The Stages of Compilation


This describes the basic stages of compilation in a very non-traditional way ...

From: Mateo.Burtch@Eng.Sun.COM (Room 101)
Subject: YADFH (Yet Another D*** Friday Hack)


Hi. This isn't exactly a hack, and thus probably doesn't deserve to be
called a "Friday Hack," but SOOOOO many people have been coming up to
me in the shower, saying, Please, tell us the secrets of how the
compilation process works, that I finally decided to share with you the
old family recipe for:


SOURCE FILE: This is the basic "code" that the engineer writes. The
compiler will take code such as

dweeble(flab, krimjaw)
sneet flab *(**smicknat[])(blugnut);
blook krimjaw;
	blark snapdaddle liederhosen ;
	if (fleb <= OAK_TREE)
		while (trousers(liederhosen))

and turn it into output that means, roughly, "point the hose away from
yourself while watering."

Obviously, the compiler is a lot smarter than we are.

PREPROCESSOR: The preprocessor takes all sorts of special directives,
like #ifdefs, and converts them into conditional statements (known as
"#ifdefs") that the compiler uses to prepare the file (or "#ifdef") for
processing. (This is a technical simplification that is in most, if not
all, aspects wrong.)

The preprocessor also strips the source file of comments (not to be
confused with #ifdefs), leaving it a shaken husk of its former self.
These comments are then pieced together by a separate function and used
for insulation in Building 12.

COMPILER: The heart of the whole process. The compiler takes the
language words, such as "if," "for," and
"help!" and turns it into the bits, bytes, and subroutines
that give employment to a whole host of socially challenged people.

OPTIMIZATION: This stage makes the generated code as efficient as
possible. The optimizer does this by carefully trimming the odd and
dangly parts off of numerals like "5" and "9" until
all the numbers are either "1"s or "0"s.

ASSEMBLER: The assembler takes "machine code" (low-level
instructions done by poorly-paid workers in windowless sweatshops
called "assembler lines") and does something with it.

LINKER: Before we're done, we must link together all the various object
files with libraries containing macros and canned routines. Don't ask
why--it's like salmon swimming upstream to spawn. You just do it.

These libraries perform a special job in the grand pageant known as
programming. The pre-written macros and functions they contain allow
the programmer to save valuable time when writing a program, while the
arcane linker options and obscure environment variables slow the
programmer down just as much. This is known as "run-time
equilibrium" and is rather similar to horizontal bungee-jumping.

EXECUTABLE: This is the final product, the finished masterpiece, the
piece de resistance of the whole process. This file is what the
engineer had in mind when he or she started out.

It's called "core."


Send email to

Department of Computer Science
University of California at Davis
Davis, CA 95616-8562

Page last modified on 5/19/99