Two questions about compiler design

I've basically read most of the dragon book, and according to that book you
should generate intermediate code then divide it into basic blocks+flow
chart then build a DAG for each basic block to do code motion etc etc. Now,
all this seems great to me. But the book i have is from -79 and i'm
wondering if this is pretty much how it's done now days as well?


Also, i want to use three-adress code for the intermediate code becuase it
seems to be easiest to do optimisation on, But i don't just want to pass it
backend to produce machine code. I also want to run the intermediate code in
an interpreter. So i was wondering if it's better to make my own
intermediate code if there already is a good standard one defined which
would suite this need. If anyone ever used quickbasic you know that you can
both run your code in interpreted mode as well as compile it. You can also
use ordinary libraries in interpreted mode. So what i had in mind is
something like QuickBASIC.


/ Blitz

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