Re: Compiler Correctness

Sid Touati <Sid.Touati@inria.fr> wrote:
>As example, suppose that the compiler executes on a 32 bits machine and
> generates a code for a 64 bits machine. If the compiler makes static
>constant propagation, it makes arithmetic computation on the 32 bit
>machine. This would not give you the same result if you generate the
>code on the 64 bits machines.


Any cross-compiler which doesn't emulate the *target* machine's arithmetic
when performing operations like constant propagation is simply broken.


Of course, if it does emulate, there is still the issue of whether it
emulates *correctly*. But this is just another case of "are there bugs in
the compiler?"; it's not a fundamental problem of missed assumptions.


>We can imagine more complex cases with floating point computations...


Emulating another machine's floating point is definitely a lot of work
(the widespread consensus on IEEE-standard floating point has helped but
doesn't entirely solve the problem). But if you signed on to write a
cross-compiler, well, you knew the job was dangerous when you took it. :-)
--
spsystems.net is temporarily off the air; | Henry Spencer
mail to henry at zoo.utoronto.ca instead. | henry@spsystems.net

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