Re: Reconstructing the Types of Stack-Machine Codes

If it's of any interest to you, I once used a related technique at a
higher level in the compiler for a proprietary language which allowed
(encouraged; very strongly) highly ambiguous expressions.


The technique involved enumerating the various possible
interpretations of an expression; then abstractly "executing" the
stack-machine code using type information instead of values. If the
execution failed (type conflict; empty stack; etc.) or completed with
more or less than 1 item on the stack, then it wasn't a reasonable
interpretation of the expression.


The type information required is very simple (parameter types and
result type) but in the very ambiguous environment I was operating
(many identically named operators with variable numbers of sometimes
optional parameters) the technique suffered from exponential execution
time problems!


Mail me (andrewf@slhosiery.com.au) if you think I can be of any
assistance.


Andrew.


Oukseh Lee wrote:
>
> Hello,
>
> Whoever knows about TYPE ANALYSIS of STACK-MACHINE CODES?
> Please make me know. Thanks in advance.

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