low level optimization

[Michael Coffin]

In article <21660 at lanl.gov> jlg at cochiti.lanl.gov (Jim Giles) writes:
>The only way around this difficulty [detecting aliasing] is
>interprocedural analysis - but C requires separate compilation.

Where in heaven's name did you get the idea that separate compilation
precludes interprocedural analysis?  Ken Kennedy's group at Rice U.
have for years been doing extremely fancy interprocedural analysis of
Fortran programs and simultaneously providing separate compilation.
Take a look at the work done on PFC back in the late 70's and early
80's.  They operate under the same constraints that C demands---that
object code has to repect the semantics of the language, regardless of
what analysis is done.  (Note, by the way, that Fortran compiler
writers seem to find interprocedural analysis necessary for producing
efficient code from Fortran in spite of the rules against aliased
parameters.)

-mike