low level optimization

[Wayne Throop]

> jlg at cochiti.lanl.gov (Jim Giles)
> In NO present implementation that does IM analysis
> is the method standard conforming.

To quote Inigo Montoya: "You keep using that word.  I do not
think it means what you think it means."

In fact an implementation that uses inter-module analysis to get
superior optimization of cases that can be proven non-aliased may very
well be standard conforming.  Certainly the inter-module optimizations
of the MIPS C compiler aren't what (if anything) keep it from being
standard conforming. 

In just what way are inter-module optimizations supposed to extend or
violate the standard?  As far as I can tell, they do neither. 

( Note that these inter-module optimizations can be performed within
  the usual capabilities of current Unix .o and .a file formats, at the
  cost of a tradeoff in multiple, differently-optimized translations
  output from the compile phase.  In much the same way as the inter-module
  type checking is done within this restriction by C++. 

  Not that there aren't better ways available if you allow the
  introduction of a better intermediate form.  It's just that this 
  conservative approach (while not totally satisfactory), would work 
  and yield 99 and 44/100ths percent of the benefits Jim is looking for.  )
--
Wayne Throop  ...!mcnc!dg-rtp!sheol!throopw