_int (tab[0]); print ("\n");
       print_int (tab[1]); print ("\n")
     end
     File 86: move-mem.tig
     
     $ tc -eL move-mem.tig >move-mem.lir
     Example 87: tc -eL move-mem.tig >move-mem.lir
     
     $ havm move-mem.lir
     100
     200
     Example 88: havm move-mem.lir
     

You also ought to get rid of nested calls:

     print (chr (ord ("\n")))
     File 89: nested-calls.tig
     
     $ tc -L nested-calls.tig
     /* == Low Level Intermediate representation. == */
     
     label l0
             "\n"
     # Routine: Main
     label l`Main'
     # Prologue
     # Body
     seq
         label l1
         move
             temp t1
             call
                 name l`ord'
                 name l0
             call end
         move
             temp t2
             call
                 name l`chr'
                 temp t1
             call end
         sxp
             call
                 name l`print'
                 temp t2
             call end
         label l2
     seq end
     # Epilogue
     label end
     Example 90: tc -L nested-calls.tig
     

In fact there are only two valid call forms: sxp (call (...)), and move (temp (...), call (...)).

Note that, contrary to C, the HIR and LIR always denote the same value. For instance the following Tiger code:

     let
       var a := 1
       function a (t: int) : int =
          (a := a + 1;
           print_int (t); print (" -> "); print_int (a); print ("\n");
           a)
       var b := a (1) + a (2) * a (3)
     in
       print_int (b); print ("\n")
     end
     File 91: seq-point.tig
     

should always produce:

     $ tc -L seq-point.tig >seq-point.lir
     Example 92: tc -L seq-point.tig >seq-point.lir
     
     $ havm seq-point.lir
     1 -> 2
     2 -> 3
     3 -> 4
     14
     Example 93: havm seq-point.lir
     

independently of the what IR you ran. Note that it has nothing to do with the precedence of the operators!

In C, you have no such guarantee: the following program can give different results with different compilers and/or on different architectures.

     #include <stdio.h>
     
     int _a = 1;
     int a (int t)
     {
       ++_a;
       printf ("%d -> %d\n", t, _a);
       return _a;
     }
     
     int main (void)
     {
       int b = a (1) + a (2) * a (3);
       printf ("%d\n", b);
       return 0;
     }
     


Node: T6 Scheduling Samples, Previous: T6 Canonicalization Samples, Up: T6 Samples

T6 Scheduling Samples

Once you have canonicalized your eseq and call, you have to canonicalize cjumps: they must always be followed by their "false" label. This goes in two steps:

  1. Split in basic blocks.

    A basic block is a sequence of code starting with a label, ending with a jump (conditional or not), and with no jumps, no labels inside.

  2. Build the traces.

    Now put all the basic blocks into a single sequence.

In the following, the result of the whole conversion is visible.

The following examples highlights the need for new labels: at least one for the entry point, and one for the exit point:

     1 & 2
     File 94: 1-and-2.tig
     
     $ tc -L 1-and-2.tig
     /* == Low Level Intermediate representation. == */
     # Routine: Main
     label l`Main'
     # Prologue
     # Body
     seq
         label l3
         cjump ne
             const 1
             const 0
             name l0
             name l1
         label l1
         label l2
         jump
             name l4
         label l0
         jump
             name l2
         label l4
     seq end
     # Epilogue
     label end
     Example 95: tc -L 1-and-2.tig
     

The following example contains many jumps. Compare the hir to the lir:

     while 10 | 20 do if 30 | 40 then break else break
     File 96: broken-while.tig
     
     $ tc -H broken-while.tig
     /* == High Level Intermediate representation. == */
     # Routine: Main
     label l`Main'
     # Prologue
     # Body
     seq
         label l1
         seq
             cjump ne
                 const 10
                 const 0
                 name l8
                 name l9
             label l8
             cjump ne
                 const 1
                 const 0
                 name l2
                 name l0
             label l9
             cjump ne
                 const 20
                 const 0
                 name l2
                 name l0
         seq end
         label l2
         seq
             seq
                 cjump ne
                     const 30
                     const 0
                     name l6
                     name l7
                 label l6
                 cjump ne
                     const 1
                     const 0
                     name l3
                     name l4
                 label l7
                 cjump ne
                     const 40
                     const 0
                     name l3
                     name l4
             seq end
             label l3
             jump
                 name l0
             jump
                 name l5
             label l4
             jump
                 name l0
             label l5
         seq end
         jump
             name l1
         label l0
     seq end
     # Epilogue
     label end
     Example 97: tc -H broken-while.tig
     
     $ tc -L broken-while.tig
     /* == Low Level Intermediate representation. == */
     # Routine: Main
     label l`Main'
     # Prologue
     # Body
     seq
         label l10
         label l1
         cjump ne
             const 10
             const 0
             name l8
             name l9
         label l9
         cjump ne
             const 20
             const 0
             name l2
             name l0
         label l0
         jump
             name l11
         label l2
         cjump ne
             const 30
             const 0
             name l6
             name l7
         label l7
         cjump ne
             const 40
             const 0
             name l3
             name l4
         label l4
         jump
             name l0
         label l3
         jump
             name l0
         label l6
         cjump ne
             const 1
             const 0
             name l3
             name l13
         label l13
         jump
             name l4
         label l8
         cjump ne
             const 1
             const 0
             name l2
             name l14
         label l14
         jump
             name l0
         label l11
     seq end
     # Epilogue
     label end
     Example 98: tc -L broken-while.tig
     


Node: T6 Given Code, Next: , Previous: T6 Samples, Up: T6

T6 Given Code

Some code is provided: 2005-tc-6.1.tar.bz2. The transition from the previous versions can be done thanks to the following diffs: 2005-tc-4.3-6.0.diff, 2005-tc-6.0-6.1.diff.

It includes most of the canonicalization.


Node: T6 Code to Write, Previous: T6 Given Code, Up: T6

T6 Code to Write

Everything you need.


Node: T7, Next: , Previous: T6, Up: Compiler Stages

T7, Instruction Selection

2005-T7 delivery is Friday, July 4th 2003 at noon.
This section was last updated for EPITA-2004 and EPITA-2005 on 2003-07-02.

Please note that the 2005-T7 delivery is an option: there will be no grade, and a single upload will be accepted. The tests from T0 to T7 tests will be run on the tarball. The goal is to help you see your mistakes, and how your T7 is running to be able to proceed in peace onto T8. There will be no penalty if you don't take advantage of this possibility.

At the end of this stage, the compiler produces the very low level intermediate representation: ASSEM. This output is target dependent, and we aim at MIPS, as we use Mipsy to run it.


Node: T7 Goals, Next: , Up: T7

T7 Goals

Things to learn during this stage that you should remember:


Node: T7 Samples, Next: , Previous: T7 Goals, Up: T7

T7 Samples

The goal of T7 is straightforward: starting from LIR, generate the MIPS instructions, except that you don't have actual registers: we still heavily use Temps. Register allocation will be done in a later stage, T9.

     1 + 2 * 3
     File 99: seven.tig
     
     $ tc --inst-display seven.tig
     # == Final assembler ouput. == #
     # Routine: Main
     t_main:
             move    t5, $s0
             move    t6, $s1
             move    t7, $s2
             move    t8, $s3
             move    t9, $s4
             move    t10, $s5
             move    t11, $s6
             move    t12, $s7
     l0:
             li      t3, 2
             mul     t2, t3, 3
             li      t4, 1
             add     t1, t4, t2
     l1:
             move    $s0, t5
             move    $s1, t6
             move    $s2, t7
             move    $s3, t8
             move    $s4, t9
             move    $s5, t10
             move    $s6, t11
             move    $s7, t12
     
     Example 100: tc --inst-display seven.tig
     

Please, note that at this stage, the control flow analysis and the liveness analysis are not performed yet, therefore the compiler cannot know what registers are really to be saved. That's why in the previous output it saves "uselessly" all the callee-save registers on main entry. The next stage, which combines control flow analysis, liveness analysis, and register allocation, will make it useless. For your information, it results in:

     $ tc -sI seven.tig
     # == Final assembler ouput. == #
     # Routine: Main
     t_main:
             sw      $fp, ($sp)
             move    $fp, $sp
             sub     $sp, $sp, 8
             sw      $ra, -4 ($fp)
     l0:
             li      $t0, 2
             mul     $t1, $t0, 3
             li      $t0, 1
             add     $t0, $t0, $t1
     l1:
     
             lw      $ra, -4 ($fp)
             move    $sp, $fp
             lw      $fp, ($fp)
             jr      $ra
     Example 101: tc -sI seven.tig
     

A delicate part of this exercise is handling the function calls:

     let function add (x: int, y: int) : int = x + y
     in
       print_int (add (1, (add (2, 3)))); print ("\n")
     end
     File 102: add.tig
     
     $ tc -e --mipsy-display add.tig
     # == Final assembler ouput. == #
     # Routine: add
     l0:
             sw      $fp, -4 ($sp)
             move    $fp, $sp
             sub     $sp, $sp, 12
             sw      $ra, -8 ($fp)
             sw      $a0, ($fp)
             move    t0, $a1
             move    t1, $a2
             move    t7, $s0
             move    t8, $s1
             move    t9, $s2
             move    t10, $s3
             move    t11, $s4
             move    t12, $s5
             move    t13, $s6
             move    t14, $s7
     l2:
             add     t6, t0, t1
             move    $v0, t6
     l3:
             move    $s0, t7
             move    $s1, t8
             move    $s2, t9
             move    $s3, t10
             move    $s4, t11
             move    $s5, t12
             move    $s6, t13
             move    $s7, t14
     
             lw      $ra, -8 ($fp)
             move    $sp, $fp
             lw      $fp, -4 ($fp)
             jr      $ra
     
     .data
     l1:
             .word 1
             .asciiz "\n"
     .text
     
     # Routine: Main
     t_main:
             sw      $fp, ($sp)
             move    $fp, $sp
             sub     $sp, $sp, 8
             sw      $ra, -4 ($fp)
             move    t19, $s0
             move    t20, $s1
             move    t21, $s2
             move    t22, $s3
             move    t23, $s4
             move    t24, $s5
             move    t25, $s6
             move    t26, $s7
     l4:
             move    $a0, $fp
             li      t15, 2
             move    $a1, t15
             li      t16, 3
             move    $a2, t16
             jal     l0
             move    t4, $v0
             move    $a0, $fp
             li      t17, 1
             move    $a1, t17
             move    $a2, t4
             jal     l0
             move    t5, $v0
             move    $a0, t5
             jal     print_int
             la      t18, l1
             move    $a0, t18
             jal     print
     l5:
             move    $s0, t19
             move    $s1, t20
             move    $s2, t21
             move    $s3, t22
             move    $s4, t23
             move    $s5, t24
             move    $s6, t25
             move    $s7, t26
     
             lw      $ra, -4 ($fp)
             move    $sp, $fp
             lw      $fp, ($fp)
             jr      $ra
     Example 103: tc -e --mipsy-display add.tig
     

Once your function calls work properly, you can start using mipsy to check the behavior of your compiler.

     $ tc -eH add.tig >add.hir
     Example 104: tc -eH add.tig >add.hir
     
     $ havm add.hir
     6
     Example 105: havm add.hir
     

Unfortunately, you need to adjust the output of tc, using t123, to mipsy conventions: $x123.

     $ tc -eR --mipsy-display add.tig >add.instr
     Example 106: tc -eR --mipsy-display add.tig >add.instr
     
     $ sed -e's/\([^$a-z]\)t\([0-9][0-9]*\)/\1$x\2/g' add.instr >add.mipsy
     Example 107: sed -e's/\([^$a-z]\)t\([0-9][0-9]*\)/\1$x\2/g' add.instr >add.mipsy
     
     $ mipsy --unlimited-regs --execute add.mipsy
     6
     Example 108: mipsy --unlimited-regs --execute add.mipsy
     

You must also complete the runtime. No difference must be observable between a run with havm and another with mipsy:

     substring ("", 1, 1)
     File 109: substring-0-1-1.tig
     
     $ tc -eH substring-0-1-1.tig >substring-0-1-1.hir
     Example 110: tc -eH substring-0-1-1.tig >substring-0-1-1.hir
     
     $ havm substring-0-1-1.hir
     substring: arguments out of bounds
     =>120
     Example 111: havm substring-0-1-1.hir
     
     $ tc -e --mipsy-display substring-0-1-1.tig
     # == Final assembler ouput. == #
     .data
     l0:
             .word 0
             .asciiz ""
     .text
     
     # Routine: Main
     t_main:
             sw      $fp, ($sp)
             move    $fp, $sp
             sub     $sp, $sp, 8
             sw      $ra, -4 ($fp)
             move    t4, $s0
             move    t5, $s1
             move    t6, $s2
             move    t7, $s3
             move    t8, $s4
             move    t9, $s5
             move    t10, $s6
             move    t11, $s7
     l1:
             la      t1, l0
             move    $a0, t1
             li      t2, 1
             move    $a1, t2
             li      t3, 1
             move    $a2, t3
             jal     substring
     l2:
             move    $s0, t4
             move    $s1, t5
             move    $s2, t6
             move    $s3, t7
             move    $s4, t8
             move    $s5, t9
             move    $s6, t10
             move    $s7, t11
     
             lw      $ra, -4 ($fp)
             move    $sp, $fp
             lw      $fp, ($fp)
             jr      $ra
     Example 112: tc -e --mipsy-display substring-0-1-1.tig
     
     $ tc -eR --mipsy-display substring-0-1-1.tig >substring-0-1-1.instr
     Example 113: tc -eR --mipsy-display substring-0-1-1.tig >substring-0-1-1.instr
     
     $ sed -e's/\([^$a-z]\)t\([0-9][0-9]*\)/\1$x\2/g' substring-0-1-1.instr >substring-0-1-1.mipsy
     Example 114: sed -e's/\([^$a-z]\)t\([0-9][0-9]*\)/\1$x\2/g' substring-0-1-1.instr >substring-0-1-1.mipsy
     
     $ mipsy --unlimited-regs --execute substring-0-1-1.mipsy
     substring: arguments out of bounds
     =>120
     Example 115: mipsy --unlimited-regs --execute substring-0-1-1.mipsy
     


Node: T7 Given Code, Next: , Previous: T7 Samples, Up: T7

T7 Given Code

Below is listed where to find the tarball depending on your class. For more information about the T7 code delivered see src/target, src/assem, src/codegen.

2004-T7
A lot of code is provided. Actually, that's a real problem: since last year, the Tiger compiler has evolved a lot, and the integration of the new features will probably be painful. The most striking difference with last year being the Task handling.

The additional code is provided as:

There are two ways to continue the projects:

minor upgrade
If you do not want to upgrade your 2004 compiler into the 2005 form, just copy the relevant files from the tarball. See below. Adjust your driver so that --inst-compute and --inst-display be recognized. Of course, --inst-compute implies --lir-compute.
major upgrade
You want to upgrade to the 2005 system. Expect massive surgery... Contrary to the previous case, I would recommend starting from the tarball we delivered, and copy your files into there. For a start, copy all the files that are not in the new tarball: it's probably not wrong.
               # Be in the new tarball before running this.
               for i in $(find .)
               do
                 if test ! -f ../my-old-working-directory/$i; then
                   cp $i ../my-old-working-directory/$i
                 fi
               done
               

And then, build it step by step.


2005-t7
The additional code is provided as:


Node: T7 Code to Write, Previous: T7 Given Code, Up: T7

T7 Code to Write

There is not much code to write:

Information on MIPS 2000 assembly instructions may be found in SPIM manual.

Be aware that you are not required to fill in the blanks in the following places, as they are needed during register allocation only:


Node: T8, Next: , Previous: T7, Up: Compiler Stages

T8, Liveness Analysis

2005-T8 delivery is Friday, July 18th 2003 at noon.
This section was last updated for EPITA-2004 and EPITA-2005 on 2003-07-02.


Node: T8 Goals, Next: , Up: T8

T8 Goals

Things to learn during this stage that you should remember:


Node: T8 Samples, Next: , Previous: T8 Goals, Up: T8

T8 Samples

Branching is of course a most interesting feature to exercise:

     1 | 2 | 3
     File 116: ors.tig
     
     $ tc -I ors.tig
     # == Final assembler ouput. == #
     # Routine: Main
     t_main:
             move    t4, $s0
             move    t5, $s1
             move    t6, $s2
             move    t7, $s3
             move    t8, $s4
             move    t9, $s5
             move    t10, $s6
             move    t11, $s7
     l5:
             li      t1, 1
             bne     t1, 0, l3
     l4:
             li      t2, 2
             bne     t2, 0, l0
     l1:
     l2:
             j       l6
     l0:
             j       l2
     l3:
             li      t3, 1
             bne     t3, 0, l0
     l7:
             j       l1
     l6:
             move    $s0, t4
             move    $s1, t5
             move    $s2, t6
             move    $s3, t7
             move    $s4, t8
             move    $s5, t9
             move    $s6, t10
             move    $s7, t11
     
     Example 117: tc -I ors.tig
     
     $ tc -F ors.tig
     Example 118: tc -F ors.tig
     
     reports.ext/119.jpg
     
     File 119: Main-Main-flow.dot
     
     $ tc -V ors.tig
     Example 120: tc -V ors.tig
     
     reports.ext/121.jpg
     
     File 121: Main-Main-liveness.dot
     
     $ tc -N ors.tig
     Example 122: tc -N ors.tig
     
     reports.ext/123.jpg
     
     File 123: Main-Main-interference.dot
     


Node: T8 Given Code, Next: , Previous: T8 Samples, Up: T8

T8 Given Code

2004-T8
EPITA-2004 student are provided with the following code:
2005-T8
EPITA-2005 student are provided with the following code:

To read the description of the new modules, see src/graph, src/liveness.


Node: T8 Code to Write, Previous: T8 Given Code, Up: T8

T8 Code to Write

src/graph/graph.hh
src/graph/graph.hxx
Implement the topological sort.
src/liveness/flowgraph.hh
Write the constructor, which is where the FlowGraph is actually constructed from the assembly fragments.
src/liveness/liveness.cc
Write the constructor, which is where the Liveness (a decorated FlowGraph) is built from assembly instructions.
src/liveness/interference-graph.cc
In InterferenceGraph::compute_liveness, build the graph.


Node: T9, Previous: T8, Up: Compiler Stages

T9, Register Allocation

2005-T9 delivery is on Monday, September 8th 2003 at noon.
This section was last updated for EPITA-2004 and EPITA-2005 on 2003-08-19.

At the end of this stage, the compiler produces code that is runnable using Mipsy.


Node: T9 Goals, Next: , Up: T9

T8 Goals

Things to learn during this stage that you should remember:


Node: T9 Samples, Next: , Previous: T9 Goals, Up: T9

T9 Samples

This section will not demonstrate the output of the option -S, --asm-display, since it includes the Tiger runtime, which is quite long. We simply use -I, --instr-display which has the same effect once the registers allocated, i.e., once -s, --asm-compute executed. In short: we use -sI instead of -S to save place.

Allocating registers in the main function, when there is no register pressure is easy, as, in particular, there are no spills. A direct consequence is that many move are now useless, and have disappeared. For instance, the file of the example 99:

     $ tc -sI seven.tig
     # == Final assembler ouput. == #
     # Routine: Main
     t_main:
             sw      $fp, ($sp)
             move    $fp, $sp
             sub     $sp, $sp, 8
             sw      $ra, -4 ($fp)
     l0:
             li      $t0, 2
             mul     $t1, $t0, 3
             li      $t0, 1
             add     $t0, $t0, $t1
     l1:
     
             lw      $ra, -4 ($fp)
             move    $sp, $fp
             lw      $fp, ($fp)
             jr      $ra
     Example 124: tc -sI seven.tig
     
     $ tc -S seven.tig >seven.s
     Example 125: tc -S seven.tig >seven.s
     
     $ mipsy --execute seven.s
     Example 126: mipsy --execute seven.s
     

Another means to display the result of register allocation consists in reporting the mapping from temps to actual registers:

     $ tc -s --tempmap-display seven.tig
     /* Temporary map. */
     t1 -> $t0
     t2 -> $t1
     t3 -> $t0
     t4 -> $t0
     t5 -> $s0
     t6 -> $s1
     t7 -> $s2
     t8 -> $s3
     t9 -> $s4
     t10 -> $s5
     t11 -> $s6
     t12 -> $s7
     
     Example 127: tc -s --tempmap-display seven.tig
     

Of course it is much better to see what is going on:

     (print_int (1 + 2 * 3); print ("\n"))
     File 128: print-seven.tig
     
     $ tc -sI print-seven.tig
     # == Final assembler ouput. == #
     .data
     l0:
             .word 1
             .asciiz "\n"
     .text
     
     # Routine: Main
     t_main:
             sw      $fp, ($sp)
             move    $fp, $sp
             sub     $sp, $sp, 8
             sw      $ra, -4 ($fp)
     l1:
             li      $t0, 2
             mul     $t1, $t0, 3
             li      $t0, 1
             add     $a0, $t0, $t1
             jal     print_int
             la      $a0, l0
             jal     print
     l2:
     
             lw      $ra, -4 ($fp)
             move    $sp, $fp
             lw      $fp, ($fp)
             jr      $ra
     Example 129: tc -sI print-seven.tig
     
     $ tc -S print-seven.tig >print-seven.s
     Example 130: tc -S print-seven.tig >print-seven.s
     
     $ mipsy --execute print-seven.s
     7
     Example 131: mipsy --execute print-seven.s
     

To torture your compiler, you ought to use many temporaries. To be honest, ours is quite slow, it spends way too much time in register allocation.

     let
       var a00 := 00      var a55 := 55
       var a11 := 11      var a66 := 66
       var a22 := 22      var a77 := 77
       var a33 := 33      var a88 := 88
       var a44 := 44      var a99 := 99
     in
       print_int (0
                  +  a00 + a00 + a55 + a55
                  +  a11 + a11 + a66 + a66
                  +  a22 + a22 + a77 + a77
                  +  a33 + a33 + a88 + a88
                  +  a44 + a44 + a99 + a99);
       print ("\n")
     end
     File 132: print-many.tig
     
     $ tc -eIs --tempmap-display -I --time-report print-many.tig
     error-->Execution times (seconds)
     error--> 6: canon-compute        : 0      (    0%)   0      (    0%)   0.01   (   20%)
     error--> 7: inst-display         : 0.01   (   20%)   0      (    0%)   0      (    0%)
     error--> 8: liveness analysis    : 0.01   (   20%)   0      (    0%)   0.01   (   20%)
     error--> 8: liveness edges       : 0.01   (   20%)   0      (    0%)   0      (    0%)
     error--> 9: assign_colors        : 0.01   (   20%)   0      (    0%)   0      (    0%)
     error--> 9: coalesce             : 0      (    0%)   0      (    0%)   0.01   (   20%)
     error--> 9: register allocation  : 0      (    0%)   0      (    0%)   0.01   (   20%)
     error--> 9: simplify             : 0.01   (   20%)   0      (    0%)   0      (    0%)
     error-->Cumulated times (seconds)
     error--> 6: canon-compute        : 0      (    0%)   0      (    0%)   0.01   (   20%)
     error--> 7: inst-display         : 0.05   (  100%)   0      (    0%)   0.04   (   80%)
     error--> 8: liveness analysis    : 0.01   (   20%)   0      (    0%)   0.01   (   20%)
     error--> 9: asm-compute          : 0.04   (   80%)   0      (    0%)   0.04   (   80%)
     error--> 9: coalesce             : 0      (    0%)   0      (    0%)   0.01   (   20%)
     error--> 9: register allocation  : 0.04   (   80%)   0      (    0%)   0.04   (   80%)
     error--> rest                    : 0.05   (  100%)   0      (    0%)   0.05   (  100%)
     error--> TOTAL (seconds)         : 0.05   user,      0      system,    0.05   wall
     # == Final assembler ouput. == #
     .data
     l0:
             .word 1
             .asciiz "\n"
     .text
     
     # Routine: Main
     t_main:
             move    t33, $s0
             move    t34, $s1
             move    t35, $s2
             move    t36, $s3
             move    t37, $s4
             move    t38, $s5
             move    t39, $s6
             move    t40, $s7
     l1:
             li      t0, 0
             li      t1, 55
             li      t2, 11
             li      t3, 66
             li      t4, 22
             li      t5, 77
             li      t6, 33
             li      t7, 88
             li      t8, 44
             li      t9, 99
             li      t31, 0
             add     t30, t31, t0
             add     t29, t30, t0
             add     t28, t29, t1
             add     t27, t28, t1
             add     t26, t27, t2
             add     t25, t26, t2
             add     t24, t25, t3
             add     t23, t24, t3
             add     t22, t23, t4
             add     t21, t22, t4
             add     t20, t21, t5
             add     t19, t20, t5
             add     t18, t19, t6
             add     t17, t18, t6
             add     t16, t17, t7
             add     t15, t16, t7
             add     t14, t15, t8
             add     t13, t14, t8
             add     t12, t13, t9
             add     t11, t12, t9
             move    $a0, t11
             jal     print_int
             la      t32, l0
             move    $a0, t32
             jal     print
     l2:
             move    $s0, t33
             move    $s1, t34
             move    $s2, t35
             move    $s3, t36
             move    $s4, t37
             move    $s5, t38
             move    $s6, t39
             move    $s7, t40
     
     /* Temporary map. */
     t0 -> $a0
     t1 -> $t9
     t2 -> $t8
     t3 -> $t7
     t4 -> $t6
     t5 -> $t5
     t6 -> $t4
     t7 -> $t3
     t8 -> $t2
     t9 -> $t1
     t11 -> $a0
     t12 -> $t0
     t13 -> $t0
     t14 -> $t0
     t15 -> $t0
     t16 -> $t0
     t17 -> $t0
     t18 -> $t0
     t19 -> $t0
     t20 -> $t0
     t21 -> $t0
     t22 -> $t0
     t23 -> $t0
     t24 -> $t0
     t25 -> $t0
     t26 -> $t0
     t27 -> $t0
     t28 -> $t0
     t29 -> $t0
     t30 -> $t0
     t31 -> $t0
     t32 -> $a0
     t33 -> $s0
     t34 -> $s1
     t35 -> $s2
     t36 -> $s3
     t37 -> $s4
     t38 -> $s5
     t39 -> $s6
     t40 -> $s7
     
     # == Final assembler ouput. == #
     .data
     l0:
             .word 1
             .asciiz "\n"
     .text
     
     # Routine: Main
     t_main:
             sw      $fp, ($sp)
             move    $fp, $sp
             sub     $sp, $sp, 8
             sw      $ra, -4 ($fp)
     l1:
             li      $a0, 0
             li      $t9, 55
             li      $t8, 11
             li      $t7, 66
             li      $t6, 22
             li      $t5, 77
             li      $t4, 33
             li      $t3, 88
             li      $t2, 44
             li      $t1, 99
             li      $t0, 0
             add     $t0, $t0, $a0
             add     $t0, $t0, $a0
             add     $t0, $t0, $t9
             add     $t0, $t0, $t9
             add     $t0, $t0, $t8
             add     $t0, $t0, $t8
             add     $t0, $t0, $t7
             add     $t0, $t0, $t7
             add     $t0, $t0, $t6
             add     $t0, $t0, $t6
             add     $t0, $t0, $t5
             add     $t0, $t0, $t5
             add     $t0, $t0, $t4
             add     $t0, $t0, $t4
             add     $t0, $t0, $t3
             add     $t0, $t0, $t3
             add     $t0, $t0, $t2
             add     $t0, $t0, $t2
             add     $t0, $t0, $t1
             add     $a0, $t0, $t1
             jal     print_int
             la      $a0, l0
             jal     print
     l2:
     
             lw      $ra, -4 ($fp)
             move    $sp, $fp
             lw      $fp, ($fp)
             jr      $ra
     Example 133: tc -eIs --tempmap-display -I --time-report print-many.tig
     


Node: T9 Given Code, Next: , Previous: T9 Samples, Up: T9

T9 Given Code

The code is provided under the following forms:

To read the description of the new module, see src/regalloc.


Node: T9 Code to Write, Next: , Previous: T9 Given Code, Up: T9

T9 Code to Write

src/liveness/interference-graph.hh
src/liveness/interference-graph.cc
Unfortunately, the way the moves were encoded in the tarball we delivered for T8 is not right for T9. We could have used glue code to provide backward compatibility, but it was a poor solution yielding low quality code. Therefore the interface of the InterferenceGraph was upgraded, which will require some modifications in your existing code.

Rest assured that little work will actually be needed: the main modification is related to the fact that moves are now encoded as a list of pairs, while before we had a map mapping a node to the set of nodes in its move-related to.

src/regalloc/color.hh
Implement the graph coloring. The skeleton we provided is an exact copy of the implementation of the code suggest by Andrew Appel in the section 11.4 "Graph Coloring Implementation" of his book. A lot of comments that are verbatim copies of his comments are left in the code. Note that there are some error in this book, reported on his web page (see Modern Compiler Implementation).

Pay attention to misc::set: there is a lot of syntactic sugar provided to implement set operations. The code of Color can range from ugly and obfuscated to readable and very close to its specification.

src/regalloc/libregalloc.cc
Run the register allocation on each code fragment. Remove the useless moves.
src/codegen/mips/codegen.cc
If your compiler supports spills, implement Codegen::rewrite_program.


Node: T9 FAQ, Previous: T9 Code to Write, Up: T9

T9 FAQ

rv vs. $v0
Our graph coloring implementation cannot support aliases for hard registers: it thinks that if it has a different name, it is different. Since this is a reasonable claim, rather than torturing the algorithm until it accepts rv and $v0 designate a single guy, we decided to change the implementation of rv and fp in the frame module to use those of the current target: $v0 and $fp for MIPS. This has a strong influence on havm, of course. It was modified to support these changes, so make sure to use 0.18 or higher.


Node: Tools, Next: , Previous: Compiler Stages, Up: Top

Tools

This chapter aims at providing some helpful information about the various tools that you are likely to use to implement tc. It does not replace the reading of the genuine documentation, nevertheless, helpful tips are given. Feel free to contribute additional information.


Node: Modern Compiler Implementation, Next: , Up: Tools

Modern Compiler Implementation

The single most important tool for implementing the Tiger Project is the original book, Modern Compiler Implementation in C/Java/ML, by Andrew W. Appel, published by Cambridge University Press (New York, Cambridge). ISBN 0-521-58388-8/.

It is not possible to finish this project without having at least one copy per group. We provide a convenient mini Tiger Compiler Reference Manual that contains some information about the language but it does not cover all the details, and sometimes digging into the original book is required. This is on purpose, by virtue of due respect to the author of this valuable book.

Several copies are available at the EPITA library.

There are three flavors of this book:

C
The code samples are written in C. I strongly recommend that you avoid this edition, as C is not appropriate to describe the elaborate algorithms involved: most of the time, the simple ideas are destroyed with longuish unpleasant lines of code.
Java
The samples are written in Java. This book is the closest to the EPITA Tiger Project, since it is written in an object oriented language. Nevertheless, the modelisation is very poor, and therefore, don't be surprised if the EPITA project is significantly different. For a start, there is no Visitors at all. Of course the main purpose of the book is compilers, but it is not a reason for such a poor modelisation.
ML
This book, which is the "original", provides code samples in ML, which is a very adequate language to write compilers. Therefore it is very readable, even if you are not fluent in ML. I recommend this edition, unless you have severe problems with functional programming.

This book addresses many more issues than the sole Tiger Project as we implement it. In other words, it is an extremely interesting book which provides insights on garbage collection, object oriented and functional languages etc.

There is a dozen copies at the EPITA library, but buying it is a good idea.

Pay extra attention: there are several errors in the books, some of which are reported on Andrew Appel's pages (C Java, and ML), some which are not.


Node: Bibliography, Next: , Previous: Modern Compiler Implementation, Up: Tools

Bibliography

Compilers: Principles, Techniques and Tools
The Dragon Book
Written by Alfred V. Aho, Ravi Sethi, and Jeffrey D. Ullman
Published by Addison-Wesley 1986; ISBN 0-201-10088-6.

This book is the bible in compiler design. It has extensive insight on the whole architecture of compilers, provides a rigorous treatment for theoretical material etc. Nevertheless I would not recommend this book to EPITA students, because

it is getting old
It doesn't mention RISC, object orientation, functional, modern optimization techniques such as ssa, register allocation by graph coloring 3 etc.
it is fairly technical
The book can be hard to read for the beginner, contrary to Modern Compiler Implementation.

Nevertheless, curious readers will find valuable information about historically important compilers, people, papers etc. Reading the last section of each chapter (Bibliographical Notes) is a real pleasure for whom is interested.

It should be noted that the French edition, "Compilateurs: Principes, techniques et outils", was brilliantly translated by Pierre Boullier, Philippe Deschamp, Martin Jourdan, Bernard Lorho and Monique Lazaud: the pleasure is as good in French as it is in English.

Effective STL
Written by Scott Meyers
Published by Addison-Wesley; ISBN: 0-201-74962-9

A remarkable book that provides deep insight on the best practice with STL. Not only does it teach what's to be done, but it clearly shows why. A book that any C++ programmer should have read. See the Effective STL Addison-Wesley Page.

Lex & Yacc
Written John R. Levine, Tony Mason, Doug Brown
Published by O'Reilly & Associates; 2nd edition (October 1992); ISBN: 1-565-92000-7.

Because the books aims at a complete treatment of Lex and Yacc on a wide range of platforms, it provides too many details on material with little interest for us (e.g., we don't care about portability to other Lexes and Yacces), and too few details on material with big interest for us (more about exclusive start condition (Flex only), more about Bison only stuff, interaction with C++ etc.).

Modern Compiler Implementation in C, Java, ML
Written by Andrew W. Appel
Published by Cambridge University Press; ISBN: 0-521-58390-X

See Modern Compiler Implementation. In my humble opinion, most books give way too much emphasis to scanning and parsing, leaving little material to the rest of the compiler, or even nothing for advanced material. This book does not suffer this flaw.

Parsing Techniques - A Practical Guide
Written by Dick Grune and Ceriel J. Jacob
Published by the authors; ISBN: 0-13-651431-6

A remarkable review of all the parsing techniques. Because the book is out of print, its authors made it freely available: Parsing Techniques - A Practical Guide.

Writing Compilers and Interpreters - An Applied Approach Using C++
Written by Ronald Mak
Published by Wiley; Second Edition, ISBN: 0-471-11353-0

This book is not very interesting for us: the compiler material is not very advanced (no real ast, not a single line on optimization, register allocation is naive as the translation is stack based etc.), and the C++ material is not convincing (for a start, it is not standard C++ as it still uses #include <iostream.h> and the like, there is no use of STL etc.).

STL Home
SGI's STL Home Page, which includes the complete documentation online.


Node: The GNU Build System, Next: , Previous: Bibliography, Up: Tools

The GNU Build System

Automake is used to facilitate the writing of power Makefile. Autoconf is required by Automake: we don't not address portability issues for this project.

You may read the Autoconf documentation, and the Automake documentation. Using info is pleasant: info autoconf on any properly set up system. The Goat Book covers the whole GNU Build System: Autoconf, Automake and Libtool.


Node: Package Name and Version, Next: , Up: The GNU Build System

Package Name and Version

To set the name and version of your package, change the AC_INIT invocation. For instance, T4 for the bardec_f group gives AC_INIT(bardec_f-tiger, 4). Warning: Autoconf 2.53 smashes the underscores into dashes. To workaround this misfeature, use:

     AC_INIT([bardec_f-tiger], 4, [bardec_f@epita.fr], [bardec_f-tiger])
     


Node: Bootstrapping the Package, Next: , Previous: Package Name and Version, Up: The GNU Build System

Bootstrapping the Package

If something goes wrong, or if it is simply the first time you create configure.ac or a Makefile.am, you need to set up the GNU Build System. The simplest invocation is:

     $ autoreconf -fvi
     

The various files (configure, Makefile.in, etc.) are created. There is no need to run make distclean, or aclocal or whatever, before running autoreconf: it knows what to do.

Then invoke configure and make (see GCC):

     $ ./configure CC=gcc-3.2 CXX=g++-3.2
     $ make
     


Node: Making a Tarball, Previous: Bootstrapping the Package, Up: The GNU Build System

Making a Tarball

Once the package autotool'ed (see Bootstrapping the Package), once you can run a simple make, then you should be able to run make distcheck to set up the package.

The mission of make distcheck is to make sure everything will work properly. In particular it:

  1. creates the tarball (via make dist)
  2. untars the tarball
  3. configures the tarball in a separate directory (to avoid cluttering the source files with the built files). Note that if you ran the top level configure with some options (e.g., ./configure CC=gcc-3.2 CXX=g++-3.2), then these options will not be taken into account here. This means that running export CC=gcc-3.2; export CXX=g++-3.2 is a better way to make sure that these compilers will be used.
  4. runs make (and following targets) in paranoid mode. This mode consists in forbidding any change in the source tree, because if, when you run make something must be changed in the sources, then it means something is broken in the tarball. If, for instance, for some reason it wants to run autoconf to recreate configure, or if it complains that autom4te.cache cannot be created, then it means the tarball is broken! So track down the reason of the failure.
  5. runs make check
  6. runs make dist again.

If you just run make dist instead of make distcheck, then you might forget to include some files in the distribution. If you don't even run make dist, then not only some files might be missing, but you have no guarantee that the tarball will compile elsewhere (not to mention that we don't care about object files etc.).

Running make distcheck is the only means for you to check that the project will properly compile on our side. Not running distcheck is like turning off the type checking of your compiler: you hide the errors, you avoid them, instead of actually getting rid of them.

At this stage, if running make distcheck does not create bardec_f-tc-4.tar.bz2, then something is wrong in your package. Do not rename it, do not create the tarball by hand: something is rotten and be sure it will break on the examiner's machine.


Node: GCC, Next: , Previous: The GNU Build System, Up: Tools

GCC, The GNU Compiler Collection

We use GCC 3.2, which includes both gcc-3.2 and g++-3.2: the C and C++ compilers. Do not use older versions as they have poor compliance with the C++ standard. You are welcome to use more recent versions of GCC if you can use one, but the tests will be done with 3.2. Using a more recent version is often a good means to get better error messages if you can't understand what 3.2 is trying to say.

There are good patches floating around to improve GCC. In particular, you might want to use the bounds checking extension available on Herman ten Brugge Home Page.


Node: Flex & Bison, Next: , Previous: GCC, Up: Tools

Flex & Bison

We use Bison 1.875a which is able to produce a C++ parser. This Bison is unpublished, as the maintainers still have issues to fix. Nevertheless, it is usable, and perfectly functional for Tiger. It is installed in ~akim/bin, under the name bison. Be aware that Bison 1.875 produces buggy C++ parsers.

If you don't use this Bison, you will be in trouble. If you are willing to work at home, use bison-1.875a.tar.bz2.

The original papers on Lex and Yacc are:

Johnson, Stephen C. [1975].
Yacc: Yet Another Compiler Compiler. Computing Science Technical Report No. 32, Bell Laboratories, Murray hill, New Jersey.
Lesk, M. E. and E. Schmidt [1975].
Lex: A Lexical Analyzer Generator. Computing Science Technical Report No. 39, Bell Laboratories, Murray Hill, New Jersey.

The following introduction guides can help beginners:

Thomas Niemann.
A Compact Guide to Lex & Yacc.

An introduction to Lex and Yacc.

Collective Work
Programming with GNU Software.

Contains information about Autoconf, Automake, Gperf, Flex, Bison, and GCC.

The Bison documentation, and the Flex documentation are available for browsing.


Node: HAVM, Next: , Previous: Flex & Bison, Up: Tools

HAVM

HAVM is a Tree (hir or lir) programs interpreter. It was written by Robert Anisko so that EPITA students could exercise their compiler projects before the final jump to assembly code. It is implemented in Haskell, a pure non strict functional language very well suited for this kind of symbolic processing. HAVM was coined on both Haskell, and VM standing for Virtual Machine.

Information about HAVM can be found on HAVM Home Page, and feedback can be sent to LRDE's Projects Address.


Node: Mipsy, Next: , Previous: HAVM, Up: Tools

Mipsy

FIXME: Ben, some words about it please.


Node: SPIM, Next: , Previous: Mipsy, Up: Tools

SPIM

The following is taken from the SPIM documentation itself.

SPIM S20 is a simulator that runs programs for the MIPS R2R3000 RISC computers. SPIM can read and immediately execute files containing assembly language. SPIM is a self-contained system for running these programs and contains a debugger and interface to a few operating system services.

The architecture of the MIPS computers is simple and regular, which makes it easy to learn and understand. The processor contains 32 general-purpose 32-bit registers and a well-designed instruction set that make it a propitious target for generating code in a compiler.

However, the obvious question is: why use a simulator when many people have workstations that contain a hardware, and hence significantly faster, implementation of this computer? One reason is that these workstations are not generally available. Another reason is that these machine will not persist for many years because of the rapid progress leading to new and faster computers. Unfortunately, the trend is to make computers faster by executing several instructions concurrently, which makes their architecture more difficult to understand and program. The MIPS architecture may be the epitome of a simple, clean RISC machine.

In addition, simulators can provide a better environment for low-level programming than an actual machine because they can detect more errors and provide more features than an actual computer. For example, SPIM has a X-window interface that is better than most debuggers for the actual machines.

Finally, simulators are an useful tool for studying computers and the programs that run on them. Because they are implemented in software, not silicon, they can be easily modified to add new instructions, build new systems such as multiprocessors, or simply to collect data.

SPIM is written and maintained by James R. Larus.


Node: Doxygen, Previous: SPIM, Up: Tools

Doxygen

We use Doxygen as the standard tool for producing the developer's documentation of the project. Its features must be used to produce good documentation, with an explanation of the role of the arguments etc. The quality of the documentation will be part of the notation. Details on how to use proper comments are given in the Doxygen Manual.

The documentation produced by Doxygen must not be included, but the target html must produce the html documentation in the doc/html directory.


Node: Appendices, Previous: Tools, Up: Top

Appendices


Node: Glossary, Next: , Up: Appendices

Glossary

Contributions to this section (as for the rest of this documentation) will be greatly appreciated.

activation block
Portion of dynamically allocated memory holding all the information a (recursive) function needs at runtime. It typically contains arguments, automatic local variables etc. Implemented by the class frame::Frame (see T5).
build
The machine/architecture on which the program is being built. For instance, EPITA students typically build their compiler on NetBSD. Contrast with "target" and "host".
HAVM
HAVM is a Tree (hir or lir) programs interpreter. See HAVM.
host
The machine/architecture on which the program is run. For instance, EPITA students typically run their Tiger Compiler on NetBSD. Contrast with "build and "target".
IA32
The official new name for the i386 architecture.
snippet
A piece of something, e.g., "code snippet".
stack frame
Synonym for "activation block".
SPIM
SPIM S20 is a simulator that runs programs for the MIPS R2R3000 RISC computers. See SPIM.
target
The machine (or language) aimed at by a compiling tool. For instance, our target is principally MIPS. Compare with "build" and "host".


Node: GNU Free Documentation License, Next: , Previous: Glossary, Up: Appendices

GNU Free Documentation License

Version 1.1, March 2000
     Copyright © 2000 Free Software Foundation, Inc.
     59 Temple Place, Suite 330, Boston, MA  02111-1307, USA
     
     Everyone is permitted to copy and distribute verbatim copies
     of this license document, but changing it is not allowed.
     
  1. PREAMBLE

    The purpose of this License is to make a manual, textbook, or other written document free in the sense of freedom: to assure everyone the effective freedom to copy and redistribute it, with or without modifying it, either commercially or noncommercially. Secondarily, this License preserves for the author and publisher a way to get credit for their work, while not being considered responsible for modifications made by others.

    This License is a kind of "copyleft", which means that derivative works of the document must themselves be free in the same sense. It complements the GNU General Public License, which is a copyleft license designed for free software.

    We have designed this License in order to use it for manuals for free software, because free software needs free documentation: a free program should come with manuals providing the same freedoms that the software does. But this License is not limited to software manuals; it can be used for any textual work, regardless of subject matter or whether it is published as a printed book. We recommend this License principally for works whose purpose is instruction or reference.

  2. APPLICABILITY AND DEFINITIONS

    This License applies to any manual or other work that contains a notice placed by the copyright holder saying it can be distributed under the terms of this License. The "Document", below, refers to any such manual or work. Any member of the public is a licensee, and is addressed as "you".

    A "Modified Version" of the Document means any work containing the Document or a portion of it, either copied verbatim, or with modifications and/or translated into another language.

    A "Secondary Section" is a named appendix or a front-matter section of the Document that deals exclusively with the relationship of the publishers or authors of the Document to the Document's overall subject (or to related matters) and contains nothing that could fall directly within that overall subject. (For example, if the Document is in part a textbook of mathematics, a Secondary Section may not explain any mathematics.) The relationship could be a matter of historical connection with the subject or with related matters, or of legal, commercial, philosophical, ethical or political position regarding them.

    The "Invariant Sections" are certain Secondary Sections whose titles are designated, as being those of Invariant Sections, in the notice that says that the Document is released under this License.

    The "Cover Texts" are certain short passages of text that are listed, as Front-Cover Texts or Back-Cover Texts, in the notice that says that the Document is released under this License.

    A "Transparent" copy of the Document means a machine-readable copy, represented in a format whose specification is available to the general public, whose contents can be viewed and edited directly and straightforwardly with generic text editors or (for images composed of pixels) generic paint programs or (for drawings) some widely available drawing editor, and that is suitable for input to text formatters or for automatic translation to a variety of formats suitable for input to text formatters. A copy made in an otherwise Transparent file format whose markup has been designed to thwart or discourage subsequent modification by readers is not Transparent. A copy that is not "Transparent" is called "Opaque".

    Examples of suitable formats for Transparent copies include plain ASCII without markup, Texinfo input format, LaTeX input format, SGML or XML using a publicly available DTD, and standard-conforming simple HTML designed for human modification. Opaque formats include PostScript, PDF, proprietary formats that can be read and edited only by proprietary word processors, SGML or XML for which the DTD and/or processing tools are not generally available, and the machine-generated HTML produced by some word processors for output purposes only.

    The "Title Page" means, for a printed book, the title page itself, plus such following pages as are needed to hold, legibly, the material this License requires to appear in the title page. For works in formats which do not have any title page as such, "Title Page" means the text near the most prominent appearance of the work's title, preceding the beginning of the body of the text.

  3. VERBATIM COPYING

    You may copy and distribute the Document in any medium, either commercially or noncommercially, provided that this License, the copyright notices, and the license notice saying this License applies to the Document are reproduced in all copies, and that you add no other conditions whatsoever to those of this License. You may not use technical measures to obstruct or control the reading or further copying of the copies you make or distribute. However, you may accept compensation in exchange for copies. If you distribute a large enough number of copies you must also follow the conditions in section 3.

    You may also lend copies, under the same conditions stated above, and you may publicly display copies.

  4. COPYING IN QUANTITY

    If you publish printed copies of the Document numbering more than 100, and the Document's license notice requires Cover Texts, you must enclose the copies in covers that carry, clearly and legibly, all these Cover Texts: Front-Cover Texts on the front cover, and Back-Cover Texts on the back cover. Both covers must also clearly and legibly identify you as the publisher of these copies. The front cover must present the full title with all words of the title equally prominent and visible. You may add other material on the covers in addition. Copying with changes limited to the covers, as long as they preserve the title of the Document and satisfy these conditions, can be treated as verbatim copying in other respects.

    If the required texts for either cover are too voluminous to fit legibly, you should put the first ones listed (as many as fit reasonably) on the actual cover, and continue the rest onto adjacent pages.

    If you publish or distribute Opaque copies of the Document numbering more than 100, you must either include a machine-readable Transparent copy along with each Opaque copy, or state in or with each Opaque copy a publicly-accessible computer-network location containing a complete Transparent copy of the Document, free of added material, which the general network-using public has access to download anonymously at no charge using public-standard network protocols. If you use the latter option, you must take reasonably prudent steps, when you begin distribution of Opaque copies in quantity, to ensure that this Transparent copy will remain thus accessible at the stated location until at least one year after the last time you distribute an Opaque copy (directly or through your agents or retailers) of that edition to the public.

    It is requested, but not required, that you contact the authors of the Document well before redistributing any large number of copies, to give them a chance to provide you with an updated version of the Document.

  5. MODIFICATIONS

    You may copy and distribute a Modified Version of the Document under the conditions of sections 2 and 3 above, provided that you release the Modified Version under precisely this License, with the Modified Version filling the role of the Document, thus licensing distribution and modification of the Modified Version to whoever possesses a copy of it. In addition, you must do these things in the Modified Version:

    1. Use in the Title Page (and on the covers, if any) a title distinct from that of the Document, and from those of previous versions (which should, if there were any, be listed in the History section of the Document). You may use the same title as a previous version if the original publisher of that version gives permission.
    2. List on the Title Page, as authors, one or more persons or entities responsible for authorship of the modifications in the Modified Version, together with at least five of the principal authors of the Document (all of its principal authors, if it has less than five).
    3. State on the Title page the name of the publisher of the Modified Version, as the publisher.
    4. Preserve all the copyright notices of the Document.
    5. Add an appropriate copyright notice for your modifications adjacent to the other copyright notices.
    6. Include, immediately after the copyright notices, a license notice giving the public permission to use the Modified Version under the terms of this License, in the form shown in the Addendum below.
    7. Preserve in that license notice the full lists of Invariant Sections and required Cover Texts given in the Document's license notice.
    8. Include an unaltered copy of this License.
    9. Preserve the section entitled "History", and its title, and add to it an item stating at least the title, year, new authors, and publisher of the Modified Version as given on the Title Page. If there is no section entitled "History" in the Document, create one stating the title, year, authors, and publisher of the Document as given on its Title Page, then add an item describing the Modified Version as stated in the previous sentence.
    10. Preserve the network location, if any, given in the Document for public access to a Transparent copy of the Document, and likewise the network locations given in the Document for previous versions it was based on. These may be placed in the "History" section. You may omit a network location for a work that was published at least four years before the Document itself, or if the original publisher of the version it refers to gives permission.
    11. In any section entitled "Acknowledgments" or "Dedications", preserve the section's title, and preserve in the section all the substance and tone of each of the contributor acknowledgments and/or dedications given therein.
    12. Preserve all the Invariant Sections of the Document, unaltered in their text and in their titles. Section numbers or the equivalent are not considered part of the section titles.
    13. Delete any section entitled "Endorsements". Such a section may not be included in the Modified Version.
    14. Do not retitle any existing section as "Endorsements" or to conflict in title with any Invariant Section.

    If the Modified Version includes new front-matter sections or appendices that qualify as Secondary Sections and contain no material copied from the Document, you may at your option designate some or all of these sections as invariant. To do this, add their titles to the list of Invariant Sections in the Modified Version's license notice. These titles must be distinct from any other section titles.

    You may add a section entitled "Endorsements", provided it contains nothing but endorsements of your Modified Version by various parties--for example, statements of peer review or that the text has been approved by an organization as the authoritative definition of a standard.

    You may add a passage of up to five words as a Front-Cover Text, and a passage of up to 25 words as a Back-Cover Text, to the end of the list of Cover Texts in the Modified Version. Only one passage of Front-Cover Text and one of Back-Cover Text may be added by (or through arrangements made by) any one entity. If the Document already includes a cover text for the same cover, previously added by you or by arrangement made by the same entity you are acting on behalf of, you may not add another; but you may replace the old one, on explicit permission from the previous publisher that added the old one.

    The author(s) and publisher(s) of the Document do not by this License give permission to use their names for publicity for or to assert or imply endorsement of any Modified Version.

  6. COMBINING DOCUMENTS

    You may combine the Document with other documents released under this License, under the terms defined in section 4 above for modified versions, provided that you include in the combination all of the Invariant Sections of all of the original documents, unmodified, and list them all as Invariant Sections of your combined work in its license notice.

    The combined work need only contain one copy of this License, and multiple identical Invariant Sections may be replaced with a single copy. If there are multiple Invariant Sections with the same name but different contents, make the title of each such section unique by adding at the end of it, in parentheses, the name of the original author or publisher of that section if known, or else a unique number. Make the same adjustment to the section titles in the list of Invariant Sections in the license notice of the combined work.

    In the combination, you must combine any sections entitled "History" in the various original documents, forming one section entitled "History"; likewise combine any sections entitled "Acknowledgments", and any sections entitled "Dedications". You must delete all sections entitled "Endorsements."

  7. COLLECTIONS OF DOCUMENTS

    You may make a collection consisting of the Document and other documents released under this License, and replace the individual copies of this License in the various documents with a single copy that is included in the collection, provided that you follow the rules of this License for verbatim copying of each of the documents in all other respects.

    You may extract a single document from such a collection, and distribute it individually under this License, provided you insert a copy of this License into the extracted document, and follow this License in all other respects regarding verbatim copying of that document.

  8. AGGREGATION WITH INDEPENDENT WORKS

    A compilation of the Document or its derivatives with other separate and independent documents or works, in or on a volume of a storage or distribution medium, does not as a whole count as a Modified Version of the Document, provided no compilation copyright is claimed for the compilation. Such a compilation is called an "aggregate", and this License does not apply to the other self-contained works thus compiled with the Document, on account of their being thus compiled, if they are not themselves derivative works of the Document.

    If the Cover Text requirement of section 3 is applicable to these copies of the Document, then if the Document is less than one quarter of the entire aggregate, the Document's Cover Texts may be placed on covers that surround only the Document within the aggregate. Otherwise they must appear on covers around the whole aggregate.

  9. TRANSLATION

    Translation is considered a kind of modification, so you may distribute translations of the Document under the terms of section 4. Replacing Invariant Sections with translations requires special permission from their copyright holders, but you may include translations of some or all Invariant Sections in addition to the original versions of these Invariant Sections. You may include a translation of this License provided that you also include the original English version of this License. In case of a disagreement between the translation and the original English version of this License, the original English version will prevail.

  10. TERMINATION

    You may not copy, modify, sublicense, or distribute the Document except as expressly provided for under this License. Any other attempt to copy, modify, sublicense or distribute the Document is void, and will automatically terminate your rights under this License. However, parties who have received copies, or rights, from you under this License will not have their licenses terminated so long as such parties remain in full compliance.

  11. FUTURE REVISIONS OF THIS LICENSE

    The Free Software Foundation may publish new, revised versions of the GNU Free Documentation License from time to time. Such new versions will be similar in spirit to the present version, but may differ in detail to address new problems or concerns. See http://www.gnu.org/copyleft/.

    Each version of the License is given a distinguishing version number. If the Document specifies that a particular numbered version of this License "or any later version" applies to it, you have the option of following the terms and conditions either of that specified version or of any later version that has been published (not as a draft) by the Free Software Foundation. If the Document does not specify a version number of this License, you may choose any version ever published (not as a draft) by the Free Software Foundation.

ADDENDUM: How to use this License for your documents

To use this License in a document you have written, include a copy of the License in the document and put the following copyright and license notices just after the title page:

       Copyright (C)  year  your name.
       Permission is granted to copy, distribute and/or modify this document
       under the terms of the GNU Free Documentation License, Version 1.1
       or any later version published by the Free Software Foundation;
       with the Invariant Sections being list their titles, with the
       Front-Cover Texts being list, and with the Back-Cover Texts being list.
       A copy of the license is included in the section entitled ``GNU
       Free Documentation License''.
     

If you have no Invariant Sections, write "with no Invariant Sections" instead of saying which ones are invariant. If you have no Front-Cover Texts, write "no Front-Cover Texts" instead of "Front-Cover Texts being list"; likewise for Back-Cover Texts.

If your document contains nontrivial examples of program code, we recommend releasing these examples in parallel under your choice of free software license, such as the GNU General Public License, to permit their use in free software.


Node: Index, Previous: GNU Free Documentation License, Up: Appendices

Index


Footnotes

  1. See the shift of language? From tarball to distribution.

  2. In the future --bound-checking will be --bounds-checking so you might want to implement both.

  3. To be fair, the Dragon Book leaves a single page (not sheet) to graph coloring.