Re: register allocation: basic blocks, liveness and next use

Indeed, Bob Morgan's book says he treats procedure calls similarly to
expression handling because they don't change the flow of control.
Similarly, Steve Muchnick prefers to do likewise because it results in
longer blocks, but makes exceptions for Fortran routines with multiple
returns and for setjmp()/longjmp(). Bill Wolf's book doesn't appear
to mention the issue at all.




[ Aside: lots of folks mention Wulf's book. To me it seems more like
nostalgia than relevance these days. Bliss on the PDP-11 doesn't do
much for me. But everyone has an old favorite like that. Mine is B.
Randell and L.J. Russell, ALGOL 60 Implementation -- it impressed me
when I first read it, but I'd probably be disappointed in it now. ]


jeff




Chris F Clark wrote:
  > I've seen more divergence in whether a call breaks a block or not.
  > . . .
  >
  > While I am not surprised you haven't found a book that covered
  > that aspect, I suspect that it isn't that it is never addressed in
  > compiler texts. For example, I would be willing to bet that Bob
  > Morgan's book covers it, as did Bill Wulf's book, and the book on the
  > VAX PL/I compiler. I would also be surprised if it wasn't addressed
  > somewhere in Steve Muchnick's book.






--


---------------------------------------------------------------------
= Jeff Kenton http://home.comcast.net/~jeffrey.kenton =
---------------------------------------------------------------------

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