This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: tree_ssa_useless_type_conversions


In message <20031111085648.GA1729@kam.mff.cuni.cz>, Jan Hubicka writes:
 >I am having problem to think of how to do type checking in such a
 >relaxed typing system.  Gimple specifies that types must match.
I think we need to update the gimple documentation.  We have relaxed
the requirement that types must match exactly.

 >If I
 >write tree_ssa_types_compatible_p predicate that duplicate the code of
 >useless_type_conversion, I run into proplem that the function is not
 >transitive nor symmetric and thus we do not produce equivalence classes.
Yup.  There are cases when it can't be symmetric, particularly in regards
to casting pointer types.

 >We may change type twice, both times doing allowed useless conversion
 >and end up with non-useless conversion.
That would be a bug.  Specifics please (ie compilable code).

jeff


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]