Re: Implementation and Optimization of exceptions ( static analysis possible? )

pardo@cs.washington.edu (David Keppel) writes:
> I believe you can also follow each `call' instruction with a word of
> data that effectively implements the address range table and/or a call
> instruction that jumps to code that handles the exception. However,
> the normal procedure call/return needs to be set up to jump past the
> exception code that follows the call and the code growth is
> proporitional to the number of callers not the number of callees.


I believe this is equivalent in space to the address range table, because
you need, for each call site, code or a table to tell the exception code
how to unwind the stack. (For C++, you'd also need some information about
what objects need to be destroyed.)


I think the only advantage to this scheme, then, is that the lookup for
finding the address range table is dead simple. Not a trivial advantage,
to be sure, but I don't see that it buys you a lot of space savings.


--
Bill Leonard
Harris Computer Systems Corporation
2101 W. Cypress Creek Road
Fort Lauderdale, FL 33309
Bill.Leonard@mail.hcsc.com
--

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