[Openmcl-devel] sharing data between fortran and OpenMCL

[Andrew P. Lentvorski, Jr.]

On Oct 1, 2004, at 8:52 PM, Hamilton Link wrote:
> 2. In languages with aliasing (i.e. all languages that you'd use), it 
> is IMPOSSIBLE in general short of GCing to know when something can be 
> freed.

What do you mean by "aliasing"?

I'm pretty sure that the Fortran standards actually make very explicit 
guarantees about "aliasing".  I seem to recall that Fortran guarantees 
that different references may not overlap in memory in any way without 
explicitly stating such.  Thus, Fortran can still optimize much more 
aggressively than most other languages.

-a