Re: Good practical language and OS agnostic text?

On 4/22/2012 5:55 AM, Tomasz Kowaltowski wrote:
>> [... And I have to
>> say that if you have CS degree and are unable to figure out what a
>> LALR parser does, there's something wrong with your CS degree. -John]
>
> I agree with our moderator and am somewhat surprised by this
> discussion. IMHO compiler construction requires knowledge of many
> different techniques and is usually an advanced course in CS
> undergraduate programs. Trying to do it without knowing the basics
> and lots of experience may be an amusing pastime but cannot be
> considered a serious endeavor. I don't mean you necessarily need a
> formal CS degree but you do have be able to read (and understand!)
> more advanced material.


well, I think it depends a lot on the material...


for example, many books are fairly straightforward:
they describe the process, general stuff going on, ...


so, then, all is good.




in another case, I went and started trying to read a book (I forget
the name): introduces general topic, starts mentioning stuff
"Hindley-Milner Type Inference" and "Type Polymorphism as applied to
the Lambda Calculus" and so on, with large volumes of rather
opaque-looking mathematical notation.


I think I didn't really get too far in this one (before brain-melting
set in), before going off and looking at other stuff.


I couldn't really see how any of this was terribly relevant in a world
where "type" generally means "int" vs "float" and maybe dealing with
things like pointer and array operations, and where "polymorphism" is
mostly "one of those words that apparently has something to do with how
the class hierarchy works or similar".


it is enough to say "int + int -> int", "int + float -> float", ...




I have personally a difficult enough time trying to fully understand how
exactly SSA-form works, much less trying to implement a code-generator
based on it, hence my continued general use of stack-machines as the
conceptual model (doesn't mean "logical" stack operations map directly
to "physical" locations or operations though). at least I generally
understand stack machines.


but, then again, my track-record for writing "good" native code
generators (of those few "sufficiently complete to work") is sadly not
very good (my first real attempts in this area starting around 2007 or so).
[LALR really isn't that hard to understand, a state machine with a
stack. I'm not saying every compiler should use it, but I am saying
that it's no more complicated than other things a CS major should have
mastered. -John]

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