Re: University Degrees in Compiler Design.

Matthew Webb <matthew.webb@net1.demon.co.uk> wrote:
> > Are there any uni placments in compiler design in the UK? I have
> >found some for AI but I would also like to look into compiler design.


Mohd-Hanafiah Abdullah wrote:
> I am not aware of any university in the US or UK that offers a degree
> or graduate program in compilers specifically.


Perhaps not, but I think that a seriously advanced course in compiler
design ought to be available for those who want it. No matter what
university you go to, all the people who teach compilers have this
same complaint, that there's too much material to cover in one
semester.


And the professional-quality compilers that major companies come out
with do use techniques and optimizations that, even in the most
advanced compiler design courses, a typical student will only be
taught a half-dozen or a dozen of.


If we had an ideal situation for teaching; ie, an honors class of grad
students, who'd already been through the Dragon Book and understood
the material (and still owned it to use for reference), and 5
classroom hours a week for a full year to teach them advanced compiler
design, What texts would you use and what techniques would you cover??


I'm asking, because that's the class I'd like to be in... And since
nobody offers it, I'm trying to do it independently. My objective is
to become able to produce the "Best Compiler Ever" for Scheme -- a
language which has garbage collection, continuations, macro syntax,
first class functions, runtime expression evaluation, dynamic types,
and extended-precision numerics. I want to take unicode source and
compile it down to straight, globally-optimized machine code.


So far, I've been through the Dragon Book, "Structure & Interpretation
of Computer Programs", "Lisp In Small Pieces", "How Debuggers Work",
"Garbage Collection Algorithms", half-a-dozen papers on macros and
syntax extension, another three on compiling in the presence of
continuations and one on the joys of incremental compilation.


So far, I've learned a lot of material but I'm still experimenting and
looking for ways to make it all (or most of it anyway) work
*together*, in the context of either a language or a target code very
different from what most of it was written assuming. It's always the
integration of ideas into a whole that's the hardest part and requires
the deepest understanding.


Right now, the book I'm coveting is "Advanced Compiler Design and
Implementation" by Steven Muchnik. It's ninety bucks, but I should be
able to do it in my next paycheck. The weird thing is, I don't even
really know what's in it -- It's just new material. Reading it, I
hope, will bump other concepts running around in my head and make them
fit together better, or show me new relationships between them, or
something. In other words, I don't know whether the specific
techniques and optimizations it covers are things I've already covered
-- I'm just hoping for the background or remark or commentary or
whatever that will give me new perspecives on this stuff, weld stuff
I've studied together and make it into a better and more seamless
whole.


What else should I be looking at? What's good for tying together the
concepts and making wholes out of them, the way the Dragon Book did
for the basic techniques?


Ray Dillinger
--

Return to the comp.compilers page.
Search the comp.compilers archives again.