This is the mail archive of the gcc-patches@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: [PATCH][RFC] Step 1 to a middle-end type-system


On 6/25/07 11:59 AM, Richard Kenner wrote:
>> Only one thought I've had which is probably naive and not very useful,
>> what if the FE were to assign "type classes" with the very simple rule
>> that all the types with the same class are compatible?
> 
> I continue to be unsure why we'd want the FE to have any role
> whatsoever in middle-end type compatibility.  Why aren't the only

I keep interchanging "aliasing conflict" with "type compatibility".  In
my original message I was thinking about alias conflicts.  We do need
the FE to set up equivalence classes for TBAA, as only the front end
will know if pointers to 'MyType' can point to objects of type 'Integer'.

So, after chatting briefly with Dan, my understanding is that the plan
is to:

1- Use classes/sets for TBAA conflicts.

2- Use structural equivalence for type compatbility (bitwidth, size,
signedness, etc)


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