Hi,
I just want to know the language in which the C compiler was written. Please say something other than C.
A:
Depending on which C compiler, it was likely written in assembly, then it eventually probably became self-compiling so then parts were written in C.
You may browse the source for GCC for yourself at http://gcc.gnu.org/viewcvs/branches/
Daniel A. White
2010-08-20 11:46:25
Which begs the question, what was used to write assembly?
JackN
2010-08-20 11:54:24
@JackN Perhaps punch cards, etc.
ChrisW
2010-08-20 12:06:09
At some point you get down to flipping switches to compose the bits of each machine instruction sent to the processor. Any lower than that and it's hardware engineering rather than programming.
Tyler McHenry
2010-08-20 12:08:12
@JackN - assembly was used to write assembly. Before that, you could argue machine code - but probably even that was hand-assembled assembly. BTW - hand assembly isn't so difficult (for a simple processor) as some might think. In the 8-bit days, it was common for people who didn't have assemblers to hand-assemble little fragments to call from Basic programs. I did some of this, though that probably means I chose the wrong kind of "misspent youth". Once upon a time, I wrote simple 6502 assemblers as a kind of next step up from "hello world" for a new language.
Steve314
2010-08-20 12:10:26
@Steve314 I wrote an assembler by writing a preprocessor in Basic (because a Basic interpreter was included with the O/S) and then piping it into the `a` (assemble) command of `debug.com`; I then wrote another in assembly.
ChrisW
2010-08-20 13:42:42
@ChrisW - I wrote a pipe-it-through-debug assembler once too! It had to do the whole thing several times over, hoping that the addresses for labels would stabilize - took forever for anything substantial. I honestly don't remember what I wrote it in - maybe GW-Basic, not sure. I never wrote a real i86 assembler because there's too many mnemonics and addressing modes to worry about - it distracts from the point of the next-step-up-from-hello-world which is arithmetic and logic, file handling, string processing and some basic data structures.
Steve314
2010-08-20 18:23:48
+2
A:
GCC is written in C. The majority of C compilers are written in C.
There is a boot-strapping phase when first producing a compiler for a language (any language that has pretensions to be able to compile its own compiler - COBOL is one plausible exception, but there are many others) on a given platform, but once you have a compiler, then you write the compiler in that language.
All else apart, doing it in assembler is too expensive.
Jonathan Leffler
2010-08-20 11:48:42
On "but once you have a compiler, then you write the compiler in that language" - it ain't necessarily so. I lot of languages *always* have their compilers/interpreters written in C, and a lot of compilers generate C rather than machine code - it's one of the reasons C gained a reputation as a kind of high level assembly language.
Steve314
2010-08-20 12:17:15
@Steve: There are languages that don't pretend to be able to compile their own compiler - those languages are not covered by my statement.
Jonathan Leffler
2010-08-20 12:20:28
GCC is being moved to C++ and Python has a pure python interpreter (pypy) but it's not used in actual work. Many languages can be used to implement themselves yet are not. clisp is largely written in C.
Nathon
2010-08-20 13:09:49
@Jonathon - just because you write a language that's well suited to writing compilers, doesn't automatically mean you want to self-host. The whole boostrap thing is a hassle for a start. Of course it's common to do it anyway, but that doesn't prove anything. There's no rule that says you have to - only a certain amount of misplaced pride.
Steve314
2010-08-21 11:50:07
+4
A:
Nearly all major C compilers are written in C. You might think there's a chicken-and-egg problem with this, but there's not. The process is called bootstrapping.
Tyler McHenry
2010-08-20 11:49:06
+6
A:
Here's an excellent read: Reflections on Trusting Trust by Ken Thompson. Starts off with an overview of how the first C compilers were written. The boot-strapping technique to be precise. May not answer your question directly but gives you some insight.
dirkgently
2010-08-20 11:50:00
It's one paper I keep going back to from time to time. A personal favorite.
dirkgently
2010-08-20 12:40:10
A:
In the old days, people would write a small subset of the C language in assembler, and then use that to "bootstrap" compile a better C compiler written in C. These days it's more common to make a C compiler for a new architecture by cross compiling from an architecture that already works. I believe there are very few bits of, for instance, the gcc compiler, that aren't written in C or C++.
Paul Tomblin
2010-08-20 11:50:29
+2
A:
The very original C compiler was written (by K&R) in a predecessor language called B, or maybe BCPL. But once the C compiler was working well enough, they converted over to C and began using each successive version to compile the next.
Many of the bizarre features of C such as pre- and post-increment operators exist because (a) they represented special addressing modes on the PDP-11 on which the first C was developed, or (b) they helped the compiler fit in memory while compiling its own next version.
So that's the rest of the story.
Ian
2010-08-20 11:52:46
@leppie - just checking - you did intend "perl!" as opposed to "peril"? I understand they're much the same thing, but...
Steve314
2010-08-20 12:00:39